DETAILED ACTION
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 Interpretation
As noted previously, elected claims are drawn to an apparatus. "Apparatus claims cover what a device is, not what a device does." Hewlett-Packard Co. v. Bausch & Lomb Inc., 909 F.2d 1464, 1469, 15 USPQ2d 1525, 1528 (Fed. Cir. 1990) (emphasis in original). A claim containing a "recitation with respect to the manner in which a claimed apparatus is intended to be employed does not differentiate the claimed apparatus from a prior art apparatus" if the prior art apparatus teaches all the structural limitations of the claim. Ex parte Masham, 2 USPQ2d 1647 (Bd. Pat. App. & Inter. 1987) (MPEP 2114). Furthermore, examiner notes that, “inclusion of material or article worked upon by a structure being claimed does not impart patentability to the claims.” (MPEP 2115). For example, specific weld quality parameter (i.e. tensile strength, fatigue, or peel strength), different types of weld data such as ultrasonic power, electrical current, voltage, force or velocity of the horn, weld graph data etc. (claims 5-8), and “bundle of wires” (claim 10) are workpiece materials which do not structurally limit the claimed apparatus/system.
Election/Restrictions
The newly added claims are directed to the following patentably distinct species:
Species A- computer system having trained data analytics model and configured to select a particular data analytics model (claims 1, 33-34, 47-48)
Species B- computer system configured to apply statistical process control (SPC) algorithm and filtering to remove anomalies (claims 41-46 and 50).
2. Examiner notes that above species were presented in the original restriction (mailed 5/17/24) and in response (filed 6/28/24), Applicant had elected Species A. The species are distinct because they recite mutually exclusive characteristics which vary from one species to another. In addition, these species are not obvious variants of each other based on the current record.
There is a search and/or examination burden for the patentably distinct species as set forth above because at least the following reason(s) apply:
the inventions have acquired a separate status in the art due to their recognized divergent subject matter
the inventions require a different field of search (e.g., searching different classes /subclasses or electronic resources, or employing different search strategies or search queries).
prior art applicable to one invention may not be applicable to another.
Since Applicant has received an action on the merits for the originally presented invention (Species A), this invention has been constructively elected by original presentation for prosecution on the merits. Accordingly, claims 41-46 and 50 are withdrawn from consideration as being directed to a non-elected invention. See 37 CFR 1.142(b) and MPEP § 821.03.
Examiner also points out that the Applicant cannot, as a matter of right, file a request for continued examination (RCE) to obtain continued examination on the basis of claims that are independent and distinct from the claims previously claimed and examined (i.e., applicant cannot switch inventions by way of an RCE as a matter of right). See MPEP 819.
Claim Rejections - 35 USC § 103
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.
Claims 1-3, 5-7, 10, 12, 39 and 47-48 are rejected under 35 U.S.C. 103 as being unpatentable over Narayanan et al. (US 10,875,125, hereafter “Narayanan”) in view of Li et al. (Online Quality Inspection of ultrasonic welding by combining artificial intelligence technologies, Materials and Design, issue 194, pg. 1-10, 2020, see NPL of record), Steinmeier (US 11,167,378) and Guo et al. (“Profile monitoring and fault diagnosis via sensors for Ultrasonic welding”, Journal of Manufacturing Science and Engineering, Aug. 2019, NPL of record, “Guo”).
Regarding claim 1, Narayanan discloses a system 100 comprising: a computer having a processor and memory, which stores executable instructions, (fig. 9) configured to: receive weld parameter data 610/710 generated during a welding process by a welder to join at least two parts with a weld (figs. 6-7); input the received weld parameter data to a data analytics model 600/700 (Classification model) to generate at least one predicted weld quality parameter (strength, ductility, etc.- col. 9, lines 18-25); compare the predicted weld quality parameter with a weld quality parameter threshold (checking specification threshold); and generate output indicating at least one of: the at least one predicted weld quality parameter and a result of the comparison between the at least one predicted weld quality parameter and the weld quality parameter threshold (col. 9, lines 5-65).
Narayanan discloses that the data analytics model is based on at least one or more machine learning algorithms selected from regression algorithm, classification algorithm, an artificial intelligence neural network algorithm, a decision tree algorithm, K nearest neighbor algorithm, a support vector machine algorithm, and a gradient boosting technique (col. 8, lines 21-38), but does not explicitly mention random forest model. However, such model is known in the art. Li (directed to Quality Inspection in ultrasonic welding using artificial intelligence technologies- title, abstract) discloses two AI models- namely, artificial neuron network (ANN) and random forest (RF) (pg. 2- right column). Prior art research has indicated that machine-learning methods (includes RF model) have exhibited higher prediction accuracy than conventional regression method (pg. 2- left column). Li discloses detailed parameters of the RF model and compares its performance (pg. 4). Li teaches that the study further combines ANN and RF models to predict weld quality and based on their advantages, it results in significant improvement in prediction accuracy (see Discussion section, pgs. 7-8). Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the invention to utilize random forest (RF) model in the system of Narayanan since such model is within purview of Narayanan’s teachings and doing so would provide significant improvement in prediction accuracy, as suggested by Li.
Narayanan discloses that different types of strengths for weldment specification may include tensile strength, yield strength, compressive strength, shear fatigue strength or impact strength (col. 5, lines 55-65), but does not mention ‘peel strength’. However, such parameter is known in the art. Analogous to Narayanan, Steinmeier is directed to techniques for determining weld quality using weld information algorithm (abstract). Steinmeier teaches that the welding system comprising a processor obtains an indication of weld quality that includes at least one of weld tensile strength, shear strength or peel strength (col. 7, lines 44-50). Accordingly, it would have been obvious to one of ordinary skill in the art before the effective filing date of the invention to include peel strength as a quality parameter in the analytic method of Narayanan because such parameter is an art-recognized indicator of weld quality, as evidenced by Steinmeier.
Narayanan does not mention the term “weld graph data” or “profiling” with respect to analyzing the weld data for weld quality evaluation. However, this concept is known in the welding art. Guo is directed to method for effective profile monitoring and fault diagnosis for ultrasonic welding (abstract). Guo teaches that sensor measurements provide time-dependent cycle-based profile data which is used to facilitate detection of system anomalies, root causes and monitoring of operational quality of the manufacturing/welding process (see pg. 1- Introduction- left column). Specifically, Guo teaches algorithms of multi-linear discriminant analysis (MLDA) and vectorized LDA (VLDA) using multilinear algebra involving matrix mathematics (see section 2). The general framework of profile monitoring and fault diagnosis using multistream signals is illustrated in fig. 5; the output label represents normal or some fault type (see pg. 5). This meets analyzing weld graph data (values over a time period) using matrix profiling to identify faults/anomalies within the data and generating output based on the comparing and analyzing of weld parameter data. In conclusion, Guo states that UMLDA outperforms VLDA in not only detecting the faults but also classifying the type of faults (see pg. 12 conclusion- first paragraph). Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the invention to employ graph profiling and ULMDA algorithm(s) to identify anomalies/faults within the weld data in the system of Narayanan because doing so would provide effective and quick diagnosis of classifying the type of faults, thereby improving the operational quality of the welding process.
Thus, Narayanan as modified by Li, Steinmeier & Guo above discloses the computer system that is configured to generate a predicted peel strength of the weld (as quality parameter) using a random forest analytics model, compare to a reference/threshold peel strength, and generate output based on the comparison & graph profiling analysis of the weld parameter data as recited for weld quality evaluation.
As to claim 2, Narayanan discloses that the data analytics model is trained during a training procedure based on actual weld data that associates weld parameter data with weld quality data (col. 5, lines 13-26, 43-60; col. 6, lines 8-20).
As to claim 3, Narayanan discloses that the data analytics model is based on at least one of a supervised learning algorithm, an unsupervised learning algorithm, a classification algorithm, an artificial intelligence neural network algorithm, K nearest neighbor algorithm, a support vector machine algorithm, and a gradient boosting regression algorithm (col. 8, lines 25-38; claim 4). Li teaches ANN and RF models. Guo also teaches that the data analytics model is based on at least one of a supervised learning algorithm, an unsupervised learning algorithm, and a classification algorithm (pg. 5). For example, tables 1-3 illustrate confusion matrix for the nearest neighbor classifier (NNC- pg. 8-9- tables 1-3).
As to claims 5-7, examiner notes that particular types of data (ultrasonic power, current, voltage, horn force or velocity, weld graph) concerns intended use of the analytics model and does not structurally limit the claimed system. Narayanan teaches that the weld parameter data generated during the welding process includes different data types such as current or voltage data, wire feed speed, welding waveform, width, travel speed data etc. (col. 3, lines 5-13). Similarly, Guo also teaches that weld parameter data includes power/voltage, force and sound waveform data (pg. 10- Table 4). Accordingly, the system of Narayanan, Li, Steinmeier & Guo above is well configured to process recited data parameters.
As to claim 7, examiner maintains official notice with respect to the weld parameter data generated by at least one sensor, which is common knowledge in the art. Guo teaches weld parameter data generated by sensors (Table 4).
As to claim 10, the weld apparatus in Narayanan is configured to join different types of parts, including a bundle of wires (note Claim Interpretation above concerning workpieces).
As to claim 12, Narayanan is silent concerning calibrating the data analytics model by utilizing a tuning procedure after operation in the field. Examiner notes that calibration/tuning as recited are generally broad and not defined by any numeric parameters or technique. Moreover, performing routine recalibration of any model is within common technical maintenance of ordinary artisan. Nonetheless, Li teaches predictive models of artificial neural network (ANN) and random forest (RF), which exhibit high accuracies for online monitoring (abstract), and also teaches calibrating the ANN model with more experiments to help further reduce the relative error of the model (pg. 7- Discussion- right column). Accordingly, it would have been obvious to one of ordinary skill in the art before the effective filing date of the invention to carry out recalibrating the data analytics model by utilizing a tuning procedure after operation in the field in the system of Narayanan & Guo motivated by reducing error rate and improving prediction accuracy.
As to claim 39, Narayana teaches obtaining images data using a digital camera (col. 5, lines 6-9).
As to claims 47-48, examiner notes that particular types of data (ultrasonic power, current, voltage, horn force or velocity, weld graph) concerns intended use of the analytics model and does not structurally limit the claimed system. Narayanan teaches that the weld parameter data generated during the welding process includes different data types such as current or voltage data, wire feed speed, welding waveform, width, travel speed data etc. (col. 3, lines 5-13). Similarly, Guo also teaches that weld parameter data includes power/voltage, force and sound waveform data (pg. 10- Table 4). Accordingly, the system of Narayanan, Li, Steinmeier & Guo above is well configured to process recited data parameters.
Claims 33-35 are rejected under 35 U.S.C. 103 as being unpatentable over Narayanan (US 10,875,125) in view of Li et al. (Online Quality Inspection of ultrasonic welding by combining artificial intelligence technologies, Materials and Design, issue 194, 2020, see NPL of record), Satpathy et al. (Ultrasonic welding of dissimilar metals: A study on joint strength by machine learning, Journal of Manufacturing Processes, 2018, issue 33, pg. 96-110, of record), and further in view of Guo et al. (“Profile monitoring and fault diagnosis via sensors for Ultrasonic welding”, Journal of Manufacturing Science and Engineering, Aug. 2019, of record).
Regarding claim 33, Narayanan discloses a system 100 comprising: a computer having a processor and memory, which stores executable instructions, (fig. 9) configured to: receive weld parameter data 610/710 generated during a welding process by a welder to join at least two parts with a weld (figs. 6-7); input the received weld parameter data to a data analytics model 600/700 (Classification model) to generate at least one predicted weld quality parameter (strength, ductility, etc.- col. 9, lines 18-25); compare the predicted weld quality parameter with a weld quality parameter threshold (checking specification threshold); and generate output indicating at least one of: the at least one predicted weld quality parameter and a result of the comparison between the at least one predicted weld quality parameter and the weld quality parameter threshold (col. 9, lines 5-65).
Narayanan discloses that the data analytics model is based on at least one or more machine learning algorithms selected from regression algorithm, classification algorithm, an artificial intelligence neural network algorithm, a decision tree algorithm, K nearest neighbor algorithm, a support vector machine algorithm, and a gradient boosting technique (col. 8, lines 21-38), but does not explicitly mention random forest model. However, such model is known in the art. Li (directed to Quality Inspection in ultrasonic welding using artificial intelligence technologies- title, abstract) discloses two AI models- namely, artificial neuron network (ANN) and random forest (RF) (pg. 2- right column). Prior art research has indicated that machine-learning methods (includes RF model) have exhibited higher prediction accuracy than conventional regression method (pg. 2- left column). Li discloses detailed parameters of the RF model and compares its performance (pg. 4). Li teaches that the study further combines ANN and RF models to predict weld quality and based on their advantages, it results in significant improvement in prediction accuracy (see Discussion section, pgs. 7-8). Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the invention to utilize random forest (RF) model in the system of Narayanan since such model is within purview of Narayanan’s teachings and doing so would provide significant improvement in prediction accuracy, as suggested by Li.
Claim 33 merely differs from claim 1 by requiring to select a particular analytics model from a plurality of analytics models based on comparison. Comparing multiple models technique is known in the art. Satpathy teaches that based on background literature, machine learning analytics techniques are quite common in different welding processes for modeling of process parameter (Introduction- pg. 97- right column). Satpathy discloses developing plurality of machine learning models such as regression, artificial neural network (ANN) and adaptive neuro-fuzzy inference system (ANFIS) for predicting and simulating joint strength for ultrasonic welding of dissimilar metals and comparing their error and performance, wherein the calculated errors are based on known weld quality data (validation and testing dataset) (figs. 12-13, pg. 106-107). Generally, the average absolute error for the 3 models (regression, ANN, ANFIS) is less than about 1-2% (pg. 107); ANFIS providing most accurate predictions for the current domain of experiments (pg. 108; Tables 8-9 show accuracy testing). Given teachings of Satpathy, one would appreciate and understand that a suitable machine learning model would be selected based on given accuracy/error criteria/threshold. Accordingly, artisan of ordinary skill in the art would have found it obvious to select a particular analytics model in Narayanan based on accuracy for the desired welding tasks to provide reliable predictions of quality parameter such as tensile strength or peel strength.
Narayanan does not mention the term “weld graph data” or “profiling” with respect to analyzing the weld data for weld quality evaluation. However, such concept is known in the welding art. Guo is directed to method for effective profile monitoring and fault diagnosis for ultrasonic welding (abstract). Guo teaches that sensor measurements provide time-dependent cycle-based profile data which is used to facilitate detection of system anomalies, root causes and monitoring of operational quality of the manufacturing/welding process (see pg. 1- Introduction- left column). Specifically, Guo teaches algorithms of multi-linear discriminant analysis (MLDA) and vectorized LDA (VLDA) using multilinear algebra involving matrix mathematics (see section 2). The general framework of profile monitoring and fault diagnosis using multistream signals is illustrated in fig. 5; the output label represents normal or some fault type (see pg. 5). This meets analyzing weld graph data (values over a time period) using matrix profiling to identify faults/anomalies within the data and generating output based on the comparing and analyzing of weld parameter data. In conclusion, Guo states that UMLDA outperforms VLDA in not only detecting the faults but also classifying the type of faults (see pg. 12 conclusion- first paragraph). Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the invention to employ graph profiling and ULMDA algorithm(s) to identify anomalies/faults within the weld data in the system of Narayanan because doing so would provide effective and quick diagnosis of classifying the type of faults, thereby improving the operational quality of the welding process.
Hence, Narayanan as modified by Li, Satpathy & Guo above discloses a system configured to: select a particular data analytics model from the plurality of data analytics RF models based on comparison; receive actual weld parameter data, which includes parameter values over a time period, generated during a welding process by a welder to join at least two parts with a weld; input the actual weld parameter data to the particular data analytics model to generate at least one predicted weld quality parameter; compare the predicted weld quality parameters with known weld quality parameter data (validation and testing dataset); analyze the weld graph data using graph profiling to identify faults/anomalies within the data; and generate output based on the comparison & analysis indicating at least one predicted weld quality parameter and a result of the comparison between the at least one predicted weld quality parameter and the weld quality parameter threshold.
As to claim 34, Narayanan discloses that the data analytics models are additionally based on at least one of a supervised learning algorithm, an unsupervised learning algorithm, a classification algorithm, an artificial intelligence neural network algorithm, a decision tree algorithm, a K nearest neighbor algorithm, a support vector machine algorithm, and a gradient boosting regression algorithm (col. 8, lines 25-38; claim 4). Guo also teaches that the data analytics model is based on at least one of a supervised learning algorithm, an unsupervised learning algorithm, and a classification algorithm (pg. 5). For example, tables 1-3 illustrate confusion matrix for the nearest neighbor classifier (NNC- pg. 8-9- tables 1-3).
As to claim 35, Narayanan discloses that one exemplary predicted weld quality parameter represents a pull/tensile strength of welded joint (col. 6, lines 56-65; col. 7, lines 6-12).
Response to Amendment and Arguments
Applicant’s arguments with respect to amended claim(s) have been considered but are moot in light of new ground(s) of rejection(s) set forth above. Current 103 rejection for claims 1 and 33 now includes Li reference, which discloses the amended feature of using random forest as the data analytics model. Examiner contends that new claims 41-46 and 50 are directed to non-elected species and therefore withdrawn. Applicant is suggested to file divisional application for the method and distinct species.
Conclusion
Applicant's amendment necessitated the new ground(s) of rejection presented in this Office action. Accordingly, THIS ACTION IS MADE FINAL. See MPEP § 706.07(a). Applicant is reminded of the extension of time policy as set forth in 37 CFR 1.136(a).
A shortened statutory period for reply to this final action is set to expire THREE MONTHS from the mailing date of this action. In the event a first reply is filed within TWO MONTHS of the mailing date of this final action and the advisory action is not mailed until after the end of the THREE-MONTH shortened statutory period, then the shortened statutory period will expire on the date the advisory action is mailed, and any nonprovisional extension fee (37 CFR 1.17(a)) pursuant to 37 CFR 1.136(a) will be calculated from the mailing date of the advisory action. In no event, however, will the statutory period for reply expire later than SIX MONTHS from the mailing date of this final action.
Inquiry
Any inquiry concerning this communication or earlier communications from the examiner should be directed to DEVANG R PATEL whose telephone number is (571) 270-3636. The examiner can normally be reached on Monday-Friday 8am-5pm, EST.
To schedule an interview, Applicant is encouraged to use the USPTO Automated Interview Request (AIR) at https://www.uspto.gov/patents/laws/interview-practice. Communications via Internet email are at the discretion of Applicant. If Applicant wishes to communicate via email, a written authorization form must be filed by Applicant: Form PTO/SB/439, available at www.uspto.gov/patent/patents-forms. The form may be filed via the Patent Center and can be found using the document description Internet Communications, see https://www.uspto.gov/patents/apply/forms. In limited circumstances, the Applicant may make an oral authorization for Internet communication. See MPEP § 502.03.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Keith Walker can be reached on 571-272-3458. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300.
Information regarding the status of an application may be obtained from the Patent Center. For more information, see https://patentcenter.uspto.gov. For questions, technical issues or troubleshooting, please contact the Patent Electronic Business Center at ebc@uspto.gov or 1-866-217-9197 (toll-free).
/DEVANG R PATEL/
Primary Examiner, AU 1735
Prosecution Insights