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 .
Priority
Acknowledgment is made of applicant’s claim for foreign priority under 35 U.S.C. 119 (a)-(d). The certified copy has been filed in parent Application No. IN202241037543, filed on 06/30/2022.
Response to Amendment
Applicant’s amendment filed 02/10/2026 has been entered. Claims 1-18 remain pending. Applicant’s amendments to the claims overcome each and every objection to the Claims. Applicant’s amendments to Claims 1-18 overcome each and every single 35 U.S.C. 112(b) rejection of Claims 1-18.
Response to Arguments
Applicant’s arguments, see Pages 20-24, filed 02/10/2026, with respect to 35 U.S.C. 101 rejection of Claims 1-18 have been fully considered and are persuasive. The 35 U.S.C. 101 rejection of Claims 1-18 has been withdrawn.
Applicant's arguments, see Pages 24-28, filed 02/10/2026 with respect to the 35 U.S.C. 103 rejection have been fully considered but they are not persuasive.
Applicant argues on page 25 that Nilakantan does not teach or suggest “computing spatial distribution based geometrical parameters across defined regions of historical products; identifying correlations between spatial geometry distributions and defect types and locations; determining localized defect regions based on such correlations; generating optimized machine parameters specifically to reduce defects at identified locations, as opposed to general part programming or toolpath optimization”. Applicant argues that Nilakantan’s details in [0006] and [0058] do not equate to the claimed spatial distribution-based geometrical parameters which require region-level geometric representation tied to defect localization. Applicant argues on Page 25 that Srivastava operates on post-manufacture scan data and not historical product geometry databases, does not compute spatial distribution-based geometrical parameters across predefined regions of historical products, does not identify correlations between geometry distributions and defect types and locations across historical data, and does not generate optimized machine parameters suggested by a ML model to proactively reduce defects in future manufacturing runs. Applicant argues that Srivastava identifies interest regions on a toolpath using statistical defect probability and that it is fundamentally different from Applicant’s claimed approach. Applicant argues on Page 26 that Abad does not compute spatial distribution-based geometrical parameters, does not tie similarity analysis to defect type and location correlations, and does not involve machine learning-driven defect region determination or parameter optimization. Applicant argues on Page 26 that Jain merely provides a generic similarity metric for identifying similar products within a portfolio and that Jain does not disclose a defect-centric, region-wise geometry-defect correlation framework. Applicant further argues on page 27 that Jain does not teach or suggest using similarity analysis as part of a machine learning pipeline that outputs defect reducing machine parameters. Applicant argues on Page 28 that the claims require a specific technical workflow and this architecture produces a technical improvement in manufacturing defect prevention and that none of the cited references disclose or suggest this integrated approach.
Examiner respectfully disagrees. In response to applicant's arguments against the references individually, one cannot show nonobviousness by attacking references individually where the rejections are based on combinations of references. See In re Keller, 642 F.2d 413, 208 USPQ 871 (CCPA 1981); In re Merck & Co., 800 F.2d 1091, 231 USPQ 375 (Fed. Cir. 1986). Previously disclosed prior art Nilakantan (US20210178697) teaches in [0057]-[0058] a machine learning model that records information relating to part programming, machine variables and specific software variables, with a list of the variables and parameters involved. Quite a few of the parameters given are directed to spatial distribution geometrical data under broadest reasonable interpretation, as the limitation “the first spatial distribution-based geometrical parameters represent spatial distribution of geometry across different regions of the one or more historical products”, including the slices of the STL model (from the CAD being converted to stereolithography format) and the distances between the slices. Thus this would be a spatial distribution based geometrical parameter, and it represents geometry across different regions of the one or more historical products. Nilakantan teaches in [0027] the feature extractor generating a feature vector that includes coefficients generated at the pattern decomposition with the machine learning model determines at least one prefabrication model adjustment parameter for a new job from the metric and Nilakantan further teaches in [0068] that the machine learning system automatically incorporates lessons learned from previous part builds. Thus Nilakantan teaches the limitation of generating optimizing parameters for a machine based on the detected defects and the adjustment parameters would reduce defects at one or more locations in the manufacturing process. Previously disclosed prior art Srivastava (US20190391562) teaches in [0029] utilizing the statistical model and analysis to provide probabilities for specific part characteristics, attributes, shapes, structures, positions, or any other delineating part feature. That is, Srivastava teaches geometrical parameters. Srivastava further teaches in [0029] using the statistical analysis on part geometries and defect locations to determine probabilistic defect locations, thus providing a correlation between spatially distributed geometric parameters and the defect types and locations. Thus the combined reference of Nilakantan in view of Srivastava teaches the amended claimed limitations.
Applicant argues on Page 27 that the cited art addresses distinct and unrelated technical problems with Nilakantan: automating additive manufacturing setup using historical records; Srivastava: post-manufacture toolpath correction based on scan data; Abad: geometric similarity comparison; Jain: numerical similarity metrics. Applicant further argues that there is no teaching, suggestion, or motivation to combine these references to arrive at a system that learns correlations between spatial geometry distributions and defect locations and uses the correlations to proactively generate optimized machine parameters for defect reduction.
Examiner respectfully disagrees. In response to applicant’s argument that there is no teaching, suggestion, or motivation to combine the references, the examiner recognizes that obviousness may be established by combining or modifying the teachings of the prior art to produce the claimed invention where there is some teaching, suggestion, or motivation to do so found either in the references themselves or in the knowledge generally available to one of ordinary skill in the art. See In re Fine, 837 F.2d 1071, 5 USPQ2d 1596 (Fed. Cir. 1988), In re Jones, 958 F.2d 347, 21 USPQ2d 1941 (Fed. Cir. 1992), and KSR International Co. v. Teleflex, Inc., 550 U.S. 398, 82 USPQ2d 1385 (2007). In this case, Nilakantan details an additive manufacturing process [0001] with the use of computer-aided design [0044] and machine learning, with adjustment to the manufacturing process based on the learning to prevent errors [0068]. Srivastava teaches a manufacturing process that utilizes computer aided design and computer-aided manufacturing to identify defects [0009]-[0011]. Thus Nilakantan and Srivastava are in related technical detail to utilize computer aided design in the manufacturing of parts and reducing errors (defects) in the manufactured parts. Thus the incorporate of the teaching of Srivastava with the statistical analysis of the part being manufactured to determine defects into that of Nilakantan would yield predictable results in the determination of the defects (errors). Similarly, Abad teaches the utilization of computer-aided design to design a product and the manufacture of the product in [0039]-[0041]. Thus it would have been obvious to one of ordinary skill in the art to modify Nilakantan in view of Srivastava to incorporate the teaching of Abad as performing the structural analysis with the geometrical models is an improvement that reduces errors generated in the process. Jain teaches product design with historical data and real time data with the utilization of machine learning. Thus it would be obvious to one of ordinary skill in the art to combine Jain with the preceding three references as Jain is in the same field of art of utilizing computers and machine learning with product design.
Claim Rejections - 35 USC § 112
The following is a quotation of 35 U.S.C. 112(b):
(b) CONCLUSION.—The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the inventor or a joint inventor regards as the invention.
The following is a quotation of 35 U.S.C. 112 (pre-AIA ), second paragraph:
The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the applicant regards as his invention.
Claims 1-18 rejected under 35 U.S.C. 112(b) or 35 U.S.C. 112 (pre-AIA ), second paragraph, as being indefinite for failing to particularly point out and distinctly claim the subject matter which the inventor or a joint inventor (or for applications subject to pre-AIA 35 U.S.C. 112, the applicant), regards as the invention.
Claims 1 and 10 details the limitations:
“wherein the one or more first geometrical parameters comprise at least one of: first single-valued geometrical parameters and first spatial distribution-based geometrical parameters, wherein the first spatial distribution-based geometrical parameters represent spatial distribution of geometry across different regions of the one or more historical products”
“wherein the one or more second geometrical parameters comprises at least one of second single-valued geometrical parameters and second spatial distribution-based geometrical parameters”
“wherein the determining, by the one or more hardware processors, comprises identifying correlations between the spatial distribution-based geometrical parameters and the one or more defect types and locations”
It is not clear nor distinct whether the limitation of (c) of “the spatial distribution-based geometrical parameters” is the “first spatial distribution-based geometrical parameters” of (a), the “second spatial distribution-based geometrical parameters” of (b), or a combination that includes the “first spatial distribution-based geometrical parameters” and the “second spatial distribution-based geometrical parameters”.
Examiner interprets the limitation as the combination of the first and second distribution-based geometrical parameters..
Claims 2-9 are rejected due to dependence on Claim 1.
Claims 11-18 are rejected due to dependence on Claim 10.
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, 6, 8, 10, 15, and 17 are rejected under 35 U.S.C. 103 as being unpatentable over Nilakantan (US20210178697) in view of Srivastava (US20190391562).
In regards to Claim 1, Nilakantan teaches “obtaining, by one or more hardware processors (processor [0006]), one or more experimental data from a machine, wherein the one or more experimental data comprise at least one of recipe data, one or more measured parameters, data associated with one or more defect types and locations, and metadata (additive manufacturing knowledge base provides additive manufacturing fabrication parameters and measured/recorded additive manufacturing production outcome data – [0025]);
obtaining, by the one or more hardware processors, one or more first geometry data associated with one or more historical products, wherein each first geometry data is associated with a respective historical product, and wherein each historical product is associated with(a provided CAD file for a part may be in a Standard for the Exchange of Product model data file format that will require conversion to a stereolithography STL file format – [0047]; part designer has a CAD model where the CAD model is passed on to the additive manufacturing operator – [0049]; machine learning model record information with the STL model – [0057]; machine learning model trained with historical records – [0023]; proof of work historical record – [0058]);
computing, by a geometry model on the one or more hardware processors, one or more first geometrical parameters, based on the one or more first geometry data associated with the one or more historical products, wherein the one or more first geometrical parameters comprise at least one of: first single-valued geometrical parameters and first spatial distribution-based geometrical parameters, wherein the first spatial distribution-based geometrical parameters represent spatial distribution of geometry across different regions of the one or more historical products (“Machine-learning model 108 , 214 , 306 , or 418 can thus record information generally relating to part programming, on the one hand, and specific machine variables and/or specific software variables on the other hand. In this context, part programming can consist of (1) configuring the 3D printer; (2) orienting the STL model (CAD converted to a stereolithography format); (3) “slicing” the STL model by intersecting the STL model with a series of horizontal planes to create slice curves; (4) creating support curves, defining where temporary supports will be built in the part, which supports will ultimately be disposed of; (5) creating toolpath fill for model and support curves; (6) saving a toolpath file; and (7) downloading the toolpath file to the printer for part building. Specific machine variables and/or specific software variables can consist of fabrication-machine-specific calibration values; head motor currents; head purge and control parameters; autoload parameters; stepper motor currents; XYZ axis speed parameters; purge and tip wipe locations; head temperature setback and no-ooze parameters; end of curve and purge following error monitoring; head and chamber temperatures used during model build; head velocities; distance between slices (in inches) [i.e. first geometrical parameters comprising spatial distribution of geometry across different regions of the one or more historical products]; minimum and maximum speeds to move the head (in inches per second); origin: X & Y coordinate location to start the model; clearance move: amount to raise the head during glueless moves (in inches); delay time and encoder tick rate (in milliseconds, flowrate); fluid relaxation time during acceleration/deceleration; desired flow accuracy during acceleration/deceleration. In turn, the machine-learning model 410 can produce the part optimization and command initiation outputs 422 , 424 based on training from the above parameters [i.e. one or more first geometrical parameters comprising first single-valued geometrical parameters and first spatial distribution-based geometrical parameters] as stored in the user experience database 410 (which can take the form of a blockchain) and the input vector 416 . In other words, the machine-learning model can be capable of performing the part programming and the parameter setting in an automated way for a new part job based on the past part job data” – [0057]; “An example operation of a machine-learning-based additive manufacturing system according to the flow diagram of FIG. 4 is not described. New part transactions 412 are broadcast to the machine-learning model 418 . The proof-of-work historical record of the user experience database 410 compiles new part transaction data into a block. The proof-of-work historical record becomes more resilient, richer with knowledge, and iteratively stronger with each block addition and feeds the machine-learning model 418 , which is iteratively re-trained either with each block addition or periodically. The machine-learning model 418 provides real-time feedback on the next part transaction based on history. The machine-learning model 418 utilizes the notion of similarity and creates a mapping between, on the one hand, voxel data, orientation data, geometry data, technical language data, and fabrication device variables and parameters, and, on the other hand, the part to be 3D printed. These mappings can, for example, be represented within a structured probability distribution. The proof-of-work historical record 410 , together with the machine-learning model 418 , learns probable model orientation, support structure, toolpaths, and boundary curves based on input. The machine-learning model 418 utilizes neural networks and pattern recognition to make suggestions, predictions, and warnings. The machine-learning model 418 provides outputs 422 , 424 that optimize user/object interaction” – [0058]);
computing, by the geometry model on one or more hardware processors, one or more second geometrical parameters, based on one or more second geometry data associated with the one or more products, wherein the one or more second geometrical parameters comprises at least one of second single-valued geometrical parameters and second spatial distribution-based geometrical parameters (“Machine-learning model 108 , 214 , 306 , or 418 can thus record information generally relating to part programming, on the one hand, and specific machine variables and/or specific software variables on the other hand. In this context, part programming can consist of (1) configuring the 3D printer; (2) orienting the STL model (CAD converted to a stereolithography format); (3) “slicing” the STL model by intersecting the STL model with a series of horizontal planes to create slice curves; (4) creating support curves, defining where temporary supports will be built in the part, which supports will ultimately be disposed of; (5) creating toolpath fill for model and support curves; (6) saving a toolpath file; and (7) downloading the toolpath file to the printer for part building. Specific machine variables and/or specific software variables can consist of fabrication-machine-specific calibration values; head motor currents; head purge and control parameters; autoload parameters; stepper motor currents; XYZ axis speed parameters; purge and tip wipe locations; head temperature setback and no-ooze parameters; end of curve and purge following error monitoring; head and chamber temperatures used during model build; head velocities; distance between slices (in inches); minimum and maximum speeds to move the head (in inches per second); origin: X & Y coordinate location to start the model [i.e. second spatial distribution-based geometrical parameters]; clearance move: amount to raise the head during glueless moves (in inches); delay time and encoder tick rate (in milliseconds, flowrate); fluid relaxation time during acceleration/deceleration; desired flow accuracy during acceleration/deceleration. In turn, the machine-learning model 410 can produce the part optimization and command initiation outputs 422 , 424 based on training from the above parameters [i.e. one or more second geometrical parameters comprising second single-valued geometrical parameters and second spatial distribution-based geometrical parameters]as stored in the user experience database 410 (which can take the form of a blockchain) and the input vector 416 . In other words, the machine-learning model can be capable of performing the part programming and the parameter setting in an automated way for a new part job based on the past part job data” – [0057]; “An example operation of a machine-learning-based additive manufacturing system according to the flow diagram of FIG. 4 is not described. New part transactions 412 are broadcast to the machine-learning model 418 . The proof-of-work historical record of the user experience database 410 compiles new part transaction data into a block. The proof-of-work historical record becomes more resilient, richer with knowledge, and iteratively stronger with each block addition and feeds the machine-learning model 418 , which is iteratively re-trained either with each block addition or periodically. The machine-learning model 418 provides real-time feedback on the next part transaction based on history. The machine-learning model 418 utilizes the notion of similarity and creates a mapping between, on the one hand, voxel data, orientation data, geometry data, technical language data, and fabrication device variables and parameters, and, on the other hand, the part to be 3D printed. These mappings can, for example, be represented within a structured probability distribution. The proof-of-work historical record 410 , together with the machine-learning model 418 , learns probable model orientation, support structure, toolpaths, and boundary curves based on input. The machine-learning model 418 utilizes neural networks and pattern recognition to make suggestions, predictions, and warnings. The machine-learning model 418 provides outputs 422 , 424 that optimize user/object interaction” – [0058]; user experience database includes data defining requirements and specifications, with a fabrication slicing resolution, i.e. second single-valued geometrical parameters – [0006]);
determining, by the one or more hardware processors running a machine learning model trained using data associated with the one or more historical products, the one or more defect of the one or more products, wherein the determining is based on at least one of: the one or more computed statistical features associated with the one or more defect types and locations, the one or more first geometrical parameters, and the one or more second geometrical parameters (machine learning model trained with the accumulated knowledge and applied to new additive manufacturing projects and tasks and include knowledge including defects and flaws – [0022]; machine learning training involves a statistical aggregation of training data – [0034]; machine learning based additive manufacturing system and method can automatically incorporate lessons learned from previous part builds to prevent the repeating of errors and the machine learning model serves as a predictive tool for potential flaws or defects that occur in the manufacturing process – [0068]; Figure 4 shows the user experience database 410 along with the database 414 and new project input vector 416 as inputs to the machine learning model 418); and
generating, by the one or more hardware processors, optimized parameters for the machine based on the determined one or more defect regions, wherein the optimized parameters are suggested by the machine learning model to reduce defects at one or more locations in the manufacturing process (“specifically to methods and systems for machine-learning-based additive manufacturing using manufacturing data” – [0001]; “Such an example can include one or more non-transitory computer-readable media storing instructions that when executed by a computer processor, cause the processor to process, with a machine-learning model trained on a user experience database comprising a plurality of entries, an input vector describing a new part transaction to provide at least one part optimization output and at least one command initiation output to configure additive manufacturing of a new part. Entries in the user experience database each include at least data defining requirements for an additively manufactured part previously fabricated or attempted to be fabricated, specifications describing an additive manufacturing fabrication device, a selection of a raw material type fed to the fabrication device for fabrication of the additively manufactured part, a fabrication spatial orientation of the additively manufactured part within the fabrication device, a fabrication slicing resolution of the additively manufactured part, and a toolpath taken by the fabrication device in fabricating the additively manufactured part” – [0007]; “A feature extractor 106 generates a feature vector that includes at least one of the set of coefficients generated at the pattern decomposition component 104 . A machine learning model 108 determines at least one prefabrication model adjustment parameter and/or print setting parameter for a new print job and/or at least one print job outcome estimate from the metric. As examples, not meant to be an exhaustive list, the prefabrication model adjustment parameter can represent a print orientation of a component or assembly to be printed in a print job, a number of layers to be used or a layer thickness or thicknesses, a geometry parameter descriptive of how a three-dimensional model is divided for printing during a prefabrication model processing phase, or information controlling the placement or geometry of support structures added to the model during a prefabrication model processing phase. As examples, not meant to be an exhaustive list, the print setting parameter can represent a material to be used for the job, a particular printer or model of printer to be used, a nozzle aperture thickness or temperature, or any of a number of other parameters that may be accessible for modification on a 3D printer or other additive manufacturing fabrication device. As examples, not meant to be an exhaustive list, the outcome estimate can be an estimate of print time for the job, an estimate of raw material usage for the print job, or an estimate of the likelihood of success of a print job. The prefabrication model adjustment parameter or print setting parameter provide or contribute to part optimization or command initiation, respectively. The prefabrication model adjustment parameter, print setting parameter, and/or outcome estimate provided by the machine learning model 108 can be stored on a non-transitory computer-readable medium associated with the system 100 and/or provided to a user at a display via a user interface (not shown in FIG. 1, but see FIG. 2)” – [0027]; “The machine-learning-based additive manufacturing systems and methods described herein can automatically incorporate lessons learned from previous part builds to prevent the repeating of errors and to enable affordability, reliability, and maintainability in an additive manufacturing production system. The machine learning model of the described systems and methods can also serve as a predictive tool for potential flaws or defects that may occur in the manufacturing process” – [0068])”
Nilakantan is silent with regards to the language of “computing, by a feature engineering model on the one or more hardware processors, one or more statistical features based on the data associated with one or more defect types and locations of the one or more experimental data obtained from the machine; determining, by the one or more hardware processors, the one or more defect of the one or more products, wherein the determining is based on at least one of: the one or more computed statistical features associated with the one or more defect types and locations, the one or more first geometrical parameters, and the one or more second geometrical parameters; wherein the determining, by the one or more hardware processors, comprises identifying correlations between the spatial distribution-based geometrical parameters and the one or more defect types and locations;.”
Srivastava teaches “computing, by a feature engineering model on the one or more hardware processors, one or more statistical features based on the data associated with one or more defect types and locations of the one or more experimental data obtained from the machine (“As yet another example, the data access engine 108 may identify specific portions along a nominal toolpath 222 at which a manufacturing defect has a higher probability of occurrence. To determine defect probabilities, the data access engine 108 may perform statistical analyses on any number of part designs as well as identified defects in physical parts manufactured from the part designs. Analyzed part designs may include similarly structured part designs that are manufactured to the same, similar, or common manufacturing processes. Through such statistical analysis processes, the data access engine 108 may produce a statistical model [i.e. feature engineering model] providing defect probabilities for specific part characteristics, attributes, shapes, structures, positions, or according to any other delineating part feature. In some instances, the data access engine 108 may perform such a statistical analysis on part geometries and defect locations of other part designs (besides the CAD model 212 ) to determine probabilistic defect locations, which may refer to any part location with a defect probability that exceeds a probability threshold (e.g., greater than 45% or any other configurable threshold). Accordingly, the data access engine 108 may identify interest regions 262 as probabilistic defect locations determined via statistical analysis” – [0029]); determining, by the one or more hardware processors, the one or more defect of the one or more products, wherein the determining is based on at least one of: the one or more computed statistical features associated with the one or more defect types and locations, the one or more first geometrical parameters, and the one or more second geometrical parameters, wherein the determining, by the one or more hardware processors, comprises identifying correlations between the spatial distribution-based geometrical parameters and the one or more defect types and locations; (“As yet another example, the data access engine 108 may identify specific portions along a nominal toolpath 222 at which a manufacturing defect has a higher probability of occurrence. To determine defect probabilities, the data access engine 108 may perform statistical analyses on any number of part designs as well as identified defects in physical parts manufactured from the part designs. Analyzed part designs may include similarly structured part designs that are manufactured to the same, similar, or common manufacturing processes. Through such statistical analysis processes, the data access engine 108 may produce a statistical model providing defect probabilities for specific part characteristics, attributes, shapes, structures, positions, or according to any other delineating part feature [i.e. first geometrical parameters and second geometrical parameters]. In some instances, the data access engine 108 may perform such a statistical analysis on part geometries and defect locations of other part designs (besides the CAD model 212 ) to determine probabilistic defect locations [i.e. defect region], which may refer to any part location with a defect probability that exceeds a probability threshold (e.g., greater than 45% or any other configurable threshold). Accordingly, the data access engine 108 may identify interest regions 262 as probabilistic defect locations determined via statistical analysis [i.e. correlating]” – [0029]).”
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify Nilakantan to incorporate the teaching of Srivastava to utilize statistical analysis of data relating to the part being manufactured to determine defects. By utilizing statistical analysis this yields predictable results in the determination of defects in a part.
In regards to Claims 6 and 15, Nilakantan in view of Srivastava discloses the claimed invention as detailed above. Nilakantan further teaches “training the machine learning model on data associated with the first spatial distribution-based geometrical parameters of the one or more historical products, and the one or more defect types and locations, of the one or more historical products (“Machine learning (ML) is a subset of artificial intelligence (AI) in which a computer uses algorithms and statistical models to accurately perform tasks without using explicitly coded instructions after having after having analyzed a learning or training data set, in effect relying on patterns and inferences to generalize from past experience” – [0003]; “The machine-learning model is trained based on entries in a user experience database, entries in the user experience database each including at least data defining requirements for an additively manufactured part previously fabricated or attempted to be fabricated, specifications describing an additive manufacturing fabrication device, a selection of a raw material type fed to the fabrication device for fabrication of the additively manufactured part previously fabricated or attempted to be fabricated, a fabrication spatial orientation of the additively manufactured part within the fabrication device, a fabrication slicing resolution of the additively manufactured part, and a toolpath taken by the fabrication device in fabricating the part” – [0005]), by:
obtaining, by the one or more hardware processors, the data associated with the first spatial distribution-based geometrical parameters of the one or more historical products, of the one or more historical products (user experience database 410 as input into ML model 418 – [0056], Figure 4; user experience database includes the fabrication spatial orientation of the additively manufactured part within the fabrication device, i.e. first spatial distribution-based geometrical parameters – [0006]);
Nilakantan is silent with regards to the language of “obtaining, by the one or more hardware processors, the data associated with the first spatial distribution-based geometrical parameters of the one or more historical products, and the one or more defect types and locations, of the one or more historical products;
determining, by the one or more hardware processors, a correlation between the data associated with the first spatial distribution-based geometrical parameters of the one or more historical products and the one or more defect types and locations, of the one or more historical products; and
training, by the one or more hardware processors, the machine learning model based on the correlation between the first spatial distribution-based geometrical parameters of the one or more historical products and the one or more defect types and locations, of the one or more historical products.”
Srivastava further teaches “obtaining, by the one or more hardware processors, the data associated with the first spatial distribution-based geometrical parameters of the one or more historical products, and the one or more defect types and locations, of the one or more historical products (“As a training set, the data access engine 108 may aggregate part geometries [i.e. first spatial distribution-based geometrical parameters] and defect locations of other part designs and provide the training set to a machine-learning algorithm to generate a machine-learning model that generates defect probabilities for an input part design or toolpath” – [0030]; parameters, databases, and other data structures may be separately stored and managed – [0054]);
determining, by the one or more hardware processors, a correlation between the data associated with the first spatial distribution-based geometrical parameters of the one or more historical products and the one or more defect types and locations, of the one or more historical products (“In some implementations, the data access engine 108 may determine interest regions 262 via machine-learning. As a training set [i.e. of the one or more historical products], the data access engine 108 may aggregate [i.e. correlate] part geometries [i.e. first spatial distribution-based geometrical parameters] and defect locations of other part designs and provide the training set to a machine-learning algorithm to generate a machine-learning model that generates defect probabilities for an input part design or toolpath. Portions of an input toolpath that exceed a threshold defect probability may be identified by the data access engine 108 as interest regions 262 . As a specific example, the data access engine 108 may provide training data comprising part geometries and defect locations of other part designs to train a neural network and input a part design (e.g., the CAD model 212 ) to the neural network to identify interest regions 262 along the nominal toolpath 222” – [0030]); and
training, by the one or more hardware processors, the machine learning model based on the correlation between the first spatial distribution-based geometrical parameters of the one or more historical products and the one or more defect types and locations, of the one or more historical products (As a specific example, the data access engine 108 may provide training data comprising part geometries and defect locations of other part designs to train a neural network and input a part design (e.g., the CAD model 212 ) to the neural network to identify interest regions 262 along the nominal toolpath 222” – [0030]).”
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify Nilakantan in view of Srivastava to incorporate the further teaching of Srivastava to training data that includes the geometry and location of defects. By utilizing training data for machine learning with the geometry and location of defects, this yields predictable results in the determination of defects in a part.
In regards to Claims 8 and 17, Nilakantan in view of Srivastava discloses the claimed invention as detailed above. Nilakantan further teaches “the machine learning model is configured to one of:
determine quality of the one or more products based on the one or more first geometry data associated with the one or more historical products, and provide optimized parameters of one or more components of at least one of: the one or more products and the one or more historical products, to reduce defects based on at least one of: the one or more components, the one or more first geometry data associated with the one or more historical products, and the data associated with the one or more defect types and locations (machine learning based additive manufacturing process takes an input of user experience database with the additive manufacturing data fields 414 as input into the machine learning module 418 to output part optimization information 422 - [0042], Figure 4).”
In regards to Claim 10, Nilakantan teaches “one or more hardware processors (processor [0006]); and
a memory coupled to the one or more hardware processors, wherein the memory comprises a set of program instructions in form of a plurality of subsystems, configured to be executed by the one or more hardware processors (system with instructions to be performed by the processor – [0006]), wherein the plurality of subsystems comprises:
a data obtaining subsystem configured to obtain one or more experimental data from a machine, wherein the one or more experimental data comprise at least one of: recipe data, one or more measured parameters, data associated with one or more defect types and locations, and metadata (additive manufacturing knowledge base provides additive manufacturing fabrication parameters and measured/recorded additive manufacturing production outcome data – [0025]);
the data obtaining subsystem configured to obtain one or more first geometry data associated with one or more historical products, wherein each first geometry data associated with a respective historical product, and wherein each historical product is associated with corresponding experimental data of the one or more experimental data obtained from the machine (a provided CAD file for a part may be in a Standard for the Exchange of Product model data file format that will require conversion to a stereolithography STL file format – [0047]; part designer has a CAD model where the CAD model is passed on to the additive manufacturing operator – [0049]; machine learning model record information with the STL model – [0057]; machine learning model trained with historical records – [0023]; proof of work historical record – [0058]);
a parameter computing subsystem configured to:
compute, by a geometry model, one or more first geometrical parameters based on the one or more first geometry data associated with the one or more historical products, wherein the one or more first geometrical parameters comprise at least one of: first single-valued geometrical parameters and first spatial distribution-based geometrical parameters, wherein the first spatial distribution-based geometrical parameters represent spatial distribution of geometry across different regions of the one or more historical products (“Machine-learning model 108 , 214 , 306 , or 418 can thus record information generally relating to part programming, on the one hand, and specific machine variables and/or specific software variables on the other hand. In this context, part programming can consist of (1) configuring the 3D printer; (2) orienting the STL model (CAD converted to a stereolithography format); (3) “slicing” the STL model by intersecting the STL model with a series of horizontal planes to create slice curves; (4) creating support curves, defining where temporary supports will be built in the part, which supports will ultimately be disposed of; (5) creating toolpath fill for model and support curves; (6) saving a toolpath file; and (7) downloading the toolpath file to the printer for part building. Specific machine variables and/or specific software variables can consist of fabrication-machine-specific calibration values; head motor currents; head purge and control parameters; autoload parameters; stepper motor currents; XYZ axis speed parameters; purge and tip wipe locations; head temperature setback and no-ooze parameters; end of curve and purge following error monitoring; head and chamber temperatures used during model build; head velocities; distance between slices (in inches) [i.e. first geometrical parameters comprising spatial distribution of geometry across different regions of the one or more historical products]; minimum and maximum speeds to move the head (in inches per second); origin: X & Y coordinate location to start the model; clearance move: amount to raise the head during glueless moves (in inches); delay time and encoder tick rate (in milliseconds, flowrate); fluid relaxation time during acceleration/deceleration; desired flow accuracy during acceleration/deceleration. In turn, the machine-learning model 410 can produce the part optimization and command initiation outputs 422 , 424 based on training from the above parameters [i.e. one or more first geometrical parameters comprising first single-valued geometrical parameters and first spatial distribution-based geometrical parameters] as stored in the user experience database 410 (which can take the form of a blockchain) and the input vector 416 . In other words, the machine-learning model can be capable of performing the part programming and the parameter setting in an automated way for a new part job based on the past part job data” – [0057]; “An example operation of a machine-learning-based additive manufacturing system according to the flow diagram of FIG. 4 is not described. New part transactions 412 are broadcast to the machine-learning model 418 . The proof-of-work historical record of the user experience database 410 compiles new part transaction data into a block. The proof-of-work historical record becomes more resilient, richer with knowledge, and iteratively stronger with each block addition and feeds the machine-learning model 418 , which is iteratively re-trained either with each block addition or periodically. The machine-learning model 418 provides real-time feedback on the next part transaction based on history. The machine-learning model 418 utilizes the notion of similarity and creates a mapping between, on the one hand, voxel data, orientation data, geometry data, technical language data, and fabrication device variables and parameters, and, on the other hand, the part to be 3D printed. These mappings can, for example, be represented within a structured probability distribution. The proof-of-work historical record 410 , together with the machine-learning model 418 , learns probable model orientation, support structure, toolpaths, and boundary curves based on input. The machine-learning model 418 utilizes neural networks and pattern recognition to make suggestions, predictions, and warnings. The machine-learning model 418 provides outputs 422 , 424 that optimize user/object interaction” – [0058]); and
compute, by the geometry model, one or more second geometrical parameters based on one or more second geometry data associated with the one or more products, wherein the one or more second geometrical parameters comprises at least one of: second single-valued geometrical parameters and second spatial distribution-based geometrical parameters (“Machine-learning model 108 , 214 , 306 , or 418 can thus record information generally relating to part programming, on the one hand, and specific machine variables and/or specific software variables on the other hand. In this context, part programming can consist of (1) configuring the 3D printer; (2) orienting the STL model (CAD converted to a stereolithography format); (3) “slicing” the STL model by intersecting the STL model with a series of horizontal planes to create slice curves; (4) creating support curves, defining where temporary supports will be built in the part, which supports will ultimately be disposed of; (5) creating toolpath fill for model and support curves; (6) saving a toolpath file; and (7) downloading the toolpath file to the printer for part building. Specific machine variables and/or specific software variables can consist of fabrication-machine-specific calibration values; head motor currents; head purge and control parameters; autoload parameters; stepper motor currents; XYZ axis speed parameters; purge and tip wipe locations; head temperature setback and no-ooze parameters; end of curve and purge following error monitoring; head and chamber temperatures used during model build; head velocities; distance between slices (in inches); minimum and maximum speeds to move the head (in inches per second); origin: X & Y coordinate location to start the model [i.e. second spatial distribution-based geometrical parameters]; clearance move: amount to raise the head during glueless moves (in inches); delay time and encoder tick rate (in milliseconds, flowrate); fluid relaxation time during acceleration/deceleration; desired flow accuracy during acceleration/deceleration. In turn, the machine-learning model 410 can produce the part optimization and command initiation outputs 422 , 424 based on training from the above parameters [i.e. one or more second geometrical parameters comprising second single-valued geometrical parameters and second spatial distribution-based geometrical parameters]as stored in the user experience database 410 (which can take the form of a blockchain) and the input vector 416 . In other words, the machine-learning model can be capable of performing the part programming and the parameter setting in an automated way for a new part job based on the past part job data” – [0057]; “An example operation of a machine-learning-based additive manufacturing system according to the flow diagram of FIG. 4 is not described. New part transactions 412 are broadcast to the machine-learning model 418 . The proof-of-work historical record of the user experience database 410 compiles new part transaction data into a block. The proof-of-work historical record becomes more resilient, richer with knowledge, and iteratively stronger with each block addition and feeds the machine-learning model 418 , which is iteratively re-trained either with each block addition or periodically. The machine-learning model 418 provides real-time feedback on the next part transaction based on history. The machine-learning model 418 utilizes the notion of similarity and creates a mapping between, on the one hand, voxel data, orientation data, geometry data, technical language data, and fabrication device variables and parameters, and, on the other hand, the part to be 3D printed. These mappings can, for example, be represented within a structured probability distribution. The proof-of-work historical record 410 , together with the machine-learning model 418 , learns probable model orientation, support structure, toolpaths, and boundary curves based on input. The machine-learning model 418 utilizes neural networks and pattern recognition to make suggestions, predictions, and warnings. The machine-learning model 418 provides outputs 422 , 424 that optimize user/object interaction” – [0058]; user experience database includes data defining requirements and specifications, with a fabrication slicing resolution, i.e. second single-valued geometrical parameters – [0006]); and
a defect determining subsystem configured to determine, using a machine learning model trained using data associated with the one or more historical products, to the one or more defect regions of the one or more products, wherein the determination is based on at least one of: the one or more computed statistical features associated with the one or more defect (machine learning model trained with the accumulated knowledge and applied to new additive manufacturing projects and tasks and include knowledge including defects and flaws – [0022]; machine learning training involves a statistical aggregation of training data – [0034]), the one or more first geometrical parameters, and the one or more second geometrical parameters (machine learning based additive manufacturing system and method can automatically incorporate lessons learned from previous part builds to prevent the repeating of errors and the machine learning model serves as a predictive tool for potential flaws or defects that occur n the manufacturing process – [0068]; Figure 4 shows the user experience database 410 along with the database 414 and new project input vector 416 as inputs to the machine learning model 418);
a parameter generation subsystem configured to generate optimized parameters for the machine based on the determined one or more defect regions, wherein the optimized parameters are suggested by the machine learning model to reduce defects at one or more locations in the manufacturing process (“specifically to methods and systems for machine-learning-based additive manufacturing using manufacturing data” – [0001]; “Such an example can include one or more non-transitory computer-readable media storing instructions that when executed by a computer processor, cause the processor to process, with a machine-learning model trained on a user experience database comprising a plurality of entries, an input vector describing a new part transaction to provide at least one part optimization output and at least one command initiation output to configure additive manufacturing of a new part. Entries in the user experience database each include at least data defining requirements for an additively manufactured part previously fabricated or attempted to be fabricated, specifications describing an additive manufacturing fabrication device, a selection of a raw material type fed to the fabrication device for fabrication of the additively manufactured part, a fabrication spatial orientation of the additively manufactured part within the fabrication device, a fabrication slicing resolution of the additively manufactured part, and a toolpath taken by the fabrication device in fabricating the additively manufactured part” – [0007]; “A feature extractor 106 generates a feature vector that includes at least one of the set of coefficients generated at the pattern decomposition component 104 . A machine learning model 108 determines at least one prefabrication model adjustment parameter and/or print setting parameter for a new print job and/or at least one print job outcome estimate from the metric. As examples, not meant to be an exhaustive list, the prefabrication model adjustment parameter can represent a print orientation of a component or assembly to be printed in a print job, a number of layers to be used or a layer thickness or thicknesses, a geometry parameter descriptive of how a three-dimensional model is divided for printing during a prefabrication model processing phase, or information controlling the placement or geometry of support structures added to the model during a prefabrication model processing phase. As examples, not meant to be an exhaustive list, the print setting parameter can represent a material to be used for the job, a particular printer or model of printer to be used, a nozzle aperture thickness or temperature, or any of a number of other parameters that may be accessible for modification on a 3D printer or other additive manufacturing fabrication device. As examples, not meant to be an exhaustive list, the outcome estimate can be an estimate of print time for the job, an estimate of raw material usage for the print job, or an estimate of the likelihood of success of a print job. The prefabrication model adjustment parameter or print setting parameter provide or contribute to part optimization or command initiation, respectively. The prefabrication model adjustment parameter, print setting parameter, and/or outcome estimate provided by the machine learning model 108 can be stored on a non-transitory computer-readable medium associated with the system 100 and/or provided to a user at a display via a user interface (not shown in FIG. 1, but see FIG. 2)” – [0027]; “The machine-learning-based additive manufacturing systems and methods described herein can automatically incorporate lessons learned from previous part builds to prevent the repeating of errors and to enable affordability, reliability, and maintainability in an additive manufacturing production system. The machine learning model of the described systems and methods can also serve as a predictive tool for potential flaws or defects that may occur in the manufacturing process” – [0068]).”
Nilakantan is silent with regards to the language of “a feature computing subsystem configured to compute, by a feature engineering model, one or more statistical features based on the data associated with one or more defect types and locations of the one or more experimental data obtained from the machine; a defect determining subsystem configured to determine the one or more defect regions in at least one of: the one or more products, based on at least one of: the one or more computed statistical features associated with the one or more defect types and locations, the one or more first geometrical parameters, and the one or more second geometrical parameters, wherein the defect determination subsystem is further configured to identify correlations between the spatial distribution-based geometrical parameters and the one or more defect types and locations.”
Srivastava teaches “a feature computing subsystem configured to compute, by a feature engineering model, one or more statistical features based on the data associated with one or more defect types and locations of the one or more experimental data obtained from the machine (“As yet another example, the data access engine 108 may identify specific portions along a nominal toolpath 222 at which a manufacturing defect has a higher probability of occurrence. To determine defect probabilities, the data access engine 108 may perform statistical analyses on any number of part designs as well as identified defects in physical parts manufactured from the part designs. Analyzed part designs may include similarly structured part designs that are manufactured to the same, similar, or common manufacturing processes. Through such statistical analysis processes, the data access engine 108 may produce a statistical model [i.e. feature engineering model] providing defect probabilities for specific part characteristics, attributes, shapes, structures, positions, or according to any other delineating part feature. In some instances, the data access engine 108 may perform such a statistical analysis on part geometries and defect locations of other part designs (besides the CAD model 212 ) to determine probabilistic defect locations, which may refer to any part location with a defect probability that exceeds a probability threshold (e.g., greater than 45% or any other configurable threshold). Accordingly, the data access engine 108 may identify interest regions 262 as probabilistic defect locations determined via statistical analysis” – [0029]); a defect determining subsystem configured to determine the one or more defect regions in at least one of: the one or more products, based on at least one of: the one or more computed statistical features associated with the one or more defect types and locations, the one or more first geometrical parameters, and the one or more second geometrical parameters, wherein the defect determination subsystem is further configured to identify correlations between the spatial distribution-based geometrical parameters and the one or more defect types and locations (“As yet another example, the data access engine 108 may identify specific portions along a nominal toolpath 222 at which a manufacturing defect has a higher probability of occurrence. To determine defect probabilities, the data access engine 108 may perform statistical analyses on any number of part designs as well as identified defects in physical parts manufactured from the part designs. Analyzed part designs may include similarly structured part designs that are manufactured to the same, similar, or common manufacturing processes. Through such statistical analysis processes, the data access engine 108 may produce a statistical model providing defect probabilities for specific part characteristics, attributes, shapes, structures, positions, or according to any other delineating part feature [i.e. first geometrical parameters and second geometrical parameters]. In some instances, the data access engine 108 may perform such a statistical analysis on part geometries and defect locations of other part designs (besides the CAD model 212 ) to determine probabilistic defect locations [i.e. defect region], which may refer to any part location with a defect probability that exceeds a probability threshold (e.g., greater than 45% or any other configurable threshold). Accordingly, the data access engine 108 may identify interest regions 262 as probabilistic defect locations determined via statistical analysis [i.e. correlating]” – [0029]).”
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify Nilakantan to incorporate the teaching of Srivastava to utilize statistical analysis of data relating to the part being manufactured to determine defects. By utilizing statistical analysis this yields predictable results in the determination of defects in a part.
Claims 4-5 and 13-14 are rejected under 35 U.S.C. 103 as being unpatentable over Nilakantan in view of Srivastava as applied to claim 1 and 10 above, and further in view of Abad (US20200192990) and Jain (US20220172258).
In regards to Claims 4 and 13, Nilakantan in view of Srivastava discloses the claimed invention as detailed above. Nilakantan further teaches “obtaining, by the one or more hardware processors, the one or more first geometry data associated with the one or more historical products (machine learning model record information with the STL model – [0057]; machine learning model trained with historical records, i.e. one or more first geometry data associated with historical products – [0023]; proof of work historical record – [0058]), and the one or more second geometry data associated with the one or more products, at the geometry model (a provided CAD file for a part may be in a Standard for the Exchange of Product model data file format that will require conversion to a stereolithography STL file format – [0047]; part designer has a CAD model where the CAD model is passed on to the additive manufacturing operator – [0049]),
computing, by the one or more hardware processors, the first one or more first geometrical parameters and the one or more second geometrical parameters, based on at least one of: the one or more first geometry data associated with the one or more historical products and the one or more second geometry data associated with the one or more products, wherein the one or more second geometrical parameters comprises at least one of: second single-valued geometrical parameters and second spatial distribution-based geometrical parameters (“Machine-learning model 108 , 214 , 306 , or 418 can thus record information generally relating to part programming, on the one hand, and specific machine variables and/or specific software variables on the other hand. In this context, part programming can consist of (1) configuring the 3D printer; (2) orienting the STL model (CAD converted to a stereolithography format); (3) “slicing” the STL model by intersecting the STL model with a series of horizontal planes to create slice curves; (4) creating support curves, defining where temporary supports will be built in the part, which supports will ultimately be disposed of; (5) creating toolpath fill for model and support curves; (6) saving a toolpath file; and (7) downloading the toolpath file to the printer for part building. Specific machine variables and/or specific software variables can consist of fabrication-machine-specific calibration values; head motor currents; head purge and control parameters; autoload parameters; stepper motor currents; XYZ axis speed parameters; purge and tip wipe locations; head temperature setback and no-ooze parameters; end of curve and purge following error monitoring; head and chamber temperatures used during model build; head velocities; distance between slices (in inches); minimum and maximum speeds to move the head (in inches per second); origin: X & Y coordinate location to start the model; clearance move: amount to raise the head during glueless moves (in inches); delay time and encoder tick rate (in milliseconds, flowrate); fluid relaxation time during acceleration/deceleration; desired flow accuracy during acceleration/deceleration. In turn, the machine-learning model 410 can produce the part optimization and command initiation outputs 422 , 424 based on training from the above parameters as stored in the user experience database 410 (which can take the form of a blockchain) and the input vector 416 . In other words, the machine-learning model can be capable of performing the part programming and the parameter setting in an automated way for a new part job based on the past part job data” – [0057]; “An example operation of a machine-learning-based additive manufacturing system according to the flow diagram of FIG. 4 is not described. New part transactions 412 are broadcast to the machine-learning model 418 . The proof-of-work historical record of the user experience database 410 compiles new part transaction data into a block. The proof-of-work historical record becomes more resilient, richer with knowledge, and iteratively stronger with each block addition and feeds the machine-learning model 418 , which is iteratively re-trained either with each block addition or periodically. The machine-learning model 418 provides real-time feedback on the next part transaction based on history. The machine-learning model 418 utilizes the notion of similarity and creates a mapping between, on the one hand, voxel data, orientation data, geometry data, technical language data, and fabrication device variables and parameters, and, on the other hand, the part to be 3D printed. These mappings can, for example, be represented within a structured probability distribution. The proof-of-work historical record 410 , together with the machine-learning model 418 , learns probable model orientation, support structure, toolpaths, and boundary curves based on input. The machine-learning model 418 utilizes neural networks and pattern recognition to make suggestions, predictions, and warnings. The machine-learning model 418 provides outputs 422 , 424 that optimize user/object interaction” – [0058]; user experience database includes the fabrication spatial orientation of the additively manufactured part within the fabrication device, i.e. first spatial distribution-based geometrical parameters, and data defining requirements and specifications, with a fabrication slicing resolution, i.e. second single-valued geometrical parameters – [0006]);
storing, by the one or more hardware processors, data associated with at least one of: the first single-valued geometrical parameters and the first spatial distribution-based geometrical parameters, for the one or more first geometry data associated with the one or more historical products, and the second single-valued geometrical parameters and the second spatial distribution-based geometrical parameters, for the one or more second geometry data associated with the one or more products (the accumulated knowledge relating to the system and to allow re-use of are stored as knowledge in a database, i.e. storing data associated with the geometrical parameters – [0022]).”
Nilakantan in view of Srivastava is silent with regards to the language of “computing, by the geometry model, a similarity index between the one or more first geometry data associated with the one or more historical products, and the one or more second geometry data associated with the one or more products, wherein computing the similarity index between the one or more first geometry data associated with the one or more historical products, and the one or more second geometry data associated with the one or more products, computing, by the one or more hardware processors, the similarity index between the one or more historical products and the one or more products, based on the at least one of: the first single-valued geometrical parameters stored for the one or more first geometry data associated with the one or more historical products, and the second single-valued geometrical parameters stored for the one or more second geometry data associated with the one or more products; selecting, by the one or more hardware processors, at least one historical product among the one or more historical products similar to the one or more products;
computing, by the one or more hardware processors, the similarity index between the selected at least one historical product and the one or more products, based on the at least one of: the first spatial distribution-based geometrical parameters and the second spatial distribution-based geometrical parameters applied between the selected at least one historical product and the one or more current products”
Abad teaches “computing, by the geometry model, a similarity index between the one or more first geometry data associated with the one or more historical products, and the one or more second geometry data associated with the one or more products, wherein computing the similarity index between the one or more first geometry data associated with the one or more historical products, and the one or more second geometry data associated with the one or more products (the similarity comparator is configured to perform a comparison of the CAD model and the FEM model to determine a similarity between the first geometry and the second geometry and the similarity comparator can compare the shapes and/or dimensions between the first geometry and the second geometry to determine the similarity – [0034]), computing, by the one or more hardware processors, the similarity index between the one or more historical products and the one or more products, based on the at least one of: the first single-valued geometrical parameters stored for the one or more first geometry data associated with the one or more historical products, and the second single-valued geometrical parameters stored for the one or more second geometry data associated with the one or more products (the similarity comparator is configured to perform a comparison of the CAD model and the FEM model to determine a similarity between the first geometry and the second geometry and the similarity comparator can compare the shapes and/or dimensions between the first geometry and the second geometry to determine the similarity – [0034]), selecting, by the one or more hardware processors, at least one historical product among the one or more historical products similar to the one or more products (the similarity comparator is configured to perform a comparison of the CAD model and the FEM model to determine a similarity – [0034]);
computing, by the one or more hardware processors, the similarity index between the selected at least one historical product and the one or more products, based on the at least one of: the first spatial distribution-based geometrical parameters and the second spatial distribution-based geometrical parameters applied between the selected at least one historical product and the one or more current products (the similarity comparator is configured to perform a comparison of the CAD model and the FEM model to determine a similarity between the first geometry and the second geometry and the similarity comparator can compare the shapes and/or dimensions between the first geometry and the second geometry to determine the similarity – [0034]).”
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify Nilakantan in view of Srivastava to incorporate the teaching of Abad to utilize the similarity comparator to determine the similarity between different model and geometries. By performing this structural analysis with the geometrical models this is an improvement that reduces the errors generated in the process and reduces the operation time for the evaluation.
Nilakantan in view of Srivastava and Abad is silent with regards to the language of “wherein the similarity index between the one or more historical products and the one or more products, is computed based on Euclidean distance; determining, by the one or more hardware processors, a historical product similar to the one or more products based on the computed similarity index between the selected at least one historical product and the one or more products;”
Jain teaches “wherein the similarity index between the one or more historical products and the one or more products, is computed based on Euclidean distance (Similarity index calculated based on a Euclidean distance – [0069]); determining, by the one or more hardware processors, a historical product similar to the one or more products based on the computed similarity index between the selected at least one historical product and the one or more products (the similarity index identify is used to find similar products from an existing product portfolio and identifies commonality between the products – [0070])”
It would have been obvious to one of ordinary skill in the art before he effective filing date of the claimed invention to modify Nilakantan in view of Srivastava and Abad to incorporate the teaching of Jain to perform a similarity index analysis between products by utilizing the Euclidean distance. By utilizing the Euclidean distance for the similarity index this is an improvement that yields predictable results in the comparison of products.
In regards to Claims 5 and 14, Nilakantan in view of Srivastava, Abad, and Jain discloses the claimed invention as detailed above. Nilakantan is silent with regards to the language of “wherein the at least one of: the first one or more first geometrical parameters and the one or more second geometrical parameters, comprises at least one of: surface-to-volume, mass, crinkliness, compactness, volume of at least one of: the one or more historical products and the one or more products, bounding box volume of at least one of: the one or more historical products and the one or more products, and principal moment in the at least three directions.”
Jain further teaches “wherein the at least one of: the first one or more first geometrical parameters and the one or more second geometrical parameters, comprises at least one of: surface-to-volume, mass, crinkliness, compactness, volume of at least one of: the one or more historical products and the one or more products, bounding box volume of at least one of: the one or more historical products and the one or more products, and principal moment in the at least three directions (process with machine learning utilizing the product volumes – [0072]).”
It would have been obvious to one of ordinary skill in the art before he effective filing date of the claimed invention to modify Nilakantan in view of Srivastava, Abad, and Jain to incorporate the further teaching of Jain to process the volume of the products for a geometrical parameter. By utilizing the volume of the product in the model, this is an improvement that yields predictable results in the evaluation and accuracy of the similarity determination.
Allowable Subject Matter
Claim 2-3, 7, 9, 11-12, 16, and 18 would be allowable if rewritten to overcome the rejection(s) under 35 U.S.C. 112(b) or 35 U.S.C. 112 (pre-AIA ), 2nd paragraph, set forth in this Office action and to include all of the limitations of the base claim and any intervening claims.
In regards to Claims 2 and 11, Nilakantan in view of Srivastava discloses the claimed invention as detailed above. Nilakantan further teaches “wherein computing, by the geometry model, the first one or more first geometrical parameters based on the one or more first geometry data associated with the one or more historical products, comprises:
obtaining, by the one or more hardware processors, the one or more first geometry data associated with the one or more historical products, in one or more formats, wherein the one or more formats comprises at least one of: an initial graphics exchange specification (IGES) format, a stereolithography (STL) format, a standard for the exchange of product data (STEP) format, and a computer aided design (CAD) format, and wherein the one or more first geometry data associated with the one or more historical products, are obtained at the geometry model in computer aided design model (“Machine-learning model 108 , 214 , 306 , or 418 can thus record information generally relating to part programming, on the one hand, and specific machine variables and/or specific software variables on the other hand. In this context, part programming can consist of (1) configuring the 3D printer; (2) orienting the STL model (CAD converted to a stereolithography format); (3) “slicing” the STL model by intersecting the STL model with a series of horizontal planes to create slice curves; (4) creating support curves, defining where temporary supports will be built in the part, which supports will ultimately be disposed of; (5) creating toolpath fill for model and support curves; (6) saving a toolpath file; and (7) downloading the toolpath file to the printer for part building.” – [0057]; “The proof-of-work historical record 410 , together with the machine-learning model 418 , learns probable model orientation, support structure, toolpaths, and boundary curves based on input. The machine-learning model 418 utilizes neural networks and pattern recognition to make suggestions, predictions, and warnings. The machine-learning model 418 provides outputs 422 , 424 that optimize user/object interaction” – [0058]; user experience database includes the fabrication spatial orientation of the additively manufactured part within the fabrication device, i.e. first spatial distribution-based geometrical parameters – [0006]);
generating, by the one or more hardware processors, a three dimensional mesh for the one or more first geometry data associated with the one or more historical products (three-dimensional model generated – [0027]);
computing, by the one or more hardware processors, the first one or more first geometrical parameters for the one or more first geometry data associated with the one or more historical products, wherein the first one or more first geometrical parameters comprises at least one of: the first single-valued geometrical parameters and the first spatial distribution-based geometrical parameters (user experience database includes the fabrication spatial orientation of the additively manufactured part within the fabrication device, i.e. first spatial distribution-based geometrical parameters – [0006]);
storing, by the one or more hardware processors, data associated with the first one or more first geometrical parameters of the one or more first geometry data associated with the one or more historical products, in a database (user experience database includes the fabrication spatial orientation of the additively manufactured part – [0006]);
Srivastava teaches the limitations “generating, by the one or more hardware processors, a three dimensional mesh (surface mesh – [0021])”
Nilakantan in view of Srivastava is silent with regards to the language of “determining, by the one or more hardware processors, whether each of the first one or more first geometrical parameters uniquely identifies each of the one or more first geometry data associated with the one or more historical products;
computing, by the one or more hardware processors, third one or more geometrical parameters when each of the first one or more first geometrical parameters is distinct from each of the one or more first geometry data associated with the one or more historical products; and
storing, by the one or more hardware processors, the data associated with the first one or more first geometrical parameters for each of the first geometry data associated with the one or more historical products when each of the first one or more first geometrical parameters uniquely identifies each of the one or more geometry data associated with the one or more historical products.”
Claim 3 is dependent on Claim 2 and Claim 12 is dependent on Claim 11.
In regards to Claims 7 and 16, Nilakantan in view of Srivastava discloses the claimed invention as detailed above. Nilakantan further teaches “wherein determining, by the machine learning model, the one or more defect regions in at least one of: the one or more products and the one or more historical products, comprises:
obtaining, by the one or more hardware processors, the one or more second geometrical parameters computed for the one or more second geometry data associated with the one or more products, at the machine learning model (“Machine-learning model 108 , 214 , 306 , or 418 can thus record information generally relating to part programming, on the one hand, and specific machine variables and/or specific software variables on the other hand. In this context, part programming can consist of (1) configuring the 3D printer; (2) orienting the STL model (CAD converted to a stereolithography format); (3) “slicing” the STL model by intersecting the STL model with a series of horizontal planes to create slice curves; (4) creating support curves, defining where temporary supports will be built in the part, which supports will ultimately be disposed of; (5) creating toolpath fill for model and support curves; (6) saving a toolpath file; and (7) downloading the toolpath file to the printer for part building. Specific machine variables and/or specific software variables can consist of fabrication-machine-specific calibration values; head motor currents; head purge and control parameters; autoload parameters; stepper motor currents; XYZ axis speed parameters; purge and tip wipe locations; head temperature setback and no-ooze parameters; end of curve and purge following error monitoring; head and chamber temperatures used during model build; head velocities; distance between slices (in inches); minimum and maximum speeds to move the head (in inches per second); origin: X & Y coordinate location to start the model; clearance move: amount to raise the head during glueless moves (in inches); delay time and encoder tick rate (in milliseconds, flowrate); fluid relaxation time during acceleration/deceleration; desired flow accuracy during acceleration/deceleration. In turn, the machine-learning model 410 can produce the part optimization and command initiation outputs 422 , 424 based on training from the above parameters as stored in the user experience database 410 (which can take the form of a blockchain) and the input vector 416 . In other words, the machine-learning model can be capable of performing the part programming and the parameter setting in an automated way for a new part job based on the past part job data” – [0057]; “An example operation of a machine-learning-based additive manufacturing system according to the flow diagram of FIG. 4 is not described. New part transactions 412 are broadcast to the machine-learning model 418 . The proof-of-work historical record of the user experience database 410 compiles new part transaction data into a block. The proof-of-work historical record becomes more resilient, richer with knowledge, and iteratively stronger with each block addition and feeds the machine-learning model 418 , which is iteratively re-trained either with each block addition or periodically. The machine-learning model 418 provides real-time feedback on the next part transaction based on history. The machine-learning model 418 utilizes the notion of similarity and creates a mapping between, on the one hand, voxel data, orientation data, geometry data, technical language data, and fabrication device variables and parameters, and, on the other hand, the part to be 3D printed. These mappings can, for example, be represented within a structured probability distribution. The proof-of-work historical record 410 , together with the machine-learning model 418 , learns probable model orientation, support structure, toolpaths, and boundary curves based on input. The machine-learning model 418 utilizes neural networks and pattern recognition to make suggestions, predictions, and warnings. The machine-learning model 418 provides outputs 422 , 424 that optimize user/object interaction” – [0058]; user experience database includes data defining requirements and specifications, with a fabrication slicing resolution, i.e. second single-valued geometrical parameters – [0006]);
Nilakantan in view of Srivastava is silent with regards to the language of “comparing, by the one or more hardware processors, the one or more second geometrical parameters computed for the one or more second geometry data associated with the one or more products, with determined data associated with the correlation between the first spatial distribution-based geometrical parameters and the one or more defect types and locations, of the one or more historical products; and
determining, by the one or more hardware processors, the one or more defect regions in at least one of: the one or more products and the one or more historical products, based on the comparison between the one or more second geometrical parameters computed for the one or more second geometry data associated with the one or more products, with the determined data associated with the correlation between the first spatial distribution-based geometrical parameters and the one or more defect types and locations, of the one or more historical products.”
In regards to Claim 9 and 18, Nilakantan in view of Srivastava discloses the claimed invention as detailed above. Nilakantan in view of Srivastava is silent with regards to the language of “wherein: the recipe data comprise a first plurality of parameters, wherein the first plurality of parameters is set for the machine to manufacture the one or more historical products, and wherein the first plurality of parameters comprises at least one of: pressure, temperature, and flow rate.
the one or more measured parameters comprises a second plurality of parameters, wherein the second plurality of parameters is measured from the machine by one or more sensors, and wherein the second plurality of parameters comprises at least one of: pressure, temperature, and flow rate.
the data associated with the one or more defect types and locations, are obtained by at least one of: visual inspection, a non-destructive system including X-ray, and one or more mechanical testing systems, and
the metadata comprise at least one of: an identity of the one or more historical products, an identity of the machine, and timestamps.”
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.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to YOSSEF KORANG-BEHESHTI whose telephone number is (571)272-3291. The examiner can normally be reached Monday - Friday 10:00 am - 6:30 pm.
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, Catherine Rastovski can be reached at (571) 270-0349. 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.
/YOSSEF KORANG-BEHESHTI/Examiner, Art Unit 2857
Prosecution Insights