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 .
Continued Examination Under 37 CFR 1.114
A request for continued examination under 37 CFR 1.114, including the fee set forth in 37 CFR
1.17(e), was filed in this application after final rejection. Since this application is eligible for continued examination under 37 CFR 1.114, and the fee set forth in 37 CFR 1.17(e) has been timely paid, the finality of the previous Office action has been withdrawn pursuant to 37 CFR 1.114. Applicant's submission filed on 01/21/2026 has been entered.
Response to Amendment
The amendment filed 01/21/2026 has been entered. As directed, claims 1, 2, 4, 6-7, 9-11, 16-17 and
19 have been amended, claim 3 has been canceled and no claim is added. Thus claims 1-2 and 4-20 remain pending in the application. However, new claim objection has been made based on the newly amendment.
Response to Arguments
With respect to the Applicant’s argued rejection under 35 U.S.C 101 in “Applicant Arguments/Remarks Made in an Amendment,”:
Applicant argues:
Claims 1-20 are rejected under 35 USC§ 101.
Applicant submits that in view of the amended claims above, the 101 rejections should now be withdrawn. In particular, Applicant submits that the amended claims are patent eligible under 35 U.S.C. §101 because the claims explicitly recite operating a manufacturing process (e.g., "producing, in a production compartment comprising a three-dimensional printer, a physical model of the structural model using the updated set of parameters" as recited in claim 1 and fully supported by at least paragraph [0036] of the original Application as filed). The computed "parameters" are not an "abstract result," rather, they are used to control a manufacturing apparatus to perform a physical transformation in the manner claimed. Therefore, the claimed subject matter integrates judicial exception into practical application under Step 2A- Prong 2, and is thus patent eligible under 35 U.S.C. §101.
(see Response filed 01/21/2026 [page 8]).
Applicant’s arguments with respect to claims 1-2 and 4-20 have been fully considered and are persuasive. The rejection under 35 U.S.C 101 of claims 1-2 and 4-20 has been withdrawn. The amended claim 1 recites limitations including “perform a topology optimization loop of the one or more design goals, … producing, in a production compartment comprising a three-dimensional printer, a physical model of the structural model using the updated set of parameters.” These limitations, when considered the claim as a whole, integrate judicial exception into a practical application by applying the optimization result to produce a physical structural model in an additive manufacturing environment. In particular, the claimed iterative feedback loop updates model parameters based on a convergence criterion, and the resulting parameters are used to generate a physical structural model, thereby improving the reliability of physical structures produced using additive manufacturing processes subject to thermal effects such as cracking and residual stress.
Accordingly, when considered claim as a whole, the additional limitations integrate judicial exception into practical application. Claim 9 and 16 recite similar limitations, are also considered to integrate the judicial exception into practical application. Therefore, the claims 1-2 and 4-20 are eligible under 35 U.S.C. § 101.
Applicant’s arguments, see pages 9-10, filed 01/21/2026, with respect to the rejection(s) under 35
U.S.C 103 of claim(s) 1, 9 and 16 under statutory basis for the previous rejection have been fully considered and are persuasive. Therefore, the rejection has been withdrawn. However, upon further consideration, a new ground(s) of rejection is made.
The previous cited reference:
Harris (US20230088537A1) teaches performing a generative design/topology optimization process in which a structural model is iteratively modified based on design variables, loading conditions, and constraints, using numerical simulation such as finite element analysis, to generate an optimized three dimensional structure for manufacturing.
The newly applied references:
Iyer (“Attention-Based 3D Neural Architectures for Predicting Cracks in Designs,” published in 2021) teaches using a deep neural network surrogate to predict a crack index for a structural design by estimating thermal and residual stresses arising during an additive manufacturing process for evaluating crack likelihood of the structural model.
Chen (“A New Topology Optimization Approach by Physics-Informed Deep Learning Process,” published in 2021) teaches a topology optimization process in which a neural network generates structural designs that are evaluated using a physics based finite element model, and the design parameters are iteratively updated using gradient information obtained through automatic differentiation to improve structural performance
Banga (“3D Topology Optimization Using Convolutional Neural Networks,” published in 2018) teaches generating multipole design instance for topology optimization by sampling parameters including volume fraction across a range, such that different volume fraction conditions are represented in the generated data used for training the network.
Therefore, the combination of Harris in view of Iyer and Chen and Banga together teach or suggest limitations of claims 1, 9 and 16. Therefore, the rejection of claims 1, 9 and 16 under 35 U.S.C. 103 is maintained.
Claim Objections
Claims 1, 5-6, 9 and 14-16 are objected to because of the following informalities:
Claim 1 recites “store one or more design goals for a structural model …” in line 6, should read as “store one or more design goals for the structural model …”.
Claims 9 and 16 also recite “a structural model”, and are objected for same reason.
Claim 5 recites “the state equation is based on at least: a performance function of the structural model under design load; a cracking index of the structural model considering thermal and residual stress during manufacturing; a weighting factor, and a target volume fraction.” should read as “the state equation is based on at least: a performance function of the structural model under design load; a cracking index of the structural model considering thermal and residual stress during manufacturing; a weighting factor; and a target volume fraction.”
Claim 14 also recites “,” and is objected for the same reason.
Claim 6 recites “The apparatus of claim 5, wherein the failure simulation of the structural model simulates an additive manufacturing process in the production compartment to:” should read as “The apparatus of claim 5, wherein the failure simulation of the structural model simulates the additive manufacturing process in the production compartment to:”.
Claim 15 also recites “an additive manufacturing process”, and is objected for the same reason.
Appropriate correction is required.
Claim Rejections - 35 USC § 103
In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and
103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis (i.e., changing from AIA to pre-AIA ) for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
Claim(s) 1-2, 7-12 and 16-20 are rejected under 35 U.S.C. 103 as being unpatentable over
Harris US20230088537A1 in view of Iyer (“Attention-Based 3D Neural Architectures for Predicting Cracks in Designs,” published in 2021) and Chen (“A New Topology Optimization Approach by Physics-Informed Deep Learning Process,” published in 2021) and Banga (“3D Topology Optimization Using Convolutional Neural Networks,” published in 2018).
Claim 1, Harris teaches An apparatus for manufacturing a structural model and computing a set of design parameters defining the structural model (Abstract, “Methods, systems, and apparatus, including medium-encoded computer program products, for computer aided design of physical structures using generative design processes. A method includes: obtaining a design space and design criteria for a modeled object including a design constraint on an acceptable likelihood of failure, …, wherein the numerical simulation includes computing the structural performance metric, which is evaluated against the design constraint; and providing the generatively designed shape of the modeled object for use in manufacturing a physical structure.”), the apparatus comprising:
a memory;
a processing device unit operatively coupled to the memory ([0050], “…A computer 110 includes a processor 112 and a memory 114 … which can include the memory 114, to store instructions of programs that run on the processor 112 …”), the processing device unit to:
store one or more design goals for a structural model, the one or more design goals including one or more parameters defining a form of the structural model and one or more performance conditions ([0008], “… obtaining a design space for a modeled object, for which a corresponding physical structure is to be manufactured using one or more materials, and design criteria for the modeled object including one or more loading cases for numerical simulation of the physical structure and at least one design constraint on an acceptable likelihood of failure for the physical structure …” [0055], “The user 160 (or other person or program) can specify a design space for a modeled object, for which a corresponding physical structure is to be manufactured, and design criteria for the modeled object. The design criteria can include one or more loading cases for numerical simulation of the physical structure … The design criteria can also include at least one design objective … and at least one design constraint …” Examiner note: the reference teaches specifying a design space for a modeled object, which defines allowable configurations of the structure and is represented using parameterized variables corresponds to parameters defining a form of the structural model, and further teaches design criteria including loading cases, design objective, and design constraints, which correspond to the design goals and performance conditions. These inputs are used in subsequent numerical simulation and optimization, and are retained by the system, indicating storage in memory);
performing a topology optimization loop of the one or more design goals ([0058], “As described herein, the CAD program(s) 116 implement at least one generative design process, which enables the CAD program(s) 116 to generate one or more portions of the 3D model(s) automatically (or the entirety of a 3D model) based on design objective(s) and constraint(s), where the geometric design is iteratively optimized based on simulation feedback.” [0059], “Various generative design processes can be used, which can optimize the shape and/or topology of at least a portion of the 3D model. The iterative optimization of the geometric design of the 3D model(s) by the CAD program(s) 116 can involve topology optimization …”), the topology optimization loop comprising:
providing an initial set of the one or more parameters ([0070], “The obtained 180 design criteria can be input by the user 160 and/or imported from another source. One or more of the design criteria can be defined over entire regions in the design space or over individual regions in the design space. Various design criteria can be obtained 180, including a setup for numerical simulation, e.g., densities of elements in an FEA model or a homogenized lattice material representation for a selected lattice topology to be used with a topology optimized 3D shape of the part being generatively designed, plus various design objectives and constraints …” [0056], “… the inputs for use in numerical simulation and generative design processes can include one or more regions of a current 3D model in which to generate new 3D geometry, loading case(s) defining one or more loads in one or more different directions to be borne by a physical structure being designed, one or more materials … one or more seed model types to use as input to a generative design process, one or more generative design processes to use, …” [0057], “In general, a set of requirements can be provided in terms of boundary conditions (e.g., structural loads and constraints), material(s), one or more starting shapes, manufacturing constraints and other parameters … Further, the design criteria for the modeled object include at least one design constraint on an acceptable likelihood of failure for the physical structure …” [0072], “Moreover, the statistical model that relates the structural performance metric to the specific likelihoods of failure is used by the CAD program(s) 116 to translate (before, during, and/or after an iterative loop 184, 186, 192) between the acceptable likelihood of failure and a value for the structural performance metric. Data used to create the statistical model can be obtained by physical testing of a specific material.” Examiner note: the reference teaches providing a set of design and simulation inputs, including parameters such as loading conditions, materials, boundary conditions, and manufacturing constraints, for use in numerical simulation and generative design processes. These parameters are provided prior to execution of the simulation and generative design process, and correspond to an initial set of one or more parameters, The reference further teaches performing numerical simulation of the structural model to compute a structural performance metric evaluated against a likelihood of failure constraint, thereby corresponding to a failure simulation of the structure model.);
computing, ([0072], “… the statistical model that relates the structural performance metric to the specific likelihoods of failure is used by the CAD program(s) 116 to translate (before, during, and/or after an iterative loop 184, 186, 192) between the acceptable likelihood of failure and a value for the structural performance metric.” [0121], “Numerical simulation of the physical response of the current model … is performed 184 using the one or more defined loads. The numerical simulation can include computing the structural performance metric, which is evaluated against the at least one design constraint for the acceptable likelihood of failure for the physical structure.” Examiner note: the reference teaches computing an output of a structural model through numerical simulation, including computing a structural performance metric evaluated against a design constraint corresponding to an acceptable likelihood of failure.);
([0101], “… More than one approach can be used to obtain the directional derivative of the objective function for use in gradient based optimization methods. Approaches … include direct differentiation, semi-analytical derivatives, adjoint method, and finite difference.” [0221], “the computing 346 involves using a gradient determined from a shape derivative evaluated for the maximum likelihood of failure design constraint at each of the different locations. The shape derivative can be an analytical expression of stress, … With the stress shape derivative in hand, the chain rule can be used to derive the derivative of the objective function with respect to reliability.” [0223], “… the maximum likelihood of failure value from the last loop iteration constitutes the predicted likelihood of failure for the physical structure…” Examiner note: the reference teaches obtaining one or more gradients for optimization, including directional and shape derivatives derived from an objective function, and further teaches computing a predicted failure value in the form of a likelihood of failure for the physical structure.); and
computing, by a topology optimization processing device, using the one or more gradients and the predicted value, an updated set of parameters representing an updated version of the structural model until a convergence criterion for changes of the updated set of parameters becoming within a threshold (fig.1, computer 110 including a processor 112 executing CAD programs 116 configured to perform generative design processes including topology optimization, which is corresponds to a topology optimization processing device. [0221] Returning to FIG. 3C, shape change velocities for an implicit surface in a level-set representation of the three dimensional shape are computed 346 based on the numerical assessment and in accordance with design criteria including the maximum likelihood of failure. In some implementations, the computing 346 involves using a gradient determined from a shape derivative evaluated for the maximum likelihood of failure design constraint … In essence, the gradient of the survivor function is computed for use in the iterative loop of the shape (and optionally topology) optimization process. [0223] But regardless of how the shape changes velocities are computed 346, the level-set representation is updated 348 using the shape change velocities to produce an updated version of the three dimensional shape of the modeled object, and the performing 344, the computing 346 and the updating 348 are repeated until a check 350 determines that a predefined number of shape modification iterations have been performed or that the generatively designed three dimensional shape of the modeled object in the design space has converged to a stable solution …the maximum likelihood of failure value from the last loop iteration constitutes the predicted likelihood of failure for the physical structure …” [0225], “…For each design constraint, a target much closer to the current value for the constraint is specified for each iteration.” [0210], “… a PID controller (e.g., an adaptive PID controller) can be used to ensure that the value is constrained to be below a configured target value.” Examiner note: the reference teaches computing gradients from shape derivatives and using a predicted likelihood of failure value, and iteratively updating the modeled object in a topology optimization loop based on the gradients and failure value until convergence to a stable solution. The iterative update of the model, including modification of the level-set representation and shape change velocities, corresponds to updating a set of parameters representing the updated version of the structural model.); and
producing, in a production compartment comprising a three-dimensional printer, a physical model of the structural model using the updated set of parameters ([0064], “… the CAD program(s) 116 can provide a document 135 (having toolpath specifications of an appropriate format) to the AM machine 170 to produce a complete structure 138 … The AM machine 170 can employ one or more additive manufacturing techniques, such as granular techniques (e.g., Powder Bed Fusion (PBF), Selective Laser Sintering (SLS) and Direct Metal Laser Sintering (DMLS)), extrusion techniques (e.g., Fused Deposition Modelling (FDM), which can include metals deposition AM).” [0200], “Once the generative design process is completed, the generatively designed three dimensional shape of the modeled object can be provided 196, e.g., by CAD program(s) 116, for use in manufacturing the physical structure. The 3D model can be provided 196 for use in manufacturing a physical structure corresponding to the object using one or more computer-controlled manufacturing systems, e.g., AM machine 170, SM machine 174, and/or other manufacturing machines/systems.” Examiner note: the reference teaches producing a physical structure by providing a generatively designed three-dimensional model to an additive manufacturing machine, which fabricates the structure using additive manufacturing techniques such as SLS or FDM, corresponding to a three-dimensional printer. The reference further teaches that 3D model used for manufacturing is the result of an iterative topology optimization process, and represents an updated version of the structural model defined by the iterative updates of the design parameters, corresponding to producing a physical model using the updated set of parameters.).
However, Harris fails to teach, but Iyer teaches a surrogate model emulating, using a trained machine-learning model, a simulation of the structural model based on one or more training datasets, wherein the simulation of the structural model simulates an additive manufacturing process of the structural model, and wherein the one or more training datasets are generated for different boundary conditions, (Abstract, “a Deep Convolutional Neural Network (DCNN) model is explored as a surrogate for the physics-based model, so that it can be used to time-efficiently estimate the crack index for a given part-design. This requires careful design of the training regime and dataset for a given design problem.” Page.182, 3.2 Training Data Generation, “A parametric data generator was designed to invoke suitably diverse variants of the design problem, thereby enriching the training data for developing the surrogate model … The generator runs a set of topology optimization (TO) simulations across a broad range of boundary conditions, loading conditions, design constraints and combinations of those to create a series of topology optimized design variants. These 3D design variants … are the inputs for training the surrogate. Given that these design variants are targeted for additive manufacturing, additional constraints like presence of overhangs, … are also accounted for … to ensure that the candidates are feasible and can be reliably printed by an AM machine.” Page.183, lines 4-5, “Each such design variant is evaluated using the physics-based additive simulation model to estimate the crack index.” Examiner note: the reference teaches generating training data using physics-based simulations of structural designs under additive manufacturing conditions, and evaluating each design variant using the simulations to estimate a crack index. Therefore, the surrogate model is trained to emulate simulation results of a structural model corresponding to an additive manufacturing process);
computing, using the surrogate model, an output that includes a crack index for the structural model (Abstract, “a Deep Convolutional Neural Network (DCNN) model is explored as a surrogate for the physics-based model, so that it can be used to time-efficiently estimate the crack index for a given part-design.”);
the output of the surrogate model and the crack index (Abstract, “a Deep Convolutional Neural Network (DCNN) model is explored as a surrogate for the physics-based model, so that it can be used to time-efficiently estimate the crack index for a given part-design.” Examiner note: the reference teaches estimating a crack index for a given part design using a surrogate mode. The crack index represents a quantitative measure of cracking behavior within the structure. Therefore, the crack index reflects structural performance characteristics of the modeled design and is included in the output of the surrogate model.).
It would have been obvious for a person of ordinary skill in the art before the effective filing date of the claimed invention to have modified Harris to incorporate the teachings of Iyer, and apply a surrogate model configured to output a crack index representing structural behavior of a design, the crack index being a quantitative output indicative of cracking behavior of the structural design, in order to improve computational efficiency and reduce the cost associated with repeated physics-based simulations during topology optimization, while enabling evaluation of structural performance including failure characteristics as performed in Harris. In this case, Harris teaches a topology optimization framework that performs numerical simulations to evaluate structural performance, including likelihood of failure, and iteratively updates design parameters based on the evaluations. Iyer teaches a surrogate model, such as trained machine learning model, that emulates physic-based simulations and outputs a crack index indicative of cracking behavior of a structural design under additive manufacturing conditions. The combination of teachings would predictably provide benefit of enabling faster evaluation of structural performance metrics, including failure characteristics, by replacing or supplementing computationally expensive simulations with surrogate predictions, thereby improving efficiency of the optimization process while maintaining high performance evaluation.
However, Harris and Iyer fail to teach, but Chen teaches apply a computational layer to the output of the model to obtain one or more gradients by automatic differentiation and a predicted value (Abstract, “… the neural network generates feasible topology designs, and then the topology performance is evaluated using the finite element method … the physics-informed neural network weights are updated directly using gradient information from the physics model, i.e., finite element analysis. The key idea is that these gradients are calculated automatically through the finite element solver and then backpropagated to the deep learning neural network during the training or intelligence building process. This integrated optimization approach is implemented in Julia programming language and can be automatically differentiated in reverse mode for gradient calculations.” Examiner note: the reference teaches generating an output using a neural network model and subsequently performing computational evaluation of the output to obtain a performance value representing structural behavior. The evaluation constitutes applying a computational operation to the model output to produce a predicted value. The reference further teaches computing gradients of the model through automatic differentiation, including reverse model differentiation of the computational process. Therefore, the reference teaches applying a computational layer to a model output to obtain both gradient and a predicted performance related value.).
It would have been obvious for a person of ordinary skill in the art before the effective filing date of the claimed invention to have modified Harris and Iyer to incorporate the teachings of Chen, and apply automatic differentiation to model outputs in order to efficiently compute gradients for use in gradient optimization and obtain corresponding performance related values. In this case, Harris teaches a topology optimization framework that performs numerical simulations to evaluate structural performance, including likelihood of failure, and iteratively updates design parameters based on the evaluations. Iyer teaches a surrogate model, such as trained machine learning model, that emulates physics based simulations and outputs a crack index indicative of cracking behavior of a structural design under additive manufacturing conditions, where the crack index corresponds to a predicted measure of structural behavior related to failure. Chen teaches applying automatic differentiation to outputs of a computational model to obtain gradients and corresponding performance related values. The combination of teachings would predictably provide benefit of enabling efficient gradient based optimization using surrogate model outputs, thereby reducing computation cost and maintaining accurate evaluation of structural performance and failure characteristics.
However, Harris and Iyer and Chen fail to teach, but Banga teaches the one or more training datasets are generated for different volume fractions (Figure 2: Parameters sampled for data generation. Page. 5, 4.1 Data Generation, “We generated synthetic data using the open source topology optimization tool ‘TopOpt” … To generate spatially variant data, we devised a sampling strategy to define the loading and boundary conditions for the topology optimization problem. Parameters which are sampled are depicted in Figure 2. … Volume fraction (V0) V0: Sampled from Normal Distribution N (μ= 0.28; σ= 0.07) The above parameters are chosen to ensure that volume fraction values primarily vary between 0.07 to 0.5.” Page.7, 4.3 Training the Network, “Out of the 6000 data samples generated using the sampling strategy, we used 4500 data samples for training the network, …” Examiner note: the reference teaches generating synthetic data using a topology optimization process, wherein parameters are sampled for data generation. The sampled parameters include volume fraction (V0), which is drawn from a distribution such that volume fraction values vary across arrange. Therefore, the generated data related to multiple instance produced under different volume fraction conditions, which correspond to training datasets generated for different volume fractions.).
It would have been obvious for a person of ordinary skill in the art before the effective filing date of the claimed invention to have modified Harris and Iyer and Chen to incorporate the teachings of Banga, and apply generating training datasets by sampling design parameters, including volume fraction, for data generation in order to improve robustness and generalization of surrogate modes used in structural design and optimization under varying material distribution conditions. In this case, Harris teaches a topology optimization framework that performs numerical simulations to evaluate structural performance, including likelihood of failure, and iteratively updates design parameters based on the evaluations. Iyer teaches a surrogate model, such as trained machine learning model, that emulates physics based simulations and outputs a crack index indicative of cracking behavior of a structural design under additive manufacturing conditions, where the crack index corresponds to a predicted measure of structural behavior related to failure. Chen teaches applying automatic differentiation to outputs of a computational model to obtain gradients and corresponding performance related values. Banga teaches generating synthetic data using topology optimization, wherein parameters including volume fractions, are sampled for data generation such that the generated dataset includes structural designs corresponding to different volume fraction conditions. The combination of teachings would predictably provide benefit of enabling training of surrogate models using datasets that capture variations in material distribution, thereby improving accuracy and reliability of performance prediction, including failure characteristics and maintaining computational efficiency in topology optimization processes.
Claim 2, Harris teaches The apparatus of claim 1, wherein the topology optimization processing device computes the initial set of the one or more parameters in a previous computation cycle of the topology optimization loop (see [0210], [0221], [0223] and [0225]), and wherein the processing device unit further to:
provide the updated set of parameters to the
identify the set of design parameters based on one or more cycles of computation upon satisfying the convergence criterion (See [0223], “the computing 346 and the updating 348 are repeated until a check 350 determines that a predefined number of shape modification iterations have been performed or that the generatively designed three dimensional shape of the modeled object in the design space has converged to a stable solution for the design criteria … Moreover, once the iterative loop ends, the maximum likelihood of failure value from the last loop iteration constitutes the predicted likelihood of failure for the physical structure …” Examiner note: the reference teaches an iterative optimization loop where updated parameters are repeatedly generated and used in subsequent iterations.).
However, Harris fails to teach surrogate model.
Iyer teaches surrogate model (Page.180, “A cracked part results in the need to refine the part-design and to repeat the print, leading to multiple iterations from design to full scale manufacture of the part … we explore the hypothesis of whether time-efficient and accurate surrogates can be designed, leveraging these expensive physics-based models, so that they can provide a reliable estimate of the crack index for a given candidate of the design problem.” Page. 181, This work leverages Deep Convolutional Neural Networks (DCNN) to construct high fidelity and time-efficient surrogates for the high fidelity physics-based models of residual stress.”).
It would have been obvious for a person of ordinary skill in the art before the effective filing date of the claimed invention to have modified Harris to incorporate the teachings of Iyer, and apply a surrogate model in order to replace computationally expensive physics based simulations with a time efficient predictive model for estimating structural performance (e.g., crack index) during iterative topology optimization.
Claim 7, Harris further teaches The apparatus of claim 1, wherein the initial set of one or more parameters comprises:
attributes defining geometric and material properties of the structural model ([0056] …the inputs for use in numerical simulation and generative design processes can include one or more regions of a current 3D model in which to generate new 3D geometry, loading case(s) defining one or more loads in one or more different directions to be borne by a physical structure being designed, one or more materials (e.g., one or more isotropic solid materials identified as a baseline material model for the design space)…); and
the performance conditions including at least one of a boundary condition, a loading condition, or a thermal condition of the structural model ([0057] … a set of requirements can be provided in terms of boundary conditions (e.g., structural loads and constraints), material(s), one or more starting shapes, manufacturing constraints and other parameters, and the CAD program(s) 116 create various shapes that satisfy the requirements using one or more generative design processes …).
Claim 8, Harris fails to teach, but Iyer teaches The apparatus of claim 1, wherein the one or more training datasets correspond to manufacturing conditions of at least one of: thermal conditions, stress conditions, or asymmetric behaviors thereof (Page.182, 3.2 Training Data Generation, “The generator runs a set of topology optimization (TO) simulations across a broad range of boundary conditions, loading conditions, design constraints and combinations of those to create a series of topology optimized design variants. These 3D design variants, voxelised at an appropriate resolution, are the inputs for training the surrogate … One way to explicitly introduce variation in the samples was to vary the direction and magnitude of the external loads applied to the coupon … Each of these four segments is subjected to an independently varying traction force. Each such design variant is evaluated using the physics-based additive simulation model to estimate the crack index.”).
It would have been obvious for a person of ordinary skill in the art before the effective filing date of the claimed invention to have modified Harris to incorporate the teachings of Iyer, and apply generating training data using simulations under varying boundary and loading conditions to produce diverse design samples for training a surrogate model in order to improve prediction accuracy and robustness of the surrogate model across different manufacturing scenarios.
The elements of claims 9-12 and 16-20 are substantially the same as those of claims 1-2 and 7-8. Therefore, the elements of claims 9-12 and 16-20 are rejected due to the same reasons as outlined above for claims 1-2 and 7-8. Further, the additional limitation of claim 16, “A non-transitory computer-readable storage medium having instructions stored thereon that, when executed by a processing device for computing a set of design parameters defining a structural model to be manufactured, cause the processing device to:” (see Harris, abstract, [0050] and [0014]).
Claim(s) 4 and 13 are rejected under 35 U.S.C. 103 as being unpatentable over Harris and Iyer and
Chen and Banga as applied to claims 1 and 9 above, and further in view of Cao (“A unified model of ductile fracture considering strain rate and temperature under the complex stress states,” published in 2021) and Lou (“New ductile fracture criterion for prediction of fracture forming limit diagrams
of sheet metals,” published in 2012).
Claim 4, Harris fail to teach, but Iyer teaches The apparatus of claim 1, wherein the failure criterion comprises (Abstract, “… a Deep Convolutional Neural Network (DCNN) model is explored as a surrogate for the physics-based model, so that it can be used to time-efficiently estimate
the crack index for a given part-design..”. Page.181, “This work leverages Deep Convolutional Neural Networks (DCNN) to construct high fidelity and time-efficient surrogates for the high fidelity physics-based models of residual stress … while it employs high-fidelity physics-based simulation, and a deep learning based model as a surrogate, to estimate stress for varying geometries …” Fig.1, “…red values indicating high propensity of cracking upon printing …” Page.183, “Each such design variant is evaluated using the physics-based additive simulation model to estimate the crack index.”).
It would have been obvious for a person of ordinary skill in the art before the effective filing date of the claimed invention to have modified Harris to incorporate the teachings of Iyer, and apply a surrogate model configured to output a crack index (stress) representing structural behavior of a design, the crack index being a quantitative output indicative of cracking behavior of the structural design, in order to improve computational efficiency and reduce the cost associated with repeated physics-based simulations during topology optimization.
However, Harris and Iyer and Chen and Banga fail to teach a maximum shear stress index (MSSI) exceeding a threshold value, and wherein the MSSI is a function of thermal related variables.
Cao teaches (page.10, “the damage value is defined as the equivalent strain divided by the critical fracture strain, and is shown as Fig. 13(b). Fracture will occur when the damage value is greater than 1.0.” Page.9, “Therefore in order to improve the prediction ability of the model, the temperature term in the original J-C fracture model is modified to be (D2 + D3exp(D4T∗), where D3 and D4 are temperature softening parameters.” Examiner note: the reference teaches that a damage index is used as a fracture criterion, where fracture occurs when the damage value exceeds a threshold, and further teaches that the fracture criterion incorporates temperature terms, including temperature softening parameters in the model, thereby indicating that the index is a function of thermal related variables such as temperature).
It would have been obvious for a person of ordinary skill in the art before the effective filing date of the claimed invention to have modified Harris and Iyer and Chen and Banga to incorporate the teachings of Cao, and apply a fracture criterion that evaluates failure based on an index exceeding a threshold value and incorporating thermal variables in order to accurately determine when structural failure occurs under varying thermal conditions. The combination of teachings would provide the benefit of improving the accuracy and reliability of predicting cracking behavior during manufacturing b incorporating a thermally depend failure criterion into the optimization framework.
However, Harris and Iyer and Chen and Banga and Cao fail to teach the index is based on a maximum shear stress index.
Lou teaches a maximum shear stress index (Abstract, “… void coalescence is controlled by the normalized maximal shear stress ...”; Page.3607, left column 2.1.3, “shear fracture is caused by the maximal shear stress, the void coalescence is modeled by the maximal shear stress normalized by the equivalent stress …” 2.2, “the coalescence of voids is described by the normalized maximal shear stress denoted as τmax/σ.” Examiner note: the reference teaches that void coalescence is describe by a normalized maximal shar stress and that void coalescence is controlled by the normalized maximal shar stress, and further teaches that shear fracture is caused by the maximal share stress. Therefore, the normalized maximal shear stress used to describe and control fracture behavior and corresponds to a maximum shear stress index, as it re[resents a quantity derived from maximum shar stress used to evaluate fracture).
It would have been obvious for a person of ordinary skill in the art before the effective filing date of the claimed invention to have modified Harris and Iyer and Chen and Banga and Cao to incorporate the teachings of Lou, and apply using a normalized maximum shear stress to define a fracture metric in order to characterize crack formation based on shear stress conditions of a material. The combination of teachings would provide the benefit of defining the crack index based on a shear stress metric that accounts for thermal fracture behavior, thereby improving the accuracy and physical relevance of crack formation prediction during manufacturing.
The elements of claim 13 is substantially the same as those of claim 4. Therefore, the elements of claim 13 is rejected due to the same reasons as outlined above for claim 4.
Allowable Subject Matter
Claims 5-6 and 14-15 are objected to as being dependent upon rejected base claims, but would be
allowable if rewritten in independent form including all of the limitations of the base claim and any intervening claims.
The following is a statement of reasons for the indication of allowable subject matter:
Claim(s) 5 and 14 are considered allowable since when reading the claims in light of the specification, none of the references of record alone or in combination disclose or suggest the combination of limitations specified in the claims, specifically, “minimizing, for a design variable such that a state equation is solved using finite element analysis when a volume constraint is satisfied; the state equation is based on at least: a performance function of the structural model under design load; a cracking index of the structural model considering thermal and residual stress during manufacturing; a weighting factor, and a target volume fraction,” as presented in claims 5 and 14 of the instant application. Dependent claims 6 and 15 are allowed as being dependent from allowed claims 5 and 14.
Prior Art of Record
The prior art references:
(Cited in the previous Office action) Chandrasekhar (“AuTO: a framework for Automatic differentiation in Topology Optimization,” published in 2021), disclose minimizing, for a design variable such that a state equation is solved using finite element analysis when a volume constraint is satisfied; the state equation is based on at least: a performance function of the structural model under design load; a weighting factor; and a target volume fraction (page.4356, algorithm 1 directly implements an optimization loop. The design parameters ρ are initialized (line 3); updated iteratively (lines 14-15); Gradients of the objective and constraint functions are computed using automatic differentiation (lines 11 and 13). The state equation K(ρ)u = f is solved using FEA (line 9); A global volume constraint with target volume fraction V* is enforced (lines 3 and 13); load vector f (line 9 and section 2.1). The framework allows objective and constraint function to include weighting factors (section 2.1); Section 2.1, “where u is the displacement (in structural problems) or temperature (in thermal problems), K is the stiffness matrix, are the pseudodensity design variables, f is the structural/thermal load and V∗ is the volume constraint… for the solid isotropic material with penalization (SIMP) (note: i.e., weighting factor) (Bendsøe and Sigmund 1995) material model, can be easily derived…”). However, the reference fails to teach cracking index of the structural model considering thermal and residual stress during manufacturing included in the state equation.
(Cited in the current Office Action) Chen (“A New Topology Optimization Approach by Physics-Informed Deep Learning Process,” published in 2021), discloses minimizing, for a design variable such that a state equation is solved using finite element analysis when a volume constraint is satisfied; the state equation is based on a performance function (Abstract, “In every iteration, the neural network generates feasible topology designs, and then the topology performance is evaluated using the finite element method.” Page.234, right column, “The objective of the SIMP method is to minimize the compliance, 𝐶, of the design domain under fixed loadings and boundary conditions. The compliance defined in (1), also described as total strain energy, is a measure of the overall displacement of a structure … Inside the optimization loop, the density of each element 𝑥𝑒 has to be updated to lower the compliance. ” 1.3. Motivation, “In this work, the goal is to integrate a physics-based model (e.g., finite element model) with a neural network for generating feasible topology designs so that the gradient information obtained from the finite element model can be used to minimize the loss function during the training process to achieve optimal topology designs.” Page.235, left column, “The 𝑈load is the displacement at the applied load point and it is the loss function, L, to be minimized … The 𝑀(𝑥) returns the mass fraction of the topology produced by generated and compare it to the target mass fraction. The 𝐾(𝑥) and F are the stiffness matrix of the topology and load vector, respectively. They are processed in the FEM component in Figure 2, and a resultant displacement U is returned.” Equation (4). Page.235, right column, “The parameters of the network are adjusted simultaneously to generate better design while satisfying the mass constraint.”). However, the reference fails to teach a cracking index of the structural model considering thermal and residual stress during manufacturing; a weighting factor; and a target volume fraction included in the state equation.
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure.
G. Johnson et al., "Fracture Characteristics of Three Metals Subjected to Various Strains, Strain Rate, Temperatures and Pressures," discloses the point of maximum stress is important inasmuch as it represents the strain at which localized instabilities may begin to occur … Fracture is then allowed to occur when D "" 1.0 … the third set of brackets represents the effect of temperature (Page.39-40).
M. Raissi et al., "Physics-informed neural networks: A deep learning framework for solving forward and inverse problems involving nonlinear partial differential equations," discloses physics-informed neural networks–neural networks that are trained to solve supervised learning tasks while respecting any given laws of physics described by general nonlinear partial differential equations.
Any inquiry concerning this communication or earlier communications from the examiner should be
directed to YI HAO whose telephone number is (571)270-1303. The examiner can normally be reached Monday - Friday.
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, Emerson Puente can be reached at (571)272-3652. 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.
/YI . HAO/
Examiner, Art Unit 2187
/EMERSON C PUENTE/Supervisory Patent Examiner, Art Unit 2187
Prosecution Insights