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 .
Status of Claims
Applicant's “Amendment” filed on 10/10/2025 has been considered.
Claims 1-2, 11-12 are amended.
Claims 8 and 18 are cancelled.
Claims 1-7, 9-17, 19-20 are currently pending and have been examined.
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 of this title, 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, 9-13, 15, 17, 19-20 are rejected under 35 U.S.C. 103 as being unpatentable over U.S. Patent Application No. 2022/0187847 A1 to Cella in view of U.S. Patent Application No. 2021/0182466 A1 to Le.
Regarding Claim 1, CELLA discloses a computer-implemented method for recommending a recipe to produce a product in a manufacturing process ([0257] agile manufacturing capabilities; [1724] a platform that prints, builds, or otherwise produces 3D parts and/or products at least in part using an additive manufacturing technique.), the computer-implemented method comprising:
obtaining, by one or more hardware processors, experimental data from a machine, wherein the experimental data comprise at least one of: recipe data, sensor data, and metadata of the machine, wherein the recipe data comprise a plurality of parameters that is set for the machine to produce the product; ([1756] The machine learning model 10213 may receive inputs of sensor data or other data as training data, including event data and state data related to one or more of the entities or assets, or other inputs noted above or throughout this disclosure. The sensor data input to the machine learning model 10213 may be used to train the machine learning model 10213 to perform the analytics, simulation, decision making, and/or predictive analytics relating to the data processing, data analysis, simulation creation, and/or simulation analysis of the one or more of the distributed manufacturing network entities or assets. [1725] calculate an optimal set of process parameters for printing or other additive manufacturing.)
generating, by the one or more hardware processors, a physics-based- simulation model ([1757] The digital twin provides one or more simulations of both physical elements and characteristics of the one or more distributed manufacturing network entities being replicated and the dynamics thereof,) based on the experimental data obtained from the machine; ([1756] The machine learning model 10213 may receive inputs of sensor data or other data as training data, including event data and state data related to one or more of the entities or assets, or other inputs noted above or throughout this disclosure. The sensor data input to the machine learning model 10213 may be used to train the machine learning model 10213 to perform the analytics, simulation, decision making, and/or predictive analytics relating to the data processing, data analysis, simulation creation, and/or simulation analysis of the one or more of the distributed manufacturing network entities or assets. [1810] In embodiment, optimization of workflows across a set of additive manufacturing entities may occur by having an artificial intelligence system undertake a set of simulations, such as simulations involving alternative scheduling sequences, design configurations, alternative output types, and the like. In embodiments, simulations may include sequences involving additive manufacturing and other manufacturing entities (such as subtractive manufacturing entities that cut, drill, or the like and/or finishing entities that polish, cure, or the like), including handoffs between sets of different manufacturing entity types, such as where handoffs are handled by robotic handling systems.)
wherein generating the physics- based-simulation model comprises: extracting, by the one or more hardware processors, simulation data from one or more sensors that are installed in the machine: ([1756] The machine learning model 10213 may receive inputs of sensor data or other data as training data, including event data and state data related to one or more of the entities or assets, or other inputs noted above or throughout this disclosure. The sensor data input to the machine learning model 10213 may be used to train the machine learning model 10213 to perform the analytics, simulation, decision making, and/or predictive analytics relating to the data processing, data analysis, simulation creation, and/or simulation analysis of the one or more of the distributed manufacturing network entities or assets.
determining, by the one or more hardware processors, an error between the extracted simulation data and experimental sensor data: ([0050] the models trained by the machine learning system are utilized by the artificial intelligence system to execute simulations on the part twin for predicting deformations or failure in a 3D printed part. In embodiments, the models may also determine a set or sequence of process control parameter adjustments that will implement a corrective action, e.g., to adjust a layer dimension or thickness, so as to correct a defect. In embodiments, the system may send a warning or error signal to an operator or a user, or automatically abort the printing process.)
generating, by the one or more hardware processors, synthetic data for a first plurality of recipes using the physics-based-simulation model, and wherein the first plurality of recipes is created based on physical ranges of each parameter from the plurality of parameters; ([0050] the models trained by the machine learning system are utilized by the artificial intelligence system to execute simulations on the part twin for predicting deformations or failure in a 3D printed part. In embodiments, the models may also determine a set or sequence of process control parameter adjustments (synthetic data) that will implement a corrective action, e.g., to adjust a layer dimension or thickness, so as to correct a defect. In embodiments, the system may send a warning or error signal to an operator or a user, or automatically abort the printing process. [2093] In some embodiments, a neural network model may be used directly to determine adjustments to optical parameters using training or learning of a neural network model. Initially, the model may be allowed to choose randomly from a range of values for each input optical control parameter or action. If the sequence of optical control parameter adjustments or actions leads to an incorrect prediction/classification, it may be scored as leading to an undesirable (or negative) outcome. Repetition of the process using different sets of randomly chosen values for each optical control parameter or action leads to reinforcement of those sequences that least to desirable (or positive) outcomes. Ultimately, the neural network model “learns” what adjustments to make to a set or sequence of optical control parameters or actions in order to achieve the target outcome i.e., a correct prediction or classification.)
determining, by the one or more hardware processors, an optimized physical range of each parameter from the physical ranges of each parameter by analyzing the experimental data and the synthetic data using a trained Al model; ([2336] , the modular AI-on-a-chip packages may be trained with models for optimizing 3D printing parameters.)
determining, by the one or more hardware processors, whether the optimized physical range of each parameter creating a second plurality of recipes is valid; ([1797] In some embodiments, a neural network model may be used directly to determine adjustments to process control parameters using training or learning of a neural network model. Initially, the model is allowed to choose randomly from a range of values for each input process control parameter or action. If the sequence of process control parameter adjustments or actions leads to a flaw or defect, it is scored as leading to an undesirable (or negative) outcome. Repetition of the process using different sets of randomly chosen values for each process control parameter or action leads to reinforcement of those sequences that least to desirable (or positive) outcomes. Ultimately, the neural network model “learns” what adjustments to make to a set or sequence of deposition process control parameters or actions in order to achieve the target outcome, i.e., a defect-free printed part.)
wherein determining whether the optimized physical range is valid comprises comparing the second plurality of recipes with a predetermined plurality of recipes created from the plurality of parameters using the process knowledge; ([1797] In some embodiments, a neural network model may be used directly to determine adjustments to process control parameters using training or learning of a neural network model. Initially, the model is allowed to choose randomly from a range of values for each input process control parameter or action. If the sequence of process control parameter adjustments or actions leads to a flaw or defect, it is scored as leading to an undesirable (or negative) outcome. Repetition of the process using different sets of randomly chosen values for each process control parameter or action leads to reinforcement of those sequences that least to desirable (or positive) outcomes. Ultimately, the neural network model “learns” what adjustments to make to a set or sequence of deposition process control parameters or actions in order to achieve the target outcome, i.e., a defect-free printed part. [0554] Examples of a behavior of a physical asset may include a state of matter of the physical asset (e.g., a solid, liquid, plasma or gas), a melting point of the physical asset, a density of the physical asset when in a liquid state, a viscosity of the physical asset when in a liquid state, a freezing point of the physical asset, a density of the physical asset when in a solid state, a hardness of the physical asset when in a solid state, the malleability of the physical asset, the buoyancy of the physical asset, the conductivity of the physical asset, electromagnetic properties of the physical asset, radiation properties, optical properties (e.g., reflectivity, transparency, opacity, albedo, and the like), wave interaction properties (e.g., transparency or opacity to radio waves, reflection properties, shielding properties, or the like), a burning point of the physical asset, the manner by which humidity affects the physical asset, the manner by which water or other liquids affect the physical asset, and the like.. )
generating, by the one or more hardware processors, recipes when the optimized physical range of each parameter creating the second plurality of recipes is valid; ([1806] the artificial intelligent system 10212 may adjust one or more features of the printer twin 10506 as a set of part twins 10504 are printed by the 3D printer. In embodiments, the artificial intelligent system 10212 may, for each set of features, execute a simulation based on the set of features and may collect the simulation outcome data resulting from the simulation. For example, in executing a simulation on the set of part twins 10504 being manufactured in the printer twin 10506, the artificial intelligent system 10212 can vary the properties of the printer twin 10506 and can execute simulations that generate outcomes. During the simulation, the artificial intelligent system 10212 may vary the ambient temperature, pressure, humidity, lighting, and/or any other properties of the printer twin 10506.)
validating, by the one or more hardware processors, the second plurality of recipes to extract an optimized recipe using the physics-based-simulation model; ([1810] In embodiment, optimization of workflows across a set of additive manufacturing entities may occur by having an artificial intelligence system undertake a set of simulations, such as simulations involving alternative scheduling sequences, design configurations, alternative output types, and the like. In embodiments, simulations may include sequences involving additive manufacturing and other manufacturing entities (such as subtractive manufacturing entities that cut, drill, or the like and/or finishing entities that polish, cure, or the like), including handoffs between sets of different manufacturing entity types, such as where handoffs are handled by robotic handling systems. [1678] the chip 9400 may initially analyze a particular data set (e.g., a data set received as inputs 9492) to determine whether one or more governance standards apply, as described in more detail below. Based on determining that one or more governance standards apply, the chip 9400 may then prioritize the applicable standards and generate and/or validate a model that enforces the governance standards.)
and recommending, by the one or more hardware processors, the optimized recipe for producing the product in the machine. ([1807] the machine-learning system 10210 trains one or more models that are utilized by the artificial intelligence system 10212 to make classifications, predictions, recommendations, and/or to generate or facilitate decisions or instructions relating to the product and the part, such as decisions or instructions governing design, configuration, material selection, shape selection, manufacturing type, job scheduling and many others. [1827] generating recommendations related to printing to a user of the platform. In embodiments, the recommendations may relate to a choice of a material for printing. In embodiments, the recommendations may relate to a choice of an additive manufacturing technique. In embodiments recommendations may relate to timing of manufacturing. [0786] The information technology system also has an artificial intelligence system that is configured to learn on the training set collected from the data sources, that simulates one or more attributes of one or more of the maritime assets, and that generates one or more sets of recommendations for a change in the one or more attributes based on the training set collected from the data sources.)
But does not explicitly disclose generating, by the one or more hardware processors, the second plurality of recipes; determining, by the one or more hardware processors, whether the error exceeds a threshold value; wherein the synthetic data is utilized only when the error is within the threshold value;
.
LE, on the other hand, teaches generating, by the one or more hardware processors, the second plurality of recipes ([0115]) simulating behavior of the self-adapting integrated circuit block over a range of input variables including at least one of: a process monitor input signal indicative of active device performance as a result of fabrication process variations; a voltage sensor signal indicative of the supply voltage; a temperature sensor signal indicative of the operating temperature; and, a changed performance specification; deriving from simulation data an artificial intelligence or machine learning model representing behavior of the self-adapting integrated circuit block over a range of such variables and digital self-adaptation control signals; calculating from the artificial intelligence or machine learning model the desired self-adaptation results; capturing as design information the trained artificial intelligence or machine learning model; and storing the design information in a design database.
LE, further teaches determining, by the one or more hardware processors, whether the error exceeds a threshold value; wherein the synthetic data is utilized only when the error is within the threshold value. ([0077] To verify the accuracy of the ML model, its inferred output is compared to that obtained from the simulation data for the same PVT condition. If the error is below a certain percentage, the ML model is deemed accurate. The process is repeated for a large number of PVT conditions. [0079] If the error is within a desired limit for all the verification data (713), the self-adaptive design is considered “locked (715),” meaning no further changes are needed. Otherwise, the ML model is further refined to bring the error to within the desired limits.)
It would have been obvious to one of ordinary skill in the art to include in the method, as taught by CELLA, the features, as taught by LE, since the claimed invention is merely a combination of old elements, and in the combination each element merely would have performed the same function as it did separately, and one of ordinary skill in the art would have recognized that the results of the combination were predictable. It further would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify CELLA, to include the teachings of LE, in order to infer appropriate changes using simulation (LE, [0001]).
Regarding Claim 2, CELLA in view of LE teaches the method of claim 1.
CELLA discloses wherein generating the physics-based-simulation model comprises: obtaining, by the one or more hardware processors, input data from the machine, wherein the input data comprise at least one of: geometries of the product in a form of computer aided design (CAD) model, ([1810] inputs indicating the item being serviced (e.g., technical specifications, CAD designs, and the like); [1882] the following part specifications and order requirements: A form or shape described by a 3D CAD solid model; Use-case loading as applied to the provided 3D CAD model; )
material properties of the product and one or more parts of the machine, ([1883] a material analysis that identifies titanium, Inconel, and 316 stainless steel as materials that meet corrosion requirements; a material analysis, assisted by simulations from the printer twin 10506 and the process and material selection twin 10702, that identifies powder bed fusion or metal material extrusion as 3D printing processes that match availability of the additive manufacturing units 10102;)
the plurality of parameters that are set for the machine for producing the product, and ([1782] printing process parameters such as nozzle orientation, flow rate and pressure, and the like)
initial and boundary conditions that are identified based on the experimental data and applied in the physics-based- simulation model, ([1797] In some embodiments, a neural network model may be used directly to determine adjustments to process control parameters using training or learning of a neural network model. Initially, the model is allowed to choose randomly from a range of values for each input process control parameter or action.)
wherein the geometries of the product comprise die and sub-parts of the machine which are assembled in the CAD model similar to an experimental setup of the product; ([1923] the product requirements may be a 3D printing instruction set including a file (e.g., a CAD file and/or an STL file) and any accompanying instructions for printing the product defined in the file.)
generating, by the one or more hardware processors, the physics-based- simulation model by providing initial values for the physical range of each parameter; ([1797] In some embodiments, a neural network model may be used directly to determine adjustments to process control parameters using training or learning of a neural network model. Initially, the model is allowed to choose randomly from a range of values for each input process control parameter or action.)
Regarding Claim 3, CELLA in view of LE teaches the method of claim 2.
However CELLA does not explicitly teach adjusting, by the one or more hardware processors, the value of the physical range of each parameter to generate the physics-based-simulation model when the error exceeds the threshold value.
LE, on the other hand, teaches adjusting, by the one or more hardware processors, the value of the physical range of each parameter to generate the physics-based-simulation model when the error exceeds the threshold value.. ([0077] To verify the accuracy of the ML model, its inferred output is compared to that obtained from the simulation data for the same PVT condition. If the error is below a certain percentage, the ML model is deemed accurate. The process is repeated for a large number of PVT conditions. [0079] If the error is within a desired limit for all the verification data (713), the self-adaptive design is considered “locked (715),” meaning no further changes are needed. Otherwise, the ML model is further refined to bring the error to within the desired limits.)
It would have been obvious to one of ordinary skill in the art to include in the method, as taught by CELLA, the features, as taught by LE, since the claimed invention is merely a combination of old elements, and in the combination each element merely would have performed the same function as it did separately, and one of ordinary skill in the art would have recognized that the results of the combination were predictable. It further would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify CELLA, to include the teachings of LE, in order to infer appropriate changes using simulation (LE, [0001]).
Regarding Claim 5, CELLA in view of LE teaches the method of claim 1.
CELLA discloses wherein validating the second plurality of recipes to extract the optimized recipe using the physics-based-simulation model comprises: simulating, by the one or more hardware processors, the second plurality of recipes to generate a plurality of outputs of the product from the machine, wherein the plurality of outputs comprises at least one of: a size, a shape and a location of a defect, a hot spot location, temperature distribution in a mold in the machine; ([1809] the artificial intelligence system 10212 may use a clustering algorithm to identify the failure pattern hidden in the failure data to train a model for detecting uncharacteristic or anomalous behavior. The failure data across multiple parts and their historical records may be clustered to understand how different patterns correlate to certain wear-down behavior. For example, if the failure happens early in the print, the failure may be due to uneven print surface. If the failure occurs later on in the print, it is likely that the part became detached from the printing surface and the cause of failure is poor bed adhesion and/or warping. All of the information gathered can be used as feedback for the model. [1792] The automated part and defect classification methods and systems of the present disclosure may be implemented using image sensors and/or machine vision systems. The machine vision systems may monitor the additive manufacturing process in real time, such as by capturing and analyzing images of the part or other item being printed. Automated image processing of the captured images may then be used to monitor any of a variety of part properties, e.g., dimensions (overall dimensions, or dimensions of specific features), feature angles, feature areas, surface finish (e.g., degree of light reflectivity, number of pits and/or scratches per unit area), and the like. The machine vision systems also track the process to detect any defects or errors in the printed part in real time while successive layers of materials are being deposited by the 3D printer.)
analyzing, by the one or more hardware processors, the plurality of outputs generated for the second plurality of recipes; and comparing, by the one or more hardware processors, the second plurality of recipes with the plurality of outputs to extract the optimized recipe from the second plurality of recipes. ([1793] Defects may be identified, e.g., by removing noise from the inspection data and subtracting a reference data set (e.g., a reference image of a defect-free part in the case that machine vision tools are being utilized for inspection), and classified using an unsupervised machine learning algorithm such as cluster analysis or an artificial neural network, to classify individual objects as either meeting or failing to meet a specified set of decision criteria (e.g., a decision boundary) in the feature space in which defects are being monitored. For example, a partially printed part may be compared with a render of the partial part and in case the partial part differs beyond a selected threshold from the render, the part may be classified as defective.)
Regarding Claim 7, CELLA in view of LE teaches the method of claim 1.
CELLA discloses wherein the first plurality of recipes are created based on the physical range of each parameter of the recipe using a process knowledge.; ([0554] Examples of a behavior of a physical asset may include a state of matter of the physical asset (e.g., a solid, liquid, plasma or gas), a melting point of the physical asset, a density of the physical asset when in a liquid state, a viscosity of the physical asset when in a liquid state, a freezing point of the physical asset, a density of the physical asset when in a solid state, a hardness of the physical asset when in a solid state, the malleability of the physical asset, the buoyancy of the physical asset, the conductivity of the physical asset, electromagnetic properties of the physical asset, radiation properties, optical properties (e.g., reflectivity, transparency, opacity, albedo, and the like), wave interaction properties (e.g., transparency or opacity to radio waves, reflection properties, shielding properties, or the like), a burning point of the physical asset, the manner by which humidity affects the physical asset, the manner by which water or other liquids affect the physical asset, and the like..)
Regarding Claim 9, CELLA in view of LE teaches the method of claim 1.
CELLA discloses wherein the recipe data comprise the plurality of parameters that are set for the machine to produce a type of the product, wherein the sensor data comprise data collected from the one or more sensors installed on the machine, and wherein the metadata comprise a label corresponding to at least one of: a defective or a non-defective part of the machine, a geometry, a location of the one or more sensors, a product type, information related to maintenance, environmental parameters, information related to replacing the part of the machine and information related to the machine, and a machine part. ([1797] In some embodiments, a neural network model may be used directly to determine adjustments to process control parameters using training or learning of a neural network model. Initially, the model is allowed to choose randomly from a range of values for each input process control parameter or action. [0708] The digital twin may predict the anticipated wear and failure of components of a system by reviewing historical and current operational data thereby reducing the risk of unplanned downtime and the need for scheduled maintenance. Instead of over-servicing or over-maintaining products to avoid costly downtime, repairs or replacement, any product performance issues predicted by the digital twin may be addressed in a proactive or just-in-time manner. [1754] the AI system is trained based on outcome factors, such as product quality and/or product defect outcomes, economic outcomes, on-time completion outcomes, and the like
Regarding Claim 10, CELLA in view of LE teaches the method of claim 1.
CELLA discloses wherein the experimental data and the synthetic data are inputted into a machine learning (ML) model to train the ML model. ([1881] Outcome data is provided to the machine learning system 10210 along with simulation, external, and training data to train or improve the initial machine learning model 10213.)
Claim 11 recites a system comprising substantially similar limitations as claim 1. The claim is rejected under substantially similar grounds as claim 1.
Claim 12 recites a system comprising substantially similar limitations as claim 2. The claim is rejected under substantially similar grounds as claim 2.
Claim 13 recites a system comprising substantially similar limitations as claim 3. The claim is rejected under substantially similar grounds as claim 3.
Claim 15 recites a system comprising substantially similar limitations as claim 5. The claim is rejected under substantially similar grounds as claim 5.
Claim 17 recites a system comprising substantially similar limitations as claim 7. The claim is rejected under substantially similar grounds as claim 7.
Claim 19 recites a system comprising substantially similar limitations as claim 9. The claim is rejected under substantially similar grounds as claim 9.
Claim 20 recites a system comprising substantially similar limitations as claim 10. The claim is rejected under substantially similar grounds as claim 10.
Claims 4, 14 are rejected under 35 U.S.C. 103 as being unpatentable over U.S. Patent Application No. 2022/0187847 A1 to Cella in view of U.S. Patent Application No. 2021/0182466 A1 to Le in view of U.S. Patent Application No. 2021/0208545 A1 to Zhang.
Regarding Claim 4, CELLA in view of LE teaches the method of claim 1.
CELLA discloses wherein determining the optimized physical range of each parameter from the physical ranges of each parameter using the trained AI model comprises: receiving, by the one or more processors, the experimental data and the synthetic data as an input at the trained AI model; ([1881] Outcome data is provided to the machine learning system 10210 along with simulation, external, and training data to train or improve the initial machine learning model 10213.)
determining, by the one or more hardware processors, one or more flaws using the trained AI model to provide recommendations for a defective product; ([1809] the artificial intelligence system 10212 may use a clustering algorithm to identify the failure pattern hidden in the failure data to train a model for detecting uncharacteristic or anomalous behavior. The failure data across multiple parts and their historical records may be clustered to understand how different patterns correlate to certain wear-down behavior. For example, if the failure happens early in the print, the failure may be due to uneven print surface. If the failure occurs later on in the print, it is likely that the part became detached from the printing surface and the cause of failure is poor bed adhesion and/or warping. All of the information gathered can be used as feedback for the model.)
extracting, by the one or more hardware processors, statistical features from the data comprising at least one of: mean, median, standard deviation, an area under curve from the one or more sensors; ([0000] an edge device 8042 may stream granular sensor data that is identified to be anomalous without compression, while the edge device 8042 may compress, summarize, or otherwise pass on a less granular data that is considered to be within a tolerance range of normal conditions or that reflects characteristics (e.g., statistical or signal characteristics) that suggest a lower likelihood that the data is likely to be of high interest.)
and recommending, by the one or more processors, the optimized recipe by analyzing the experimental data using the trained AI model.. ([1807] the machine-learning system 10210 trains one or more models that are utilized by the artificial intelligence system 10212 to make classifications, predictions, recommendations, and/or to generate or facilitate decisions or instructions relating to the product and the part, such as decisions or instructions governing design, configuration, material selection, shape selection, manufacturing type, job scheduling and many others. [1827] generating recommendations related to printing to a user of the platform. In embodiments, the recommendations may relate to a choice of a material for printing. In embodiments, the recommendations may relate to a choice of an additive manufacturing technique. In embodiments recommendations may relate to timing of manufacturing. [0786] The information technology system also has an artificial intelligence system that is configured to learn on the training set collected from the data sources, that simulates one or more attributes of one or more of the maritime assets, and that generates one or more sets of recommendations for a change in the one or more attributes based on the training set collected from the data sources.)
However CELLA does not explicitly teach performing, by the one or more processors, data cleaning and preparation processes, wherein the data cleaning and preparation processes comprise at least one of: removing outliers and handling missing data; extracting, by the one or more hardware processors, statistical features from the data comprising at least one of: mean, median, standard deviation, an area under curve from the one or more sensors; splitting, by the one or more processors, the extracted statistical features into a train set and a test set; training, by the one or more hardware processors, the AI model based on hyper-parameters of the AI model on the train set; evaluating, by the one or more hardware processors, the trained AI model on the test set;.
ZHANG, on the other hand, teaches performing, by the one or more processors, data cleaning and preparation processes, wherein the data cleaning and preparation processes comprise at least one of: removing outliers and handling missing data; ([0065] cleaned and processed real-time valid industrial process data (e.g., valid data range selected, normalized, denoised, outliers removed, data aligned across tags, missing data filled, and/or unnecessary data removed, etc.))
splitting, by the one or more processors, the extracted statistical features into a train set and a test set; ([0105] The model training controller 482 may perform validation test of forecast model 404. When a new version of forecast model 404 is trained, the performance metrics 467 of the new version of forecast model 404 tested on a validation data set is generated.)
training, by the one or more hardware processors, the AI model based on hyper-parameters of the AI model on the train set; ([0100] the machine learning module is a Python software package compatible to the model training interfaces. The Python software package contains data pre-processing and post-processing logics, TensorFlow code for training and inferencing, code for running an optimization search algorithm for optimizing the forecast model 404 (e.g., for optimizing the hyperparameter set) to predict the optimal hyperparameter set of the forecast model 404, and code for running an optimization search algorithm for searching the control space of the forecast model 404 to predict the optimal combination of control parameters and/or operation parameters of the industrial process for achieving one or more objectives.)
evaluating, by the one or more hardware processors, the trained AI model on the test set;. ([0105] the new version of forecast model 404 may pass absolute evaluation if the performance metrics 467 of the new version of forecast model 404 is within a threshold range, otherwise, the new forecast model 404 may not pass absolute evaluation, since it does not give accurate prediction.)
It would have been obvious to one of ordinary skill in the art to include in the method, as taught by CELLA and LE, the features, as taught by ZHANG, since the claimed invention is merely a combination of old elements, and in the combination each element merely would have performed the same function as it did separately, and one of ordinary skill in the art would have recognized that the results of the combination were predictable. It further would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the combination, to include the teachings of ZHANG, in order to find optimal suggestion for a given set of inputs. (ZHANG, [0065]).
Claim 14 recites a system comprising substantially similar limitations as claim 4. The claim is rejected under substantially similar grounds as claim 4.
Claims 6, 16 are rejected under 35 U.S.C. 103 as being unpatentable over U.S. Patent Application No. 2022/0187847 A1 to Cella in view of U.S. Patent Application No. 2021/0182466 A1 to Le in view of U.S. Patent No. 6298898 A1 to Mahadeva.
Regarding Claim 6, CELLA in view of LE teaches the method of claim 1.
CELLA discloses wherein the plurality of parameters comprises at least one of: molten metal temperature, pre- heat temperature, cooling channel parameters, heat transfer coefficient between the mold and a molten metal, the heat transfer coefficient between the mold and a cooling channel, pressure, and flow of air, which are set for the machine to produce the product, and ([1806] During the simulation, the artificial intelligent system 10212 may vary the ambient temperature, pressure, humidity, lighting, and/or any other properties of the printer twin 10506.)
However CELLA does not explicitly teach wherein the machine is a low pressure die casting (LPDC) machine.
ZHANG, on the other hand, teaches wherein the machine is a low pressure die casting (LPDC) machine. ([Col 4 Ln 25-35] To calibrate the finite element model for low pressure die casting against the experimental data, two phases are used: phase 1 for filling transients, and phase 2 for solidification. To simulate filling of the cavity, it is necessary to determine initial conditions for the solidification phase and an accurate fill time.)
It would have been obvious to one of ordinary skill in the art to include in the method, as taught by CELLA and LE, the features, as taught by Mahadeva, since the claimed invention is merely a combination of old elements, and in the combination each element merely would have performed the same function as it did separately, and one of ordinary skill in the art would have recognized that the results of the combination were predictable. It further would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the combination, to include the teachings of Mahadeva, in order to optimize casting quality. (Mahadeva, [Col 1 Ln 45-46]).
Claim 16 recites a system comprising substantially similar limitations as claim 6. The claim is rejected under substantially similar grounds as claim 6.
Response to Arguments
Applicant’s arguments with respect to rejection of the claim under 35 USC 103 have been considered but are moot in view of new grounds of rejection, necessitated by Applicant’s amendment.
Applicant argues that claims 1 and 11 recite a two-step validation process – a process knowledge feasibility check preceding a simulation-based validation – providing a distinct operational sequence not disclosed or suggested by the cited references.
Specifically, Mahadeva’s method neither calibrates simulation fidelity via error thresholds nor filters results through process knowledge before simulation.
Examiner disagrees. Regarding pre-simulation validation, Cella discloses wherein determining whether the optimized physical range is valid comprises comparing the second plurality of recipes with a predetermined plurality of recipes created from the plurality of parameters using the process knowledge; (Cella discloses [1797] repeating using different sets of values (recipes) for each input process control parameter or action and scoring based on positive or negative outcomes. Further, at [1520], Cella discloses In example embodiments, analytics produced by the analytics module 8818 may facilitate quantification of system performance as compared to a set of goals and/or metrics. The goals and/or metrics may be preconfigured, determined dynamically from operating results, and the like.
Further, LE teaches [0082] only meaningful relationships are modeled, and dominant components are taken into account in the creation of the ML model. In the process (step 803), correlations between input variables (PVT, dominant components, etc.) and output variables (current, voltage, etc.) are calculated from the simulation data (801). If the correlation between a certain input variable and the outputs is above a certain threshold (805), that variable is added to the “effective feature set” (809). The process is repeated for all the input variables and output variables then the results are used to build and train the ML model (811). This method gives a feature set with enough variables to make accurate prediction.
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).
Any inquiry concerning this communication or earlier communications from the examiner should be directed to Michelle T. Kringen whose telephone number is (571)270-0159. The examiner can normally be reached M-F: 11am-7pm.
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, Marissa Thein can be reached at (571)272-6764. 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.
/MICHELLE T KRINGEN/ Primary Examiner, Art Unit 3689
Prosecution Insights