Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Claim Rejections - 35 USC § 101
35 U.S.C. 101 reads as follows:
Whoever invents or discovers any new and useful process, machine, manufacture, or composition of matter, or any new and useful improvement thereof, may obtain a patent therefor, subject to the conditions and requirements of this title.
Claims 1-13 are rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more. The claim(s) recite(s) mental processes – concepts performed in the human mind, as well as mathematical concepts – specifically, mathematical calculations.
Subject Matter Eligibility Analysis
Step 1: Do the Claims Specify a Statutory Category?
Claims 1-13 recite methods, therefore satisfying Step 1 of the analysis.
Step 2 Analysis
Regarding claim 8,
Step 2A – Prong 1: Is a Judicial Exception Recited?
For step 2A eligibility prong one(does the claim recite a judicial exception?), the claim(s) recite(s) “training, by a computer, the MLM based on input data and a selected training algorithm to generate a trained MLM, wherein the selected training algorithm includes a fault detection and classification algorithm”(training a MLM using a selected training algorithm can contain mathematical calculations. The broadest reasonable interpretation of training the MLM includes training a gradient-boosting model(specification par 17), which contains using mathematical functions to train the MLM(NPL “gradient boosting” – Wikipedia June 2020 pg 3-5). The process following the training algorithm instructions is also a series of mental steps); limitation “detecting, by the trained MLM, at least a first anomalous pattern of traces in the multiplicity of time-series traces at a first time-based location in a stable region or a transition region during operation of the semiconductor process; defining, by the trained MLM, an anomaly window around the first time-based location and containing the first anomalous pattern; determining, by the trained MLM, a plurality of key features for portions of the multiplicity of time-series traces within the anomaly window; comparing, by the trained MLM, the plurality of determined key features with prior key features associated with prior anomalous patterns stored in a database; based on the comparison of key features, determining, by the trained MLM, a likelihood that the first anomalous pattern matches one of the prior anomalous patterns;”, “if the likelihood exceeds a threshold”, and “taking the corrective action in the semiconductor process to correct the root cause”. Defining a plurality of windows, detecting an anomalous pattern, defining a second window within the first window, determining key features, comparing key features with prior patterns in a database, determining a match with the prior patterns, comparing a likelihood to a threshold, and taking an action to correct the root cause are mental processes of observation, evaluation, judgment, opinion [see MPEP 2106.04(a)(2) III. “mental processes”]. Reference US 20200057689 A1(Farahat) teaches that predictive maintenance has historically been done manually (par 6 “In related art implementations, one of the main objectives of predictive maintenance is to prevent failures before they happen. Related art implementations attempt to prevent failures by monitoring the equipment and searching for any pre-failure patterns. Such related art monitoring processes are conducted manually through visual inspection of equipment, or through using monitoring tools such as vibration monitoring and other devices.”). As claimed, this process can practically be performed either in the human mind or using a computer as a tool.
Even if the limitations require a computer, it can still be a mental process [see MPEP 2106.04(a)(2) III. C. "A Claim That Requires a Computer May Still Recite a Mental Process"]. Defining windows, detecting anomalous patterns, determining key features, looking up features in a database for similar situations in the past, following the instructions stored in the database on how the issue was fixed in the past, are directed to mental processes because the steps are recited at a high level of generality and merely use computers as a tool to perform the processes.
Step 2A – Prong 2: Is the Judicial Exception Integrated into a Practical Application?
For step 2A eligibility prong two(does the claim recite additional elements that integrate the judicial exception into a practical application?), This judicial exception is not integrated into a practical application because the additional limitations of “receiving,… a multiplicity of time-series traces obtained from a corresponding multiplicity of semiconductor processing equipment sensors during steps of a semiconductor process;”, and “retrieving a root cause and a corrective action associated with the matched prior anomalous pattern” are insignificant extra-solution activities of data gathering, data sending, and presentation[see MPEP 2106.05(g) Whether the limitation amounts to necessary data gathering and outputting. This is considered in Step 2A Prong Two and Step 2B.]
The additional computer parts(a computer, semiconductor processing equipment sensors, database, semiconductor processing equipment, a machine learning model) are well known components recited at a high level of generality[see MPEP 2106.05(b) “If applicant amends a claim to add a generic computer or generic computer components and asserts that the claim recites significantly more because the generic computer is 'specially programmed' (as in Alappat, now considered superseded) or is a 'particular machine' (as in Bilski), the examiner should look at whether the added elements integrate the exception into a practical application or provide significantly more than the judicial exception. Merely adding a generic computer, generic computer components, or a programmed computer to perform generic computer functions does not automatically overcome an eligibility rejection. Alice Corp. Pty. Ltd. v. CLS Bank Int’l, 573 U.S. 208, 223-24, 110 USPQ2d 1976, 1983-84 (2014). See In re Alappat, 33 F.3d 1526, 1545, 31 USPQ2d 1545, 1558 (Fed. Cir. 1994); In re Bilski, 545 F.3d 943, 88 USPQ2d 1385 (Fed. Cir. 2008)”]. As a whole, the claims are directed to several abstract mental processes implemented on a generic computer, but are not integrated into a practical application[see MPEP 2106.05(f) “implementing an abstract idea on a generic computer, does not integrate the abstract idea into a practical application in Step 2A Prong Two”].
The claim’s “obtained from a corresponding multiplicity of semiconductor processing equipment sensors during steps of a semiconductor process” do not integrate the judicial exception into a practical application. The limitations are specified at a high level of generality, and does not meaningfully limit the claim by going beyond generally linking the use of the judicial exception to a particular technological environment(“semiconductor processing equipment”). The application specification paragraph 4 describes “Typically, an FDC method starts with breaking a complex trace into logical "windows" and then computing statistics (frequently called indicators or key numbers) on the trace data in the windows. The indicators can be monitored using statistical process control ("SPC") techniques to identify anomalies, based primarily on engineering knowledge, and the indicators can be utilized as inputs for predictive models and root cause analysis. … However, the analysis of the indicators for anomaly detection is still primarily univariate in nature, with anomalies considered on a feature by feature basis, and is generally insufficient to identify equipment fail modes related to the detected anomaly.”. The claims simply apply the abstract mental process to multiple sensors. Reference US 20210042570 A1 (Iskandar) describes multivariate analysis of sensor values from multiple sensors as a conventional system(par 14 “In other conventional systems, multivariate analysis is performed to receive sensor values from multiple sensors,…”) The claims generally link the abstract idea of anomaly detection and correction, to the field of semiconductor processing equipment. The same process except for the “semiconductor processing equipment” descriptor would also work for a heart pacemaker, cloud computing systems, a car status monitoring system(check engine light), an air conditioning system. [See MPEP 2106.04(d)(1) “Evaluating Improvements in the Functioning of a Computer, or an Improvement to Any Other Technology or Technical Field in Step 2A Prong Two” and also MPEP 2106.05(h)“Field of Use and Technological Environment”]
Step 2B: Do the Claims Provide an Inventive Concept?
For step 2B eligibility (Whether a Claim Amounts to Significantly More), The claim(s) does/do not include additional elements that are sufficient to amount to significantly more than the judicial exception because additional elements are either gathering/storing data(“receiving, into the trained MLM, a multiplicity of time-series traces obtained from a corresponding multiplicity of semiconductor processing equipment sensors during steps of a semiconductor process;”, and “retrieving a root cause and a corrective action associated with the matched prior anomalous pattern”), or are additional computer parts that are well known components recited at a high level of generality(“receiving a multiplicity of time-series traces obtained from a corresponding multiplicity of semiconductor processing equipment sensors during steps of a semiconductor process;”, and “patterns stored in a database”).
The data gathering/storing limitations are insignificant extra-solution activity because these limitations amount to necessary data gathering and outputting, (i.e., all uses of the recited judicial exception require such data gathering or data output) [see MPEP 2106.05(g) “(1) Whether the extra-solution limitation is well known. “, “(2) Whether the limitation is significant (i.e. it imposes meaningful limits on the claim such that it is not nominally or tangentially related to the invention).”, “(3) Whether the limitation amounts to necessary data gathering and outputting, (i.e., all uses of the recited judicial exception require such data gathering or data output).”]
The data gathering/storing limitations are also well-understood, routine, conventional computer functions, recited at a high level of generality functions as recognized by the court decisions listed in MPEP § 2106.05(d). The application specification paragraph 4 describes “Typically, an FDC method starts with breaking a complex trace into logical "windows" and then computing statistics (frequently called indicators or key numbers) on the trace data in the windows. The indicators can be monitored using statistical process control ("SPC") techniques to identify anomalies, based primarily on engineering knowledge, and the indicators can be utilized as inputs for predictive models and root cause analysis. … However, the analysis of the indicators for anomaly detection is still primarily univariate in nature, with anomalies considered on a feature by feature basis, and is generally insufficient to identify equipment fail modes related to the detected anomaly.”. The claims simply apply the abstract mental process to multiple sensors. The claims generally link the abstract idea to the field of semiconductor processing equipment. The same process except for the “semiconductor processing equipment” descriptor would also work for a pacemaker, cloud computing systems, a car warning system, an air conditioning system. [See MPEP 2106.05(h) “Field of Use and Technological Environment”] Automating a mental process and adding well-understood, routine, conventional activities previously known to the industry, specified at a high level of generality does not qualify as “significantly more” [see MPEP 2106.05 “Limitations that the courts have found not to be enough to qualify as "significantly more" when recited in a claim with a judicial exception include: … ii. Simply appending well-understood, routine, conventional activities previously known to the industry, specified at a high level of generality, to the judicial exception, e.g., a claim to an abstract idea requiring no more than a generic computer to perform generic computer functions that are well-understood, routine and conventional activities previously known to the industry, as discussed in Alice Corp., 573 U.S. at 225, 110 USPQ2d at 1984 (see MPEP § 2106.05(d));”].
Combined and considered as a whole, the claim describes a system that gathers sensor data, analyses it to find anomalies, compares the anomalies to a database with fixes for the anomalies, and then the system implements the fixes. This system comprises only well-understood, routine, conventional mental steps recited at a high level of generality[MPEP 2106.05 “ii Simply appending well-understood, routine, conventional activities previously known to the industry, specified at a high level of generality, to the judicial exception, e.g., a claim to an abstract idea requiring no more than a generic computer to perform generic computer functions that are well-understood, routine and conventional activities previously known to the industry, as discussed in Alice Corp., 573 U.S. at 225, 110 USPQ2d at 1984 (see MPEP § 2106.05(d));”], and insignificant extra-solution activity[see MPEP 2106.05(g) “(1) Whether the extra-solution limitation is well known. “, “(2) Whether the limitation is significant (i.e. it imposes meaningful limits on the claim such that it is not nominally or tangentially related to the invention).”, “(3) Whether the limitation amounts to necessary data gathering and outputting, (i.e., all uses of the recited judicial exception require such data gathering or data output).”]. Aside from the limitation that states that the sensors are “semiconductor processing equipment sensors”, and “taking the corrective action in the semiconductor process” there are no other steps or limitations that tie the claims to semiconductor manufacturing. If you replace “semiconductor process” with a different field, the claims could apply to any system that gathers sensor data for analysis and corrective actions. [see MPEP 2106.05(h) “Examples of limitations that the courts have described as merely indicating a field of use or technological environment in which to apply a judicial exception include: … iv. Specifying that the abstract idea of monitoring audit log data relates to transactions or activities that are executed in a computer environment, because this requirement merely limits the claims to the computer field, i.e., to execution on a generic computer, FairWarning v. Iatric Sys., 839 F.3d 1089, 1094-95, 120 USPQ2d 1293, 1295 (Fed. Cir. 2016);”].
Conclusion: In light of the above, the limitations in claim 8 recite and are directed to an abstract idea and recite no additional elements that would amount to significantly more than the identified abstract idea. Claim 8 is therefore not patent eligible.
Regarding claim 9, limitations “calculating a plurality of statistical indicators for portions of the traces contained within the anomaly window” (this is a mental process of observation, evaluation, judgment, opinion [MPEP 2106.04(a)(2) III. “Mental processes”]). Calculating a plurality of statistical indicators is also a mathematical calculation type of abstract idea as well as a mental process [MPEP 2106.04(a)(2) “Mathematical concepts”]. As claimed, this process can practically be performed either in the human mind or using a computer as a tool.
Regarding claim 10, limitations “identifying a plurality of wafer attributes for the semiconductor process” (this is a mental process of observation, evaluation, judgment, opinion [MPEP 2106.04(a)(2) III. “Mental processes”]). As claimed, this process can practically be performed either in the human mind or using a computer as a tool.
The additional computer parts(wafer) are well known components recited at a high level of generality[see MPEP 2106.05(b) “If applicant amends a claim to add a generic computer or generic computer components and asserts that the claim recites significantly more because the generic computer is 'specially programmed' (as in Alappat, now considered superseded) or is a 'particular machine' (as in Bilski), the examiner should look at whether the added elements integrate the exception into a practical application or provide significantly more than the judicial exception. Merely adding a generic computer, generic computer components, or a programmed computer to perform generic computer functions does not automatically overcome an eligibility rejection. Alice Corp. Pty. Ltd. v. CLS Bank Int’l, 573 U.S. 208, 223-24, 110 USPQ2d 1976, 1983-84 (2014). See In re Alappat, 33 F.3d 1526, 1545, 31 USPQ2d 1545, 1558 (Fed. Cir. 1994); In re Bilski, 545 F.3d 943, 88 USPQ2d 1385 (Fed. Cir. 2008)”]. As a whole, the claims are directed to several abstract mental processes implemented on a generic computer, but are not integrated into a practical application [see MPEP 2106.05(f) “implementing an abstract idea on a generic computer, does not integrate the abstract idea into a practical application in Step 2A Prong Two”].
Regarding claim 11, limitation “performing feature engineering to select a subset of the key features determined to be important to detecting and identifying the first anomalous pattern.” (this is a mental process of observation, evaluation, judgment, opinion [MPEP 2106.04(a)(2) III. “Mental processes”]). As claimed, feature engineering is generic enough that it could also possibly include mathematical calculations as regression analysis algorithms are a common statistical modeling technique used for estimating the relationships between a dependent variable and one or more independent variables(Regression analysis - Wikipedia July 2020). As claimed, this process can practically be performed either in the human mind or using a computer as a tool.
Regarding claim 12, limitation “the step of detecting an anomalous pattern further comprising detecting a rapid change in values of the traces in the stable region or the transition region of operation.” (this is a mental process of observation, evaluation, judgment, opinion [MPEP 2106.04(a)(2) III. “Mental processes”]). Calculating change over time also contains mathematical calculations. As claimed, this process can practically be performed either in the human mind or using a computer as a tool.
Regarding claim 13, limitation “the step of detecting the first anomalous pattern further comprising detecting a rapid change in the rate of change for values of the traces in the stable region or the transition region of operation.” (this is a mental process of observation, evaluation, judgment, opinion [MPEP 2106.04(a)(2) III. “Mental processes”]). Calculating rate of change and differences in rates of change over time are also mathematical calculations. As claimed, this process can practically be performed either in the human mind or using a computer as a tool.
Regarding claim 1,
Step 2A – Prong 1: Is a Judicial Exception Recited?
For step 2A eligibility prong one(does the claim recite a judicial exception?), the claim(s) recite(s) “training, by a computer, a machine learning model (MLM) based on input data and a selected training algorithm to generate a trained MLM, wherein the selected training algorithm includes a fault detection and classification algorithm”(training a MLM using a selected training algorithm can contain mathematical calculations. The broadest reasonable interpretation of training the MLM includes training a gradient-boosting model(specification par 17), which contains using mathematical functions to train the MLM(NPL “gradient boosting” – Wikipedia June 2020 pg 3-5). The process of following the training algorithm instructions is also a series of mental steps); limitation “detecting, by the trained MLM, an anomalous pattern of traces in a stable region or a transition region of process operation at a first time-based location in the multiplicity of time-series traces; defining, by the trained MLM, an anomaly window around the first time-based location and containing the anomalous pattern; determining, by the trained MLM, a plurality of key features for the anomalous pattern of traces contained within the anomaly window; comparing, by the trained MLM, the plurality of determined key features with prior key features associated with prior anomalous patterns; determining that the anomalous pattern matches one of the prior anomalous patterns;”, and “taking the corrective action in the semiconductor process to correct the root cause for the anomalous pattern”. Defining a plurality of windows, detecting an anomalous pattern, defining a second window within the first window, determining key features, comparing key features with prior patterns, determining a match with the prior patterns, and taking an action to correct the root cause are mental processes of observation, evaluation, judgment, opinion [see MPEP 2106.04(a)(2) III. “mental processes”]. Reference US 20200057689 A1(Farahat) teaches that predictive maintenance has historically been done manually (par 6 “In related art implementations, one of the main objectives of predictive maintenance is to prevent failures before they happen. Related art implementations attempt to prevent failures by monitoring the equipment and searching for any pre-failure patterns. Such related art monitoring processes are conducted manually through visual inspection of equipment, or through using monitoring tools such as vibration monitoring and other devices.”). As claimed, this process can practically be performed either in the human mind or using a computer as a tool.
Even if the limitations require a computer, it can still be a mental process [see MPEP 2106.04(a)(2) III. C. "A Claim That Requires a Computer May Still Recite a Mental Process"]. Defining windows, detecting anomalous patterns, determining key features, looking up features in a database for similar situations in the past, following the instructions stored in the database on how the issue was fixed in the past, are directed to mental processes because the steps are recited at a high level of generality and merely use computers as a tool to perform the processes.
Step 2A – Prong 2: Is the Judicial Exception Integrated into a Practical Application?
For step 2A eligibility prong two(does the claim recite additional elements that integrate the judicial exception into a practical application?), This judicial exception is not integrated into a practical application because the additional limitations of “receiving,… a multiplicity of time-series traces obtained from a corresponding multiplicity of equipment sensors during steps of a semiconductor process;”, and “retrieving a root cause and a corrective action associated with the matched prior anomalous pattern” are insignificant extra-solution activities of data gathering, data sending, and presentation[see MPEP 2106.05(g) Whether the limitation amounts to necessary data gathering and outputting. This is considered in Step 2A Prong Two and Step 2B.]
The additional computer parts(a computer, semiconductor processing equipment sensors, semiconductor processing equipment, a machine learning model) are well known components recited at a high level of generality[see MPEP 2106.05(b) “If applicant amends a claim to add a generic computer or generic computer components and asserts that the claim recites significantly more because the generic computer is 'specially programmed' (as in Alappat, now considered superseded) or is a 'particular machine' (as in Bilski), the examiner should look at whether the added elements integrate the exception into a practical application or provide significantly more than the judicial exception. Merely adding a generic computer, generic computer components, or a programmed computer to perform generic computer functions does not automatically overcome an eligibility rejection. Alice Corp. Pty. Ltd. v. CLS Bank Int’l, 573 U.S. 208, 223-24, 110 USPQ2d 1976, 1983-84 (2014). See In re Alappat, 33 F.3d 1526, 1545, 31 USPQ2d 1545, 1558 (Fed. Cir. 1994); In re Bilski, 545 F.3d 943, 88 USPQ2d 1385 (Fed. Cir. 2008)”]. As a whole, the claims are directed to several abstract mental processes implemented on a generic computer, but are not integrated into a practical application[see MPEP 2106.05(f) “implementing an abstract idea on a generic computer, does not integrate the abstract idea into a practical application in Step 2A Prong Two”].
The claim’s “obtained from a corresponding multiplicity of semiconductor processing equipment sensors during steps of a semiconductor process” do not integrate the judicial exception into a practical application. The limitations are specified at a high level of generality, and does not meaningfully limit the claim by going beyond generally linking the use of the judicial exception to a particular technological environment(“semiconductor processing equipment”). The application specification paragraph 4 describes “Typically, an FDC method starts with breaking a complex trace into logical "windows" and then computing statistics (frequently called indicators or key numbers) on the trace data in the windows. The indicators can be monitored using statistical process control ("SPC") techniques to identify anomalies, based primarily on engineering knowledge, and the indicators can be utilized as inputs for predictive models and root cause analysis. … However, the analysis of the indicators for anomaly detection is still primarily univariate in nature, with anomalies considered on a feature by feature basis, and is generally insufficient to identify equipment fail modes related to the detected anomaly.”. The claims simply apply the abstract mental process to multiple sensors. Reference US 20210042570 A1 (Iskandar) describes multivariate analysis of sensor values from multiple sensors as a conventional system(par 14 “In other conventional systems, multivariate analysis is performed to receive sensor values from multiple sensors,…”) The claims generally link the abstract idea of anomaly detection and correction, to the field of semiconductor processing equipment. The same process except for the “semiconductor processing equipment” descriptor would also work for a heart pacemaker, cloud computing systems, a car status monitoring system(check engine light), an air conditioning system. [See MPEP 2106.04(d)(1) “Evaluating Improvements in the Functioning of a Computer, or an Improvement to Any Other Technology or Technical Field in Step 2A Prong Two” and also MPEP 2106.05(h)“Field of Use and Technological Environment”]
Step 2B: Do the Claims Provide an Inventive Concept?
For step 2B eligibility (Whether a Claim Amounts to Significantly More), The claim(s) does/do not include additional elements that are sufficient to amount to significantly more than the judicial exception because additional elements are either gathering/storing data(“receiving, into the trained MLM, a multiplicity of time-series traces obtained from a corresponding multiplicity of semiconductor processing equipment sensors during steps of a semiconductor process;”, and “retrieving a root cause and a corrective action associated with the matched prior anomalous pattern”), or are additional computer parts that are well known components recited at a high level of generality(“receiving a multiplicity of time-series traces obtained from a corresponding multiplicity of semiconductor processing equipment sensors during steps of a semiconductor process;”, and “training, by a computer, a machine learning model (MLM)”).
The data gathering/storing limitations are insignificant extra-solution activity because these limitations amount to necessary data gathering and outputting, (i.e., all uses of the recited judicial exception require such data gathering or data output) [see MPEP 2106.05(g) “(1) Whether the extra-solution limitation is well known. “, “(2) Whether the limitation is significant (i.e. it imposes meaningful limits on the claim such that it is not nominally or tangentially related to the invention).”, “(3) Whether the limitation amounts to necessary data gathering and outputting, (i.e., all uses of the recited judicial exception require such data gathering or data output).”]
The data gathering/storing limitations are also well-understood, routine, conventional computer functions, recited at a high level of generality functions as recognized by the court decisions listed in MPEP § 2106.05(d). The application specification paragraph 4 describes “Typically, an FDC method starts with breaking a complex trace into logical "windows" and then computing statistics (frequently called indicators or key numbers) on the trace data in the windows. The indicators can be monitored using statistical process control ("SPC") techniques to identify anomalies, based primarily on engineering knowledge, and the indicators can be utilized as inputs for predictive models and root cause analysis. … However, the analysis of the indicators for anomaly detection is still primarily univariate in nature, with anomalies considered on a feature by feature basis, and is generally insufficient to identify equipment fail modes related to the detected anomaly.”. The claims simply apply the abstract mental process to multiple sensors. The claims generally link the abstract idea to the field of semiconductor processing equipment. The same process except for the “semiconductor processing equipment” descriptor would also work for a pacemaker, cloud computing systems, a car warning system, an air conditioning system. [See MPEP 2106.05(h) “Field of Use and Technological Environment”] Automating a mental process and adding well-understood, routine, conventional activities previously known to the industry, specified at a high level of generality does not qualify as “significantly more” [see MPEP 2106.05 “Limitations that the courts have found not to be enough to qualify as "significantly more" when recited in a claim with a judicial exception include: … ii. Simply appending well-understood, routine, conventional activities previously known to the industry, specified at a high level of generality, to the judicial exception, e.g., a claim to an abstract idea requiring no more than a generic computer to perform generic computer functions that are well-understood, routine and conventional activities previously known to the industry, as discussed in Alice Corp., 573 U.S. at 225, 110 USPQ2d at 1984 (see MPEP § 2106.05(d));”].
Combined and considered as a whole, the claim describes a system that gathers sensor data, analyses it to find anomalies, compares the anomalies to a database with fixes for the anomalies, and then the system implements the fixes. This system comprises only well-understood, routine, conventional mental steps recited at a high level of generality[MPEP 2106.05 “ii Simply appending well-understood, routine, conventional activities previously known to the industry, specified at a high level of generality, to the judicial exception, e.g., a claim to an abstract idea requiring no more than a generic computer to perform generic computer functions that are well-understood, routine and conventional activities previously known to the industry, as discussed in Alice Corp., 573 U.S. at 225, 110 USPQ2d at 1984 (see MPEP § 2106.05(d));”], and insignificant extra-solution activity[see MPEP 2106.05(g) “(1) Whether the extra-solution limitation is well known. “, “(2) Whether the limitation is significant (i.e. it imposes meaningful limits on the claim such that it is not nominally or tangentially related to the invention).”, “(3) Whether the limitation amounts to necessary data gathering and outputting, (i.e., all uses of the recited judicial exception require such data gathering or data output).”]. Aside from the limitation that states that the sensors are “semiconductor processing equipment sensors”, and “taking the corrective action in the semiconductor process” there are no other steps or limitations that tie the claims to semiconductor manufacturing. If you replace “semiconductor process” with a different field, the claims could apply to any system that gathers sensor data for analysis and corrective actions. [see MPEP 2106.05(h) “Examples of limitations that the courts have described as merely indicating a field of use or technological environment in which to apply a judicial exception include: … iv. Specifying that the abstract idea of monitoring audit log data relates to transactions or activities that are executed in a computer environment, because this requirement merely limits the claims to the computer field, i.e., to execution on a generic computer, FairWarning v. Iatric Sys., 839 F.3d 1089, 1094-95, 120 USPQ2d 1293, 1295 (Fed. Cir. 2016);”].
Conclusion: In light of the above, the limitations in claim 1 recite and are directed to an abstract idea and recite no additional elements that would amount to significantly more than the identified abstract idea. Claim 1 is therefore not patent eligible.
Regarding claims 2-6 they are the same as claims 9-13 and are rejected for the same reasons.
Regarding claim 20,
Step 2A – Prong 1: Is a Judicial Exception Recited?
For step 2A eligibility prong one(does the claim recite a judicial exception?), the claim(s) recite(s) “training, by a computer, the MLM based on input data and a selected training algorithm to generate a trained MLM, wherein the selected training algorithm includes a fault detection and classification algorithm”(training a MLM using a selected training algorithm can contain mathematical calculations. The broadest reasonable interpretation of training the MLM includes training a gradient-boosting model(specification par 17), which contains using mathematical functions to train the MLM(NPL “gradient boosting” – Wikipedia June 2020 pg 3-5). The process following the training algorithm instructions is also a series of mental steps); limitation “detecting one or more anomalous patterns of time-series traces in the multiplicity of timeseries traces at a time-based location in a stable region of process operation using the trained MLM; defining a window around the first time-based location and containing at least a first anomalous pattern of the one or more anomalous patterns using the trained MLM; calculating a plurality of statistical indicators for the first anomalous pattern of traces contained within the window; determining that the first anomalous pattern is a likely match with at least one prior anomalous pattern using the MLM; based on the comparison of key features,”, and “taking the corrective action in the semiconductor process”. Defining a plurality of windows, detecting an anomalous pattern, defining a second window within the first window, determining key statistical indicators, comparing key statistical indicators with prior patterns in a database, determining a match with the prior patterns, and taking an action to correct the root cause are mental processes of observation, evaluation, judgment, opinion [see MPEP 2106.04(a)(2) III. “mental processes”]. Reference US 20200057689 A1(Farahat) teaches that predictive maintenance has historically been done manually (par 6 “In related art implementations, one of the main objectives of predictive maintenance is to prevent failures before they happen. Related art implementations attempt to prevent failures by monitoring the equipment and searching for any pre-failure patterns. Such related art monitoring processes are conducted manually through visual inspection of equipment, or through using monitoring tools such as vibration monitoring and other devices.”). As claimed, this process can practically be performed either in the human mind or using a computer as a tool. Under broadest reasonable interpretation, the process of “calculating a plurality of statistical indicators for the first anomalous pattern”, also contains statistical functions which is also considered a mathematical calculation [see MPEP 2106.04(a)(2) 1. “Mathematical concepts”])
Even if the limitations require a computer, it can still be a mental process [see MPEP 2106.04(a)(2) III. C. "A Claim That Requires a Computer May Still Recite a Mental Process"]. Defining windows, detecting anomalous patterns, determining key features, looking up features in a database for similar situations in the past, following the instructions stored in the database on how the issue was fixed in the past, are directed to mental processes because the steps are recited at a high level of generality and merely use computers as a tool to perform the processes.
Step 2A – Prong 2: Is the Judicial Exception Integrated into a Practical Application?
For step 2A eligibility prong two(does the claim recite additional elements that integrate the judicial exception into a practical application?), This judicial exception is not integrated into a practical application because the additional limitations of “receiving,… a multiplicity of time-series traces obtained from a corresponding multiplicity of semiconductor processing equipment sensors during steps of a semiconductor process;”, and “retrieving a root cause and a corrective action associated with the prior anomalous pattern” are insignificant extra-solution activities of data gathering, data sending, and presentation[see MPEP 2106.05(g) Whether the limitation amounts to necessary data gathering and outputting. This is considered in Step 2A Prong Two and Step 2B.]
The additional computer parts(a computer, semiconductor processing equipment sensors, semiconductor processing equipment, a machine learning model) are well known components recited at a high level of generality[see MPEP 2106.05(b) “If applicant amends a claim to add a generic computer or generic computer components and asserts that the claim recites significantly more because the generic computer is 'specially programmed' (as in Alappat, now considered superseded) or is a 'particular machine' (as in Bilski), the examiner should look at whether the added elements integrate the exception into a practical application or provide significantly more than the judicial exception. Merely adding a generic computer, generic computer components, or a programmed computer to perform generic computer functions does not automatically overcome an eligibility rejection. Alice Corp. Pty. Ltd. v. CLS Bank Int’l, 573 U.S. 208, 223-24, 110 USPQ2d 1976, 1983-84 (2014). See In re Alappat, 33 F.3d 1526, 1545, 31 USPQ2d 1545, 1558 (Fed. Cir. 1994); In re Bilski, 545 F.3d 943, 88 USPQ2d 1385 (Fed. Cir. 2008)”]. As a whole, the claims are directed to several abstract mental processes implemented on a generic computer, but are not integrated into a practical application[see MPEP 2106.05(f) “implementing an abstract idea on a generic computer, does not integrate the abstract idea into a practical application in Step 2A Prong Two”].
The claim’s “obtained from a corresponding multiplicity of semiconductor processing equipment sensors during steps of a semiconductor process” do not integrate the judicial exception into a practical application. The limitations are specified at a high level of generality, and does not meaningfully limit the claim by going beyond generally linking the use of the judicial exception to a particular technological environment(“semiconductor processing equipment”). The application specification paragraph 4 describes “Typically, an FDC method starts with breaking a complex trace into logical "windows" and then computing statistics (frequently called indicators or key numbers) on the trace data in the windows. The indicators can be monitored using statistical process control ("SPC") techniques to identify anomalies, based primarily on engineering knowledge, and the indicators can be utilized as inputs for predictive models and root cause analysis. … However, the analysis of the indicators for anomaly detection is still primarily univariate in nature, with anomalies considered on a feature by feature basis, and is generally insufficient to identify equipment fail modes related to the detected anomaly.”. The claims simply apply the abstract mental process to multiple sensors. Reference US 20210042570 A1 (Iskandar) describes multivariate analysis of sensor values from multiple sensors as a conventional system(par 14 “In other conventional systems, multivariate analysis is performed to receive sensor values from multiple sensors,…”) The claims generally link the abstract idea of anomaly detection and correction, to the field of semiconductor processing equipment. The same process except for the “semiconductor processing equipment” descriptor would also work for a heart pacemaker, cloud computing systems, a car status monitoring system(check engine light), an air conditioning system. [See MPEP 2106.04(d)(1) “Evaluating Improvements in the Functioning of a Computer, or an Improvement to Any Other Technology or Technical Field in Step 2A Prong Two” and also MPEP 2106.05(h)“Field of Use and Technological Environment”]
Step 2B: Do the Claims Provide an Inventive Concept?
For step 2B eligibility (Whether a Claim Amounts to Significantly More), The claim(s) does/do not include additional elements that are sufficient to amount to significantly more than the judicial exception because additional elements are either gathering/storing data(“receiving, into the trained MLM, a multiplicity of time-series traces obtained from a corresponding multiplicity of semiconductor processing equipment sensors during steps of a semiconductor process;”, and “retrieving a root cause and a corrective action associated with the prior anomalous pattern”), or are additional computer parts that are well known components recited at a high level of generality(“receiving, into the trained MLM, a multiplicity of time-series traces obtained from a corresponding multiplicity of semiconductor processing equipment sensors during steps of a semiconductor process;”).
The data gathering/storing limitations are insignificant extra-solution activity because these limitations amount to necessary data gathering and outputting, (i.e., all uses of the recited judicial exception require such data gathering or data output) [see MPEP 2106.05(g) “(1) Whether the extra-solution limitation is well known. “, “(2) Whether the limitation is significant (i.e. it imposes meaningful limits on the claim such that it is not nominally or tangentially related to the invention).”, “(3) Whether the limitation amounts to necessary data gathering and outputting, (i.e., all uses of the recited judicial exception require such data gathering or data output).”]
The data gathering/storing limitations are also well-understood, routine, conventional computer functions, recited at a high level of generality functions as recognized by the court decisions listed in MPEP § 2106.05(d). The application specification paragraph 4 describes “Typically, an FDC method starts with breaking a complex trace into logical "windows" and then computing statistics (frequently called indicators or key numbers) on the trace data in the windows. The indicators can be monitored using statistical process control ("SPC") techniques to identify anomalies, based primarily on engineering knowledge, and the indicators can be utilized as inputs for predictive models and root cause analysis. … However, the analysis of the indicators for anomaly detection is still primarily univariate in nature, with anomalies considered on a feature by feature basis, and is generally insufficient to identify equipment fail modes related to the detected anomaly.”. The claims simply apply the abstract mental process to multiple sensors. The claims generally link the abstract idea to the field of semiconductor processing equipment. The same process except for the “semiconductor processing equipment” descriptor would also work for a pacemaker, cloud computing systems, a car warning system, an air conditioning system. [See MPEP 2106.05(h) “Field of Use and Technological Environment”] Automating a mental process and adding well-understood, routine, conventional activities previously known to the industry, specified at a high level of generality does not qualify as “significantly more” [see MPEP 2106.05 “Limitations that the courts have found not to be enough to qualify as "significantly more" when recited in a claim with a judicial exception include: … ii. Simply appending well-understood, routine, conventional activities previously known to the industry, specified at a high level of generality, to the judicial exception, e.g., a claim to an abstract idea requiring no more than a generic computer to perform generic computer functions that are well-understood, routine and conventional activities previously known to the industry, as discussed in Alice Corp., 573 U.S. at 225, 110 USPQ2d at 1984 (see MPEP § 2106.05(d));”].
Combined and considered as a whole, the claim describes a system that gathers sensor data, analyses it to find anomalies, compares the anomalies to a database with fixes for the anomalies, and then the system implements the fixes. This system comprises only well-understood, routine, conventional mental steps recited at a high level of generality[MPEP 2106.05 “ii Simply appending well-understood, routine, conventional activities previously known to the industry, specified at a high level of generality, to the judicial exception, e.g., a claim to an abstract idea requiring no more than a generic computer to perform generic computer functions that are well-understood, routine and conventional activities previously known to the industry, as discussed in Alice Corp., 573 U.S. at 225, 110 USPQ2d at 1984 (see MPEP § 2106.05(d));”], and insignificant extra-solution activity[see MPEP 2106.05(g) “(1) Whether the extra-solution limitation is well known. “, “(2) Whether the limitation is significant (i.e. it imposes meaningful limits on the claim such that it is not nominally or tangentially related to the invention).”, “(3) Whether the limitation amounts to necessary data gathering and outputting, (i.e., all uses of the recited judicial exception require such data gathering or data output).”]. Aside from the limitation that states that the sensors are “semiconductor processing equipment sensors”, and “taking the corrective action in the semiconductor process” there are no other steps or limitations that tie the claims to semiconductor manufacturing. If you replace “semiconductor process” with a different field, the claims could apply to any system that gathers sensor data for analysis and corrective actions. [see MPEP 2106.05(h) “Examples of limitations that the courts have described as merely indicating a field of use or technological environment in which to apply a judicial exception include: … iv. Specifying that the abstract idea of monitoring audit log data relates to transactions or activities that are executed in a computer environment, because this requirement merely limits the claims to the computer field, i.e., to execution on a generic computer, FairWarning v. Iatric Sys., 839 F.3d 1089, 1094-95, 120 USPQ2d 1293, 1295 (Fed. Cir. 2016);”].
Conclusion: In light of the above, the limitations in claim 20 recite and are directed to an abstract idea and recite no additional elements that would amount to significantly more than the identified abstract idea. Claim 20 is therefore not patent eligible.
Claim Rejections - 35 USC § 103
In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis (i.e., changing from AIA to pre-AIA ) for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
Claim(s) 1-6,8-13,20 is/are rejected under 35 U.S.C. 103 as being unpatentable over by US 20210042570 A1 (Iskandar) in view of US 20200057689 A1 (Farahat).
Regarding claim 8, Iskandar teaches
A method of using a machine learning model (MLM) for predicting semiconductor processing equipment failure and taking corrective action,(par 4 “The method further includes processing the current trace data to identify a plurality of features of the current trace data and providing the plurality of features of the current trace data as input to a trained machine learning model that uses a hyperplane limit for product classification. The method further includes obtaining, from the trained machine learning model, one or more outputs indicative of predictive data associated with the hyperplane limit. A corrective action associated with the manufacturing equipment is to be performed based on the predictive data.”) comprising:
training, by a computer, the MLM based on input data and a selected training algorithm to generate a trained MLM(fig 4A:408; par 93 “At block 408, the processing logic trains a machine learning model using training data including the features (e.g., selected feature parameters of the features) of the historical trace data and the product data to generate a trained machine learning model.”), wherein the selected training algorithm includes a fault detection and classification algorithm;(fig 6A-C; par 19 “In some embodiments, the processing logic may generate ( e.g., based on the predictive data, as part of the corrective action) a plot that has two axes (e.g., corresponding to two features) and different regions ( e.g., normal, abnormal, and gray regions) indicated by an FDC limit (see FIGS. 6B-C). The normal region may be on a first side of the FDC limit, the abnormal region may be on a second side of the FDC limit, and the gray region may be within the FDC limit.” FDC stands for “Fault Detection and Classification”(par 14 “Described herein are technologies directed to automatic and adaptive fault detection and classification (FDC) limits.”))
receiving, into the trained MLM, a multiplicity(par 25 “The present disclosure may automatically generate FDC limits for sensor values from multiple sensors and take into account the interactions between the sensors ( e.g., via training and using the machine learning model) which provides more accurate results than conventional systems that compare sensor values from a single sensor or output of a set algorithm to a set limit.”) of time-series traces obtained from a corresponding multiplicity of semiconductor processing equipment sensors during steps of a semiconductor process(par 28 “The sensors 126 may provide sensor data 142 associated with manufacturing equipment 124 (e.g., associated with producing, by manufacturing equipment 124, corresponding products, such as wafers).” fig 4C:440; par 107 “Referring to FIG. 4C, at block 440, the processing logic receives current trace data including current sensor values associated with producing, by manufacturing equipment, second products.”; par 67 “Although embodiments of the disclosure are discussed in terms of generating predictive data 169 to perform a corrective action in manufacturing facilities (e.g., semiconductor manufacturing facilities), embodiments may also be generally applied to generating limits to perform an action.”);
detecting, by the trained MLM at least a first anomalous pattern of traces in the multiplicity of time-series traces at a first time-based location in a stable region or a transition region during operation of the semiconductor process;( Par 102 “At block 422, the processing logic segments (e.g., windows), via a sliding window, the historical trace data to generate a boundary between steady state and transient segments (e.g., to generate segmented trace data).”)
identifying, by the trained MLM, an anomaly window while sliding around the first time-based location and containing the first anomalous pattern; (Par 102 “At block 422, the processing logic segments (e.g., windows), via a sliding window, the historical trace data to generate a boundary between steady state and transient segments (e.g., to generate segmented trace data).”; par 149 “The sensor data values and/or feature patterns may be used to generate a hyperplane limit (e.g., FIG. 4A) and/or a plot of the FDC limit and regions ( e.g., normal region, abnormal region, gray region) (e.g., FIG. 4C).”; par 72 “In some embodiments, the data set generator 272 may discretize (e.g., segment) one or more of the data input 210 or the target output 220 (e.g., to use in classification algorithms for regression problems). Discretization (e.g., segmentation via a sliding window) of the data input 210 or target output 220 may transform continuous values of variables into discrete values.”; par 97 “The trained machine learning model may provide multivariate limit optimization ( e.g., via sliding window, incremental learning, filtering mechanism, etc.)”
determining, by the trained MLM, a plurality of key features for portions of the multiplicity of time-series traces within the anomaly window(par 53 “In some embodiments, the predictive server 112 (e.g., predictive component 114) may generate features ( e.g., historical features 148, current features 154) from trace data (e.g., historical trace data 144, current trace data 150) and store the features in the data store. In some embodiments, the features are a pattern in the trace data (e.g., slope, width, height, peak, etc.) or a combination of sensor values from the trace data ( e.g., power derived from voltage and current, etc.).”);
comparing, by the trained MLM, the plurality of determined key features with prior key features associated with prior anomalous patterns stored in a database;(par 48 “In some embodiments, the predictive component 114 may use a trained machine learning model 190 to determine the output for performing the corrective action based on the current features 154. The trained machine learning model 190 may be trained using the historical features 148 and historical product data 158 to learn key process and hardware parameters.”)
based on the comparison of key features, determining, by the trained MLM, a likelihood that the first anomalous pattern matches one of the prior anomalous patterns;(par 59 “Predictive component 114 may provide current features 154 to the trained machine learning model 190 and may run the trained machine learning model 190 on the input to obtain one or more outputs. … The predictive component 114 or corrective action component 122 may use the confidence data to decide whether to cause a corrective action associated with the manufacturing equipment 124 based on the predictive data 169.”)
retrieving a root cause and a corrective action associated with the matched prior anomalous pattern(par 24 “A corrective action associated with the manufacturing equipment may be performed based on the predictive data.” Par 18 “The processing device may predict, based on the FDC limit, one or more causes of classification within the products (e.g., causes of abnormal wafers) so that a corrective action associated with the manufacturing equipment can be performed.”) if the likelihood exceeds a threshold(par 59 “The predictive component 114 or corrective action component 122 may use the confidence data to decide whether to cause a corrective action associated with the manufacturing equipment 124 based on the predictive data 169.”); and
taking the corrective action to correct the root cause in the semiconductor process.(par 82 “At block 320, system 300 uses the trained model (e.g., selected model 308) to receive current features 354 (e.g., current features 154 of FIG. 1) and determines (e.g., extracts), from the output of the trained model, predictive data 369 (e.g., predictive data 169 of FIG. 1, FDC limit) to perform corrective actions associated with the manufacturing equipment 124.”)
However, Iskandar does not specifically teach defining, by the trained MLM, an anomaly window within the first window around the first time-based location and containing the first anomalous pattern.
On the other hand, Farahat teaches,
A method of using a machine learning model (MLM) for predicting general equipment failure and taking corrective action,(par 8 “Example implementations described herein involve a system for maintenance recommendation based on data-driven failure prediction.”) comprising:
training, by a computer, the MLM based on input data and a selected training algorithm to generate a trained MLM, wherein the selected training algorithm includes a fault detection and classification algorithm;(fig 1(b):110; par 43 “The first pipeline represents the model training 110, … . Model training 110 will execute a process similar to that illustrated in FIG. 2(a) to perform a preparation of data 111 and generate parameter values that is utilized in the data preparation 121 of batch processing 120 and data preparation 131 of stream processing 130.” Par 40 “The extracted windows are then used for training the failure prediction model.”)
receiving, into the trained MLM, a multiplicity of time-series traces obtained from a corresponding multiplicity of semiconductor processing equipment sensors during steps of a semiconductor process;(fig 1(b):101; par 39 “The maintenance recommendation system 102 can intake raw sensor, event and failure data 101 for processing by the following modules.”)
detecting, by the trained MLM, at least a first anomalous pattern of traces in the multiplicity of time-series traces at a first time-based location in a stable region or a transition region during operation of the semiconductor process;(fig 1(a):104; par 40 “Evidence window extraction 104 is responsible for extracting windows of sensor measurements and events from historical data which correspond to failure and normal cases. … . The extracted windows will be inspected by the failure prediction model for any signs of failures.”; par 56 “After the data is processed, the next procedure in failure prediction is to extract windows of data which are believed to contain pre-failure patterns.” )
defining, by the trained MLM, an anomaly window around the first time-based location and containing the first anomalous pattern; (fig 5(a); par 56 “After the data is processed, the next procedure in failure prediction is to extract windows of data which are believed to contain pre-failure patterns. After that, features are extracted from these windows, and these features are applied to the failure prediction module…”; par 57 “For each past failure instance, windows of observations are extracted. The windows can include the Alert Window ([ta, 1t], ta =t-W J, which is the period of time just before the failure during which countermeasures are required to prevent the failure from happening.”; par 58 “The windows can further include the Evidence Window … which is the period of time before the alert window during which the pre-failure patterns happen. The evidence time windows are considered as the positive class for the failure prediction model.”)
determining, by the trained MLM, a plurality of key features for portions of the multiplicity of time-series traces within the anomaly window;(par 41 “Feature extraction 105 extracts features from each window of measurements. The extracted features are then provided to the failure prediction model for training/application.”)
comparing, by the trained MLM, the plurality of determined key features with prior key features associated with prior anomalous patterns stored in the machine learning model; (par 74 “After the model is trained using the historical failure instances, sensor measurements and events, the model is deployed to work on real-time sensor measurements and events obtained from the equipment.”)
based on the comparison of key features, determining, by the trained MLM, a likelihood that the first anomalous pattern matches one of the prior anomalous patterns;(par 75 “For each classification algorithm, there is a tradeoff between the recall and the false alarm rate. If the recall is high, the algorithm tends to generate more false alarms. On the other hand, if the false alarm is very low, the algorithm will miss more failure cases (i.e., lower recall.) This trade-off can be controlled by changing some the aforementioned model parameters.”)
retrieving attributes related to the cause of the failure and a corrective action associated with the matched prior anomalous pattern if the likelihood exceeds a threshold;(par 36 “Event data: Event data involves sequences of non-failure events that can involve different types of events, such as maintenance actions, alarms, and so on.”; par 37 “Failure data: Failure data involves discrete failure events that happened in the past. Each failure event is associated with a time-stamp that specifies the date and sometimes the time of the failure. The failure event might also have associated attributes like the type of the failure, the particular component that failed, other attributes related to the cause of the failure, the repair actions, and other attributes according to the desired implementation.”) and
taking the corrective action in the semiconductor process to correct the predicted failure. (par 5 “Predictive Maintenance (also known as condition- based maintenance): continually monitoring the condition of the equipment to determine the maintenance actions need to be taken at certain times.”; Par 6 “one of the main objectives of predictive maintenance is to prevent failures before they happen. Related art implementations attempt to prevent failures by monitoring the equipment and searching for any pre-failure patterns.”)
Therefore, it would have been obvious to a person of ordinary skill in the art before the effective filing date of the claimed invention to further modify Iskandar to incorporate defining, by the trained MLM, an anomaly window within the first window of Farahat. One of ordinary skill in the art would have been motivated to remedy the shortcomings of Iskandar -- a need for how to implement predictive maintenance to prevent failures before they happen(Farahat par 6 “In related art implementations, one of the main objectives of predictive maintenance is to prevent failures before they happen. Related art implementations attempt to prevent failures by monitoring the equipment and searching for any pre-failure patterns.”). -- with Farahat providing a known method to solve a similar problem. Farahat provides “implementations described herein involve a system for maintenance recommendation based on data-driven failure prediction. The example implementations can involve estimating the probability of having a failure event in the near future given sensor measurements and events from the equipment, and then alerts the system user or maintenance staff if the probability of failure exceeds a certain threshold.”(Farahat par 8)
Regarding claim 9, Iskandar and Farahat teaches
The modeling method of claim 8,
Iskandar further teaches,
the step of determining key features comprising calculating a plurality of statistical indicators for portions of the traces contained within the anomaly window.(Par 136 “At block 474, the processing logic creates scatterplots based on the hyperplane limit for each of the features generated from the current trace data. In some embodiments, the processing logic creates a scatterplot matrix (e.g., pairwise scatterplots, pairwise sensor-statistics scatterplots) based on the features generated from the current trace data.”; Par 102 “At block 422, the processing logic segments (e.g., windows), via a sliding window, the historical trace data to generate a boundary between steady state and transient segments (e.g., to generate segmented trace data).”)
Regarding claim 10, Iskandar and Farahat teaches
The modeling method of claim 9,
Iskandar further teaches,
the step of determining key features further comprising identifying a plurality of wafer attributes for the semiconductor process.(par 140 “At block 482, the processing logic generates the FDC limit that is a 2D linear limit based on the scatterplot. The FDC limit of block 482 may be a two-variable limit ( e.g., two features determine whether the product is normal or abnormal). In some embodiments, the FDC limit may separate an abnormal region and normal region on a plot. The FDC limit may be a gray region that indicates that at least one additional feature causes the product in the gray region to be normal or abnormal.”)
Regarding claim 11, Iskandar and Farahat teaches
The modeling method of claim 8,
Iskandar further teaches,
further comprising performing feature engineering, by the trained MLM, to select a subset of the key features determined to be important to detecting and identifying the first anomalous pattern.(par 135 “At block 472, the processing logic projects from the hyperplane limit to a FDC limit that is a ID limit or a 2D limit. For example, the hyperplane may only be affected by one or two features, so the hyperplane may be projected onto a ID (e.g., one axis is the feature and one axis is time or run) or 2D ( e.g., one axis is one feature and another axis is another feature).”; fig 4C:446 par 110 “At block 446, the processing logic obtains, from the trained machine learning model, one or more outputs indicative of predictive data ( e.g., indication of normal and abnormal products) associated with the hyperplane limit.”; fig 6A-B; par 112 “The processing logic may use the predictive data to generate one or more plots including the FDC limit ( e.g., see plots 600A-B of FIGS. 6A-B). The processing logic may generate a plot that has three regions: 1) abnormal region indicating abnormal products based on the displayed one or more variables; 2) normal region indicating normal products based on the displayed one or more variables; and 3) gray region indicating abnormal or normal product prediction but not just dependent on the displayed variables ( e.g., based at least on one or more additional variables).” )
Regarding claim 12, Iskandar and Farahat teaches
The modeling method of claim 8,
Iskandar further teaches,
the step of detecting the first anomalous pattern further comprising detecting a rapid change in values of the traces in the stable region or the transition region of operation.(par 53 “In some embodiments, the predictive server 112 (e.g., predictive component 114) may generate features ( e.g., historical features 148, current features 154) from trace data (e.g., historical trace data 144, current trace data 150) and store the features in the data store. In some embodiments, the features are a pattern in the trace data (e.g., slope, width, height, peak, etc.) or a combination of sensor values from the trace data ( e.g., power derived from voltage and current, etc.).” Slope in time series data is a measure of change in value over time. A larger slope means a more rapid change in values.)
Regarding claim 13, Iskandar and Farahat teaches
The modeling method of claim 8,
Iskandar further teaches,
the step of detecting the first anomalous pattern further comprising detecting a rapid change in the rate of change for values of the traces in the stable region or the transition region of operation.(par 89 “Features may include one or more of combinations of sensor values from sensors (e.g., summing, multiplying, subtracting, dividing, etc.), …, slope of sensor values from a sensor, properties (e.g., width, height, etc.) of a peak of sensor values from a sensor, patterns of sensor values from a sensor, frequency of sensor values from a sensor, etc. …. The processing logic may one or more of use SME to determine a feature, look for patterns ( e.g., sine waves) to determine a feature, or the like.”; par 102 “For example, a portion of the historical trace data that has a change in slope that is less than a threshold change in slope (e.g., steady state) may be in a corresponding segment … a third segment, 10 to 15 seconds is not steady state ( e.g., has a change of slope that is greater than the threshold change in slope) and is a second segment).”)
Regarding claim 1, see the teachings of Iskandar and Farahat with respect to claim 8 above. Claim 1 is rejected for the same reasons as claim 8.
Regarding claims 2-6 they are the same as claims 9-13 and are rejected for the same reasons.
Regarding claim 20, Iskandar teaches
A method of using a machine learning model (MLM) for predicting and correcting semiconductor processing equipment failure,(par 4 “The method further includes processing the current trace data to identify a plurality of features of the current trace data and providing the plurality of features of the current trace data as input to a trained machine learning model that uses a hyperplane limit for product classification. The method further includes obtaining, from the trained machine learning model, one or more outputs indicative of predictive data associated with the hyperplane limit. A corrective action associated with the manufacturing equipment is to be performed based on the predictive data.”) comprising:
training, by a computer, the MLM based on input data and a selected training algorithm to generate a trained MLM, (fig 4A:408; par 93 “At block 408, the processing logic trains a machine learning model using training data including the features (e.g., selected feature parameters of the features) of the historical trace data and the product data to generate a trained machine learning model.”), wherein the selected training algorithm includes a fault detection and classification algorithm;(fig 6A-C; par 19 “In some embodiments, the processing logic may generate ( e.g., based on the predictive data, as part of the corrective action) a plot that has two axes (e.g., corresponding to two features) and different regions ( e.g., normal, abnormal, and gray regions) indicated by an FDC limit (see FIGS. 6B-C). The normal region may be on a first side of the FDC limit, the abnormal region may be on a second side of the FDC limit, and the gray region may be within the FDC limit.” FDC stands for “Fault Detection and Classification”(par 14 “Described herein are technologies directed to automatic and adaptive fault detection and classification (FDC) limits.”))
receiving, into the trained MLM, a multiplicity(par 25 “The present disclosure may automatically generate FDC limits for sensor values from multiple sensors and take into account the interactions between the sensors ( e.g., via training and using the machine learning model) which provides more accurate results than conventional systems that compare sensor values from a single sensor or output of a set algorithm to a set limit of time-series traces obtained from a corresponding multiplicity of semiconductor processing equipment sensors during a semiconductor process; (par 28 “The sensors 126 may provide sensor data 142 associated with manufacturing equipment 124 (e.g., associated with producing, by manufacturing equipment 124, corresponding products, such as wafers).” fig 4C:440; par 107 “Referring to FIG. 4C, at block 440, the processing logic receives current trace data including current sensor values associated with producing, by manufacturing equipment, second products.”; par 67 “Although embodiments of the disclosure are discussed in terms of generating predictive data 169 to perform a corrective action in manufacturing facilities (e.g., semiconductor manufacturing facilities), embodiments may also be generally applied to generating limits to perform an action.”)
detecting one or more anomalous patterns of time-series traces in the multiplicity of time- series traces at a time-based location in a stable region of process operation using the trained MLM; ( Par 102 “At block 422, the processing logic segments (e.g., windows), via a sliding window, the historical trace data to generate a boundary between steady state and transient segments (e.g., to generate segmented trace data).”)
identifying a window while sliding around the first time-based location and containing at least a first anomalous pattern of the one or more anomalous patterns using the trained MLM; (Par 102 “At block 422, the processing logic segments (e.g., windows), via a sliding window, the historical trace data to generate a boundary between steady state and transient segments (e.g., to generate segmented trace data).”; par 149 “The sensor data values and/or feature patterns may be used to generate a hyperplane limit (e.g., FIG. 4A) and/or a plot of the FDC limit and regions ( e.g., normal region, abnormal region, gray region) (e.g., FIG. 4C).”; par 72 “In some embodiments, the data set generator 272 may discretize (e.g., segment) one or more of the data input 210 or the target output 220 (e.g., to use in classification algorithms for regression problems). Discretization (e.g., segmentation via a sliding window) of the data input 210 or target output 220 may transform continuous values of variables into discrete values.”; par 97 “The trained machine learning model may provide multivariate limit optimization ( e.g., via sliding window, incremental learning, filtering mechanism, etc.)”
calculating a plurality of statistical indicators for the first anomalous pattern of traces contained within the window; (par 53 “In some embodiments, the predictive server 112 (e.g., predictive component 114) may generate features ( e.g., historical features 148, current features 154) from trace data (e.g., historical trace data 144, current trace data 150) and store the features in the data store. In some embodiments, the features are a pattern in the trace data (e.g., slope, width, height, peak, etc.) or a combination of sensor values from the trace data ( e.g., power derived from voltage and current, etc.).”);
determining that the first anomalous pattern is a likely match with at least one prior anomalous pattern using the MLM; (par 48 “In some embodiments, the predictive component 114 may use a trained machine learning model 190 to determine the output for performing the corrective action based on the current features 154. The trained machine learning model 190 may be trained using the historical features 148 and historical product data 158 to learn key process and hardware parameters.”; par 59 “Predictive component 114 may provide current features 154 to the trained machine learning model 190 and may run the trained machine learning model 190 on the input to obtain one or more outputs. … The predictive component 114 or corrective action component 122 may use the confidence data to decide whether to cause a corrective action associated with the manufacturing equipment 124 based on the predictive data 169.”)
retrieving a root cause and a corrective action associated with the prior anomalous pattern; (par 24 “A corrective action associated with the manufacturing equipment may be performed based on the predictive data.” Par 18 “The processing device may predict, based on the FDC limit, one or more causes of classification within the products (e.g., causes of abnormal wafers) so that a corrective action associated with the manufacturing equipment can be performed.”); and
taking the corrective action in the semiconductor process.(par 82 “At block 320, system 300 uses the trained model (e.g., selected model 308) to receive current features 354 (e.g., current features 154 of FIG. 1) and determines (e.g., extracts), from the output of the trained model, predictive data 369 (e.g., predictive data 169 of FIG. 1, FDC limit) to perform corrective actions associated with the manufacturing equipment 124.”)
However, Iskandar does not specifically teach defining a window around the first time-based location and containing at least a first anomalous pattern of the one or more anomalous patterns using the trained MLM.
On the other hand, Farahat teaches,
A method of using a machine learning model (MLM) for predicting and correcting semiconductor processing equipment failure,(par 8 “Example implementations described herein involve a system for maintenance recommendation based on data-driven failure prediction.”) comprising:
training, by a computer, the MLM based on input data and a selected training algorithm to generate a trained MLM, wherein the selected training algorithm includes a fault detection and classification algorithm;(fig 1(b):110; par 43 “The first pipeline represents the model training 110, … . Model training 110 will execute a process similar to that illustrated in FIG. 2(a) to perform a preparation of data 111 and generate parameter values that is utilized in the data preparation 121 of batch processing 120 and data preparation 131 of stream processing 130.” Par 40 “The extracted windows are then used for training the failure prediction model.”)
receiving, into the trained MLM, a multiplicity of time-series traces obtained from a corresponding multiplicity of semiconductor processing equipment sensors during a semiconductor process;(fig 1(b):101; par 39 “The maintenance recommendation system 102 can intake raw sensor, event and failure data 101 for processing by the following modules.”)
detecting one or more anomalous patterns of time-series traces in the multiplicity of time- series traces at a time-based location in a stable region of process operation using the trained MLM; (fig 1(a):104; par 40 “Evidence window extraction 104 is responsible for extracting windows of sensor measurements and events from historical data which correspond to failure and normal cases. … . The extracted windows will be inspected by the failure prediction model for any signs of failures.”; par 56 “After the data is processed, the next procedure in failure prediction is to extract windows of data which are believed to contain pre-failure patterns.” )
defining a window around the first time-based location and containing at least a first anomalous pattern of the one or more anomalous patterns using the trained MLM; (fig 5(a); par 56 “After the data is processed, the next procedure in failure prediction is to extract windows of data which are believed to contain pre-failure patterns. After that, features are extracted from these windows, and these features are applied to the failure prediction module…”; par 57 “For each past failure instance, windows of observations are extracted. The windows can include the Alert Window ([ta, 1t], ta =t-W J, which is the period of time just before the failure during which countermeasures are required to prevent the failure from happening.”; par 58 “The windows can further include the Evidence Window … which is the period of time before the alert window during which the pre-failure patterns happen. The evidence time windows are considered as the positive class for the failure prediction model.”)
calculating a plurality of statistical indicators for the first anomalous pattern of traces contained within the window; (par 41 “Feature extraction 105 extracts features from each window of measurements. The extracted features are then provided to the failure prediction model for training/application.”)
determining that the first anomalous pattern is a likely match with at least one prior anomalous pattern using the MLM; (par 74 “After the model is trained using the historical failure instances, sensor measurements and events, the model is deployed to work on real-time sensor measurements and events obtained from the equipment.”; par 75 “For each classification algorithm, there is a tradeoff between the recall and the false alarm rate. If the recall is high, the algorithm tends to generate more false alarms. On the other hand, if the false alarm is very low, the algorithm will miss more failure cases (i.e., lower recall.) This trade-off can be controlled by changing some the aforementioned model parameters.”)
retrieving attributes related to the cause of the failure and a corrective action associated with the prior anomalous pattern; (par 36 “Event data: Event data involves sequences of non-failure events that can involve different types of events, such as maintenance actions, alarms, and so on.”; par 37 “Failure data: Failure data involves discrete failure events that happened in the past. Each failure event is associated with a time-stamp that specifies the date and sometimes the time of the failure. The failure event might also have associated attributes like the type of the failure, the particular component that failed, other attributes related to the cause of the failure, the repair actions, and other attributes according to the desired implementation.”) and
taking the corrective action in the semiconductor process. (par 5 “Predictive Maintenance (also known as condition- based maintenance): continually monitoring the condition of the equipment to determine the maintenance actions need to be taken at certain times.”; Par 6 “one of the main objectives of predictive maintenance is to prevent failures before they happen. Related art implementations attempt to prevent failures by monitoring the equipment and searching for any pre-failure patterns.”)
Therefore, it would have been obvious to a person of ordinary skill in the art before the effective filing date of the claimed invention to further modify Iskandar to incorporate defining, by the trained MLM, an anomaly window within the first window of Farahat. One of ordinary skill in the art would have been motivated to remedy the shortcomings of Iskandar -- a need for how to implement predictive maintenance to prevent failures before they happen(Farahat par 6 “In related art implementations, one of the main objectives of predictive maintenance is to prevent failures before they happen. Related art implementations attempt to prevent failures by monitoring the equipment and searching for any pre-failure patterns.”). -- with Farahat providing a known method to solve a similar problem. Farahat provides “implementations described herein involve a system for maintenance recommendation based on data-driven failure prediction. The example implementations can involve estimating the probability of having a failure event in the near future given sensor measurements and events from the equipment, and then alerts the system user or maintenance staff if the probability of failure exceeds a certain threshold.”(Farahat par 8)
Response to Arguments
Applicant's arguments filed 9/17/2025 regarding the rejection under 35 U.S.C. 101 have been fully considered but they are not persuasive.
With respect to the independent claims, the applicant has argued that the mental process steps can not practically be performed in the human mind, and so should not fall under the mental process category. Applicant explains that limitations “detecting, by the trained MLM, an anomalous pattern of traces in a stable region or a transition region of process operation at a first time-based location in the multiplicity of time-series traces; Defining, by the trained MLM, an anomaly window… determining, by the trained MLM, a plurality of key features … comparing by the trained MLM the plurality of determined key features… ” infers that there is too much data for a human to process.
The examiner respectfully disagrees. Reference US 20200057689 A1(Farahat) teaches that predictive maintenance has historically been done manually (par 6 “In related art implementations, one of the main objectives of predictive maintenance is to prevent failures before they happen. Related art implementations attempt to prevent failures by monitoring the equipment and searching for any pre-failure patterns. Such related art monitoring processes are conducted manually through visual inspection of equipment, or through using monitoring tools such as vibration monitoring and other devices.”). [See MPEP 2106.04(a)(2) “Examples of claims that recite mental processes include: a claim to "collecting information, analyzing it, and displaying certain results of the collection and analysis," where the data analysis steps are recited at a high level of generality such that they could practically be performed in the human mind, Electric Power Group v. Alstom, S.A., 830 F.3d 1350, 1353-54, 119 USPQ2d 1739, 1741-42 (Fed. Cir. 2016);”]. Also, Machine Learning Models(MLM) are generic components designed to process large amounts of data, and are also the tool used to perform the process.[See MPEP 2106.04(a)(2);“A Claim That Requires a Computer May Still Recite a Mental Process … 1. Performing a mental process on a generic computer.”]. Also, for the sake of argument, even if performing statistical calculations were not mental processes, they are still also considered mathematical calculations, and would still be considered to contain an abstract idea for step 2A prong one, regardless of whether a human can perform the mathematical calculation or not. [see MPEP 2106.04(a)(2) 1. “Mathematical concepts”])
With respect to the independent claims, the applicant has argued that the claims integrate the judicial exception into a practical application under prong two of step 2A. The examiner respectfully disagrees. Although limitation “a multiplicity of time-series traces obtained from a corresponding multiplicity of semiconductor processing equipment sensors during steps of a semiconductor process” is recited, the claims only generally apply the abstract idea of anomaly detection and correction to the field of semiconductor processing equipment. The same process except for the “semiconductor processing equipment” and “semiconductor process” descriptors would also work for a heart pacemaker, cloud computing systems, a car status monitoring system(check engine light), an air conditioning system. [See MPEP 2106.04(d)(1) “Evaluating Improvements in the Functioning of a Computer, or an Improvement to Any Other Technology or Technical Field in Step 2A Prong Two” and MPEP 2106.05(h)“Field of Use and Technological Environment”]
With respect to the independent claims, the applicant has further argued that under prong one of step 2A, the claims do not recite a judicial exception, but instead merely involve a judicial exception, explaining that detecting an anomalous pattern of traces is not reasonably capable of being performed in the human mind due to the inherent numerosity of such data from sensors installed in semiconductor processing equipment. The examiner respectfully disagrees. Even if the limitations require a computer, it can still be a mental process [see MPEP 2106.04(a)(2) III. C. "A Claim That Requires a Computer May Still Recite a Mental Process"]. Defining windows, detecting anomalous patterns, determining key features, looking up features in a database for similar situations in the past, following the instructions stored in the database on how the issue was fixed in the past, are directed to mental processes because the steps are recited at a high level of generality and merely use computers as a tool to perform the processes.
Applicant’s arguments, see remarks pg 7-10, filed 9/17/2025, with respect to the rejection(s) of claim(s) 1-13 under 35 U.S.C. 102(a)(2) as being anticipated by US 20210042570 A1 (Iskandar) have been fully considered and are persuasive. Therefore, the rejection has been withdrawn. However, upon further consideration, a new ground(s) of rejection is made in view of 35 U.S.C. 103 as being unpatentable over by US 20210042570 A1 (Iskandar) in view of US 20200057689 A1 (Farahat).
With respect to the independent claims, the applicant has argued that Iskandar does not teach limitation “defining, by the trained MLM, an anomaly window around the first time-based location and containing the first anomalous pattern.” The newly cited Farahat teaches in the cited (fig 5(a); par 56 “After the data is processed, the next procedure in failure prediction is to extract windows of data which are believed to contain pre-failure patterns. After that, features are extracted from these windows, and these features are applied to the failure prediction module…”; par 57,58;). The examiner interprets this as limitation “defining, by the trained MLM, an anomaly window around the first time-based location and containing the first anomalous pattern.”.
With respect to the independent claims, the applicant has also argued that Iskandar does not teach comparing key features of the current anomaly with key features of prior anomalies, or determining key features within the anomaly window, described by limitations “determining, by the trained MLM, a plurality of key features for portions of the multiplicity of time-series traces within the second anomaly window;” and “comparing, by the trained MLM, the plurality of determined key features with prior key features associated with prior anomalous patterns stored in a database;” (claim 8). The examiner respectfully disagrees.
Iskandar teaches, in the cited par 53 “In some embodiments, the predictive server 112 (e.g., predictive component 114) may generate features ( e.g., historical features 148, current features 154) from trace data (e.g., historical trace data 144, current trace data 150) and store the features in the data store. In some embodiments, the features are a pattern in the trace data (e.g., slope, width, height, peak, etc.) or a combination of sensor values from the trace data ( e.g., power derived from voltage and current, etc.).”. The examiner interprets this as “determining, by the trained MLM, a plurality of key features for portions of the multiplicity of time-series traces within the second anomaly window;”.
Iskandar teaches, in the cited par 48 “In some embodiments, the predictive component 114 may use a trained machine learning model 190 to determine the output for performing the corrective action based on the current features 154. The trained machine learning model 190 may be trained using the historical features 148 and historical product data 158 to learn key process and hardware parameters.”. The examiner interprets this as “comparing, by the trained MLM, the plurality of determined key features with prior key features associated with prior anomalous patterns stored in a database;”.
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure.
US 9952921 B2 - Kim – analyzes time series data in windows to predict failures
US 20180025483 A1 - Finlay - semiconductor manufacturing error prediction system
US 7676775 B2 - Chen - tests wafers after production to find defects and find root cause to improve manufacturing process. Matches fail patterns to find root cause and implements resolutions.
US 20230122653 A1 - Yoshida - error cause estimation when manufacturing computer parts. Looks at features and determines anomalies
US 20190379589 A1 - Ryan - analyzes time series data to find anomalous spikes within time windows.
US 20190324831 A1 - Gu - analyzes log data to create key feature error labels.
US 20160246662 A1 - Meng- diagnosis window and refined diagnosis window for troubleshooting in general. Also extracts features and discovers normal and abnormal "drifts"/changes
Any inquiry concerning this communication or earlier communications from the examiner should be directed to MICHAEL XU whose telephone number is (571)272-5688. The examiner can normally be reached Monday-Friday 8:00am - 5:00pm.
Examiner interviews are available via telephone, in-person, and video conferencing using a USPTO supplied web-based collaboration tool. To schedule an interview, applicant is encouraged to use the USPTO Automated Interview Request (AIR) at http://www.uspto.gov/interviewpractice.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Bryce Bonzo can be reached at (571) 272-3655. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300.
Information regarding the status of published or unpublished applications may be obtained from Patent Center. Unpublished application information in Patent Center is available to registered users. To file and manage patent submissions in Patent Center, visit: https://patentcenter.uspto.gov. Visit https://www.uspto.gov/patents/apply/patent-center for more information about Patent Center and https://www.uspto.gov/patents/docx for information about filing in DOCX format. For additional questions, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free). If you would like assistance from a USPTO Customer Service Representative, call 800-786-9199 (IN USA OR CANADA) or 571-272-1000.
/MICHAEL XU/Examiner, Art Unit 2113
Prosecution Insights