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 .
Response to Amendment
The amendment filed 07/10/2026 has been entered. As directed, claims 1, 6, 9, 15-16, and 18
have been amended, claims 3, 5 and 14 have been canceled, no claim has been added. Thus claims 1-2, 4, 6-13 and 15-20 remain pending in the application.
Response to Arguments
Applicant' s arguments, see “Applicant Arguments/Remarks Made in an Amendment”, filed
7/10/2026, with respect to the rejection(s) of claim(s) 1, 9 and 16 under 35 U.S.C. § 103 have been fully considered. The prior rejection has been withdrawn. However, upon further consideration, a new ground(s) of rejection are made based on newly cited prior art as set forth below. Accordingly, Applicant’s assertion that claims 1, 9, and 16 and their respective dependent claims are in condition for allowance is not persuasive. Further, an alternative rejection under 35 U.S.C. § 103 is made relying on N. Iyer et al., “PATO: Producibility-Aware Topology Optimization using Deep Learning for Metal Additive Manufacturing,” published in 2021. Applicant may overcome this rejection under 35 U.S.C. 102(a)(1) by a showing under 37 CFR 1.130(a) that the subject matter disclosed in the reference was obtained directly or indirectly from the inventor or a joint inventor of this application, and is therefore, not prior art as set forth in 35 U.S.C. 102(b)(1)(A). Alternatively, applicant may rely on the exception under 35 U.S.C. 102(b)(1)(B) by providing evidence of a prior public disclosure via an affidavit or declaration under 37 CFR 1.130(b).
Claim Rejections - 35 USC § 112
The following is a quotation of 35 U.S.C. 112(b):
(b) CONCLUSION.—The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the inventor or a joint inventor regards as the invention.
The following is a quotation of 35 U.S.C. 112 (pre-AIA ), second paragraph:
The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the applicant regards as his invention.
Claims 1-2, 4, 6-13, and 15-20 are rejected under 35 U.S.C. 112(b) or 35 U.S.C. 112 (pre-AIA ), second paragraph, as being indefinite for failing to particularly point out and distinctly claim the subject matter which the inventor or a joint inventor (or for applications subject to pre-AIA 35 U.S.C. 112, the applicant), regards as the invention.
Claim 1 recites “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; the crack index of the structural model considering thermal and residual stress during manufacturing; a weighting factor; and a target volume fraction,” which renders the claim indefinite because it is unclear what value or quantity is being minimized, and it is further unclear what relationship is required between the state equation and the recited performance function, crack index, weighting factor, and target volume fraction. For the purpose of substantive examination, the examiner presumes that the limitation requires minimizing, for a design variable, an objective value based on a performance function of the structural model under design load and the crack index of the structural model considering thermal and residual stress during manufacturing, as weighted by a weighting factor, subject to a state equation solved using finite element analysis and a volume constraint associate with a target volume fraction.
Claims 9 and 16 recite limitations similar to those of claim 1, and are rejected for the same reasons discussed above.
The remaining claims are dependent upon one of the claims listed above and are rejected for the same reasons.
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) and Giraldo-londono WO2020159812A1.
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 the 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);
perform 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.” [0198], “To determine whether the generative design has converged to a stable solution, the check 192 can identify the condition in which all the design constraints are met and no design objectives have improved significantly since the last one or more iterations.” 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.),
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.180, “Such rapid heating and cooling can induce high thermal stresses in the part resulting in the generation of significant thermal residual stresses. These residual stresses manifest in the form of part deformation or if significantly high, it can result in the part cracking at multiple locations [3]. 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 … These stresses can then be used to synthesize a 3D crack index field for the entire part, capturing the likelihood of failure around specific locations in the part.” 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, “High fidelity, physics-based numerical models of the additive melting process exist that can simulate the thermal gradients and consequent stresses produced during manufacturing, which can then be used to synthesize a 3D crack index field for the entire part volume, capturing the likelihood that a region in a part will crack upon heat treatment … 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.180, Introduction, “Several commercially available high fidelity, physics-based software tools can simulate the additive melting process, the resulting thermal gradients and consequent residual stresses during manufacturing. These stresses can then be used to synthesize a 3D crack index field for the entire part, capturing the likelihood of failure around specific locations in the part.”);
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.);
considering thermal and residual stress during manufacturing; (Page.180, “Such rapid heating and cooling can induce high thermal stresses in the part resulting in the generation of significant thermal residual stresses. These residual stresses manifest in the form of part deformation or if significantly high, it can result in the part cracking at multiple locations [3]. 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 … These stresses can then be used to synthesize a 3D crack index field for the entire part, capturing the likelihood of failure around specific locations in the part.”).
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 applying 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 arrangements. 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.
However, Harris and Iyer and Chen and Banga fail to teach, but Giraldo-londono teaches the computing the updated set of parameters comprises: 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 (See rejection under 112(b); [0045], “The topology optimization framework 200 includes a model setup step which involves creating a CAD model 202 geometry, discretizing the geometry into a mesh (meshing 205), defining boundary conditions 207 and material properties 210, a design domain 212 and creating objective functions. Then, a design variable field of the design domain 212 will be initialized … a primal solution 215 will be computed with a physics solver 217 to solve for physical field variables (e.g. displacement) … The resulting sensitivity 220 will be used as input to an optimizer 225 along with design variable to compute a new set of design variable 230. This process will be iterated until the optimization convergence criteria is being met, resulting in an optimal design.” [0046], “… the objective function is written as a weighted sum of a mechanical objective function (mechanical compliance) and a thermal objective function (thermal compliance or temperature variance).” [0051], “… the objective function, J … is defined in terms of a mechanical objective, Jm, and a thermal objective, Jo, where 0 < w < 1 is a weight factor. The design variables, zi, . . . , zm, correspond to vectors of element densities … gj, j = 1, . . . , Nc is the jth volume constraint … vj is the upper limit for volume constraint j; Ku = f is the linear elastic equilibrium equation, where K, u, and f are the stiffness matrix, displacement vector, and load vector, respectively.” [0034], “A volume constraint is a maximum fraction of volume that should be filled with material in the outcome of the optimization process.” [0035] “… the mechanical performance 132(1) is measured using a structural or mean compliance 145 of a system.” See also [0050], “min J (Z1 ..., Zm) = wJm + (1-w)Jo” [0070], “The bridge is discretized with 75,000 polygonal finite elements.” Examiner note: the reference teaches a density-based topology optimization in which element density design variables are iteratively updated to minimize a weighted multi-objective function. The mechanical performance objective is structural compliance, and the objective includes a weight factor w, and further teaches the linear elastic equilibrium equation Ku = f, where in K, u an f respectively correspond to the stiffness matrix, state/displacement vector, and load vector, and solves the structural model using a discretized finite element representation. The reference also teaches imposing volume constraints defining a maximum allowable fraction of material in the optimized design.).
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 Giraldo-londono, and apply a density based multi-objective topology optimization formulation in which a design variable is optimized based on a weighted consideration of structural performance and a manufacturing related objective, subject to a finite-element state equation and a volume constraint ,and incorporate the crack index taught by Iyer as the manufacturing related objective, in order to account for susceptibility to cracking caused by thermal gradients and residual stress during additive manufacturing while optimizing structural performance. 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. Giraldo-londono teaches a density based multi-objective topology optimization formulation that minimizes a weighted objective based on structural performance and another objective, subject to a finite-element equilibrium equation and volume constraints. The combination of teachings would predictably provide benefit of enabling the structural design to be optimized while accounting for both structural performance and susceptibility to cracking caused by thermal and residual stresses during additive manufacturing.
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 ([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 ([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.182-183, 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, 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 … A set of 116 distinct design samples were generated and evaluated as described above to generate the training data for the crack index surrogate.” Page.180, Introduction, “Several commercially available high fidelity, physics-based software tools can simulate the additive melting process, the resulting thermal gradients and consequent residual stresses during manufacturing. These stresses can then be used to synthesize a 3D crack index field for the entire part, capturing the likelihood of failure around specific locations in the part. Examiner note: the reference teaches that the training data comprise design variants generated under different simulation conditions, wherein each design variant is evaluated using an additive manufacturing physics based simulation to estimate the crack index. The additive manufacturing simulation includes thermal gradients and consequent residual stressed during manufacturing. Therefore, the training datasets correspond to at least thermal and stress manufacturing conditions).
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 from topology optimization simulations performed under varying boundary and loading conditions for use in training the surrogate model, in order to provide sufficiently diverse training samples representative of different manufacturing related operating conditions and thereby improve the accuracy and robustness of crack prediction by the surrogate model.
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 and Giraldo-londono as applied to claims 1 and 9 above, and further in view of Han US20220075911A1 and Sakai US20040050318A1.
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.180, “Several commercially available high fidelity, physics-based software tools can simulate the additive melting process, the resulting thermal gradients and consequent residual stresses during manufacturing. These stresses can then be used to synthesize a 3D crack index field for the entire part, capturing the likelihood of failure around specific locations in the part.” 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 …” Page.182 “Fig. 1. Example coupons showing the design problem tackled in this paper. The second, blue volume in each subfigure shows the crack index distribution with red values indicating high propensity of cracking upon printing. The first design-instance clearly shows a high likelihood of cracking at the bottom wall of the hole (red values), while the second design-instance looks to have low crack index values. (Color figure online)” 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 representing a likelihood or propensity of cracking of the structural design during additive manufacturing in order to efficiently evaluate cracking behavior of candidate structural designs during topology optimization and reduce the computational cost associated the repeatedly performing high fidelity physics based additive manufacturing simulations.
However, Harris and Iyer and Chen and Banga and Giraldo-londono fail to teach a maximum shear stress index (MSSI) exceeding a threshold value, and wherein the MSSI is a function of thermal related variables.
Han teaches a maximum shear stress index (MSSI) exceeding a threshold value for indicating cracking of the structural model during manufacturing, ([0022] The strength criterion of the structural materials is summarized as: θ(σ(x))≥θcrit and x∈Ω, where Ω denotes the set of all points in the structure, θ(σ(x)) denotes the strength function of the stress tensor σ(x) at point x of the stress field, θcrit is the strength of the structural material.” [0048], “… define it as θ(σ(x))=σ1−σ3 according to the theory of maximum shear stress …” [0050], “when the complex stress reaches a specific limit state, the structure will begin to fail. The strength criterion is the particular characterization of these ultimate stress states.” [0051], “… strength theory … can be used to predict the initiation of cracks on the structure … It uses the material strength as a clear determination indicator to introduce a peridynamic model to predict the cracks' initiation and propagation process on the structure.” Examiner note: the reference defines a strength criterion θ(σ(x))≥θcrit, wherein θ(σ(x)) is a strength function of the stress tensor and θcrit is the strength of the structural material. The reference further defines the same strength function as θ(σ(x))=σ1−σ3 according to maximum shear stress theory. Thus, the maximum shear stress based strength function is quantitatively evaluated against a threshold corresponding to material strength. The reference further teaches that reaching this limit state indicates structural failure and that the strength criterion is used to predict crack initiation. Accordingly, the maximum shear stress based strength function corresponds to the claimed maximum shear stress index used for indicating cracking).
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 Giraldo-londono to incorporate the teachings of Han, and apply a maximum shear stress based strength criterion that compares the maximum shear stress based strength function with a threshold corresponding to material strength, in order to provide a clear indicator for determining structural failure and predicting crack initiation.
However, Harris and Iyer and Chen and Banga and Giraldo-londono and Han fail to teach the index is based on a maximum shear stress index.
Sakai teaches the MSSI is a function of thermal related variables ([0049], “In the embodiment according to the present invention, the entire cylindrical surface of CaF 2 crystal was isothermal, and the maximum shear stress among thermal stress that is determined by the size of CaF2 crystal and cooling rate is approximated as follows by considering parameters including thermal diffusivity, specific heat, Young's modulus, and Poisson's ratio of CaF2 single crystal: τMAX=9E−6·exp(2E−3·T)·D 2 ·R (3)).” [0050]
where τ MAX is the maximum shear stress [MPa], D is a thickness of crystal, R is a cooling rate, and T is temperature [K] of CaF2.” Examiner note: the reference defines the maximum shear stress τMAX as a function including temperature T, wherein T is the temperature of the material. Thus. The maximum shear stress equality varies as a function of a thermal variable.
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 Giraldo-londono and Han to incorporate the teachings of Sakai, and apply determining the maximum shear stress as a function of temperature in the failure criterion, in order to account for changes in maximum shear stress caused by thermal conditions when evaluating structural failure, thereby improving the accuracy of predicting thermally induced structural failure under varying temperature conditions.
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.
Claim(s) 6 and 15 are rejected under 35 U.S.C. 103 as being unpatentable over Harris and
Iyer and Chen and Banga and Giraldo-londono as applied to claims 1 and 9 above, and further in view of “Optimization of support structures for selective laser melting” by Zeng, published on Aug. 2015.
However, Claim 6, Harris and Iyer and Chen and Banga and Giraldo-londono fail to teach, but Zeng teaches The apparatus of claim 1, wherein the failure simulation of the structural model simulates an additive manufacturing process in the additive production compartment (Problem statement, page.4, “… A thermo-mechanical simulation tool for SLM developed at the University of Louisville is being commercialized by 3DSIM, LLC. It is able to simulate the temperature and thermal stress distribution in the SLM process in a fine-scale … and makes possible simulating the complete problem for an SLM process.” Page.5, “the 3DSIM model able to simulate the SLM process using the same scan pattern as used in an SLM machine to study the influence of various scan patterns on the temperature, stress distribution and microstructure of parts made using SLM.”Page.99, 6.1 Introduction, “Selective laser melting (SLM) is an Additive Manufacturing (AM) technology whereby a part is built by melting successive layers of metal powder. The part is typically built upon a plate in an enclosed chamber.”) to:
form the structural model layer by layer (Page.19, “The computational domain consists of five layers of 1x1x0.03mm Titanium powder deposited on the mild steel substrate … The model was also built in ANSYS and the functionality of element death and birth is applied for new layer deposition.” Page.119, “This model will be solved using a thermomechanical eigenmodal strategy layer-by-layer with a contact model to account for snapping of supports at their interface with the baseplate and/or the part. At the end of each simulation, the dimensions of the solidified struts will be periodically updated as a function of layer-by-layer buildup of the fabricated structure. In the meanwhile, it results in updated element properties from powder to solid with an updated total stiffness matrix.”);
for a position in each layer, heat a material from a first solid state to a fluid state (Page. 10, 2.3.1 SLM Thermal Phenomenon, “A high energy density laser beam is applied in SLM to melt the powder material.” Page.11, “During the heating and cooling process, the material undergoes phase changes from solid to liquid and back to solid …”); and
allow the material in the fluid state to dissipate heat and return to a second solid state (Page. 10-11, 2.3.1 SLM Thermal Phenomenon, “A high energy density laser beam is applied in SLM to melt the powder material. The melt pool solidifies quickly to form the bulk part … The laser energy is dissipated in several ways as shown in Figure 6 … Most of the remaining energy is conducted by the powder and air trapped among powder particles through the powder bed and the base plate. During the heating and cooling process, the material undergoes phase changes from solid to liquid and back to solid …”); wherein the updated version of the structural model is prevented from cracking due to thermal conditions related to phase changes (Page. Vii, “Supports are optimized and designed based on the thermal stress accumulated in parts as they are made … The support structure is designed to be withstand the thermal stress at locations where it could cause damage to the part and support structure …” Page.119, “Thermally induced stress is calculated using 3DSIM tools and optimized support parameters are calculated …The final support structure parameters are the maximal value of all scenarios and it is updated iteratively.” Page. 41, “Two optimization runs of support structure stiffening led to a reduction of thermal stress of about 72% … The experimental results clearly show that the specimens with standard block supports have a higher stress level than the optimized support designed specimens.”).
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 Giraldo-londono to incorporate the teachings of Zeng, and apply a simulated layer by layer additive manufacturing process including melting and solidification phase changes, while updating the structural design based on thermally induced stresses so as to reduce or avoid cracking or damaging during fabrication, in order to account for the thermal stresses generated during the additive manufacturing process when evaluating and optimizing the structural design, thereby producing an updated structural design having reduced thermal stress and improved resistance to cracking or damaging during additive manufacturing.
The elements of claim 15 is substantially the same as those of claim 6. Therefore, the elements of claim 15 is rejected due to the same reasons as outlined above for claim 6.
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
Iyer (“PATO: Producibility-Aware Topology Optimization using Deep Learning for Metal Additive Manufacturing,” published in 2021) in view of Harris US20230088537A1.
Claim 1, Iyer (PATO) teaches An apparatus for manufacturing a structural model and computing a set of design parameters defining the structural model (Pages. 1, Abstract, In this paper, we propose PATO - a producibility-aware topology optimization (TO) framework to help efficiently explore the design space of components fabricated using metal additive manufacturing (AM), while ensuring manufacturability with respect to cracking … Further, we employ automatic differentiation to directly compute the gradient of maximum MSSI with respect to the input design variables and augment it with the performance-based sensitivity field to optimize the design while considering the trade-off between weight, manufacturability, and functionality.” See also 5. Results), the apparatus comprising:
a memory;
a processing device unit operatively coupled to the memory, the processing device unit to:
store one or more design goals for the 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 (Page.13, 5. Result, “desktop machine”. Page.2, “Topology optimization (TO) is a powerful automated design paradigm that identifies optimized realizations of a conceptual design problem defined in terms of loading conditions, and performance and manufacturing objectives and constraints.” Page.4, top right, “In our experiment, the primary parameter is the shape and volume of the channel …”);
perform a topology optimization loop of the one or more design goals (Page.2, left column, “We employ automatic differentiation to compute the gradients, a.k.a. the sensi-tivity field, of the producibility criterion and augment it with the performance sensitivity fields. Subsequently, the augmented sensitivity field is used to optimized the design with respect to both performance and producibility. We demonstrate that closing the loop from manufacturing to design in this manner allows for TO to successfully stay away from crack-prone designs as illustrated in Fig. 1.” Page13, 5. Results, “ At every TO iteration, the design variables are updated using the Method of Moving Asymptotes (MMA) …”), the topology optimization loop comprising:
providing an initial set of the one or more parameters to a surrogate model emulating, using a trained machine-learning model, a failure simulation of the structural model based on one or more training datasets, wherein the failure 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, volume fractions, and manufacturing constraints (Figure 1: … PATO leverages an attention-based neural network surrogate model trained on a diverse set of optimized designs …” Page.3, top left, “To efficiently predict cracking for each candidate design generated within the TO loop, we train an attention-based CNN surrogate model [5].” Page.6, 4.3. Training Data Generation, “Since the surrogate model will be used within the TO loop, it is essential that sufficient training data is pro-vided by TO under different physics and boundary condi-tions, while considering relevant design and manufactur-ing constraints …” Page.7, 4.5. Training Data Evaluation, “Training data evaluation involves simulating the addi-tive process for each training sample … The build process is simulated using com-mercially available Simufact Additive 2020 FP1 software package [66].” Page.8, Figure 7: Training data was generated for different boundary conditions, volume fractions, and design and manufacturing constraints.; Page.12, Figure 18.);
computing, using the surrogate model, an output that includes a crack index for the structural model (Page.1, Abstract, “Multiple crack indices are explored and using experimental validation, maximum shear strain index (MSSI) is shown to be an accurate crack index … We leverage the current advances in deep convolutional neural networks (DCNN) and present a high-delity surrogate model based on an Attention-based U-Net architecture to predict the MSSI values as a spatially varying field over the part's domain.” Page.13, bottom left and top right, “To compute the maximum MSSI sensitivity field, we ex-tend the surrogate model using a max pooling layer with a pool size as large as the domain size. This essentially al-lows the surrogate model to predict the peak MSSI value rather than the full MSSI field.” Page.6, left column, “the MSSI criterion indicates a high value for the observed crack location with good precision. As a result, MSSI was chosen as the crack index for the outcomes described in this paper.”);
applying a computational layer to the output of the surrogate model to obtain one or more gradients by automatic differentiation and a predicted value of a failure criterion associated with the crack index (Page.13, bottom left, “To compute the maximum MSSI sensitivity field, we ex-tend the surrogate model using a max pooling layer with a pool size as large as the domain size. This essentially al-lows the surrogate model to predict the peak MSSI value rather than the full MSSI field. Subsequently, we use au-tomatic differentiation capabilities in TensorFlowTM [71] to compute the change of the output of the NN (i.e., max-imum MSSI) with respect to hypothetical change in the in-put design variable (i.e., pseudo-density).” See also Figure 18); 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 (Page.13, “This essentially al-lows the surrogate model to predict the peak MSSI value rather than the full MSSI field. Subsequently, we use au-tomatic differentiation capabilities in TensorFlowTM [71] to compute the change of the output of the NN (i.e., max-imum MSSI) with respect to hypothetical change in the in-put design variable (i.e., pseudo-density) … At every TO iteration, the design variables are updated using the Method of Moving Asymptotes (MMA) [72], where the problem is approximated by a number of con-vex sub-problems which are solved using the Interior Point Method [73].), wherein the computing the updated set of parameters comprises:
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; the crack index of the structural model considering thermal and residual stress during manufacturing; a weighting factor; and a target volume fraction (
PNG
media_image1.png
692
604
media_image1.png
Greyscale
Page.1, Abstract, “Specifically, parts fabricated through Laser Powder Bed Fusion (LPBF) are prone to defects such as warpage or cracking due to high residual stress values generated from the steep thermal gradients produced during the build process. To ensure that the design is crack free during optimization, producibility is explicitly encoded within the standard formulation of TO, using a crack index. Multiple crack indices are explored and using experimental validation, maximum shear strain index (MSSI) is show to be an accurate crack index. Simulating the build process, in order to estimate MSSI, is a coupled, multi-physics, time-complex computation …” Introduction, “These rapid heating and cooling cycles can lead to steep thermal gradients in the part that can result in generation of high residual stresses. These residual stresses can manifest in part deformation or if sufficiently high, result in cracking at multiple locations of the part [3].” Page.6, “After adjusting the thresholds, it can be seen that the second criterion, i.e., the MSSI criterion indicates a high value for the observed crack location with good precision. As a result, MSSI was chosen as the crack index for the outcomes described in this paper.” Page.13, “For compliance minimization problems, since Ꝕ = uT f … Figure 19 illustrates the optimized designs at different volume fractions for TO without considering producibility (w = 0:0) and the proposed PATO (w = 0:95) while min-imizing thermal compliance under the loading condition of Fig. 7a. Examiner note: the reference teaches minimizing, with respect to design variable, a weighted objective considering performance and MSSI cracking index, with a weighting factor. The reference further teaches minimizing thermal compliance under a loading condition, thereby teaching the recited performance function of the structural model under design load. The reference also teaches that high residual stresses generated from steep thermal gradients produced during the additive manufacturing build process can result in cracking, and further teaches MSSI as an accurate crack index estimated by simulating the build process, thereby teaching the recited crack index considering thermal and residual tress during manufacturing, and further teaches a state equation solved using finite element analysis and a volume constraint, wherein Vtarget is the target volume fraction); and
producing, (Page.14, “We also conducted experimental validation of our outcome by printing multiple coupons of the optimal design discovered by our approach. It is clear from the figure (bottom right), and based on a thorough inspection, that this coupon has no cracks either in the channel or any other region of the coupon, further validating that our approach converges to a design that is truly crack free. Page.3, bottom left, “we focus on TO for LPBF-AM, where a laser beam melts and sinters metal powder in a layer-by-layer fashion to fabricate the part.”).
However, Iyer (PATO) fails to teach computing 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; producing, in a production compartment comprising a three-dimensional printer, a physical model.
Harris teaches computing 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.” [0198], “To determine whether the generative design has converged to a stable solution, the check 192 can identify the condition in which all the design constraints are met and no design objectives have improved significantly since the last one or more iterations.” 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.);
producing, in a production compartment comprising a three-dimensional printer, a physical model ([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.).
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 Iyer (PATO) to incorporate the teachings of Harris, and apply convergence check during the iterative topology optimization process, in order to control convergence and reduce oscillations and radical changes during optimization, thereby obtaining and producing a stable optimized structural design.
Claim 2, Iyer (PATO) 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, and wherein the processing device unit further to:
provide the updated set of parameters to the surrogate model in a next computation cycle; and
identify the set of design parameters based on one or more cycles of computation upon satisfying the convergence criterion (Page.2, bottom right, “We show that PATO is able to successfully converge towards the discovery of crack-free designs …” Fig.18 showing the design variable/density supplied to the NN surrogate within the iterative PATP topology optimization framework; Page.13, 5. Results, “At every TO iteration, the design variables are updated using the Method of Moving Asymptotes (MMA) [72] …” Page. 13-14, Figs.19-22, showing the resulting optimized designs and starting that “the design suggested by PATO” provide the optimized crack-free design)).
Claim 7, Iyer (PATO) 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; and
the performance conditions including at least one of a boundary condition, a loading condition, or a thermal condition of the structural model (Page.4, “In our experiment, the primary parameter is the shape and volume of the channel.” Page.5, “Only the room temperature mechanical proper-ties such as modulus, Poisson's ratio and flow stress ver-sus plastic strain is sufficient to fully define the material.” Page.6, “Since the surrogate model will be used within the TO loop, it is essential that sufficient training data is pro-vided by TO under different physics and boundary condi-tions, while considering relevant design and manufactur-ing constraints … Figures 7(a) and (b) illustrate the thermal conduction problem … In addition to the constant thermal and pressure loading, a set of 3D TO problems has been considered featuring varying dis-tribution of surface loading on the surface of the channels …” Figure 7: Training data was generated for different boundary conditions, volume fractions, and design and manufacturing constraints.).
Claim 8, Iyer (PATO) 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.6, “Since the surrogate model will be used within the TO loop, it is essential that sufficient training data is pro-vided by TO under different physics and boundary condi-tions, while considering relevant design and manufactur-ing constraints … Figures 7(a) and (b) illustrate the thermal conduction problem … Figure 7c shows the hydrostatic pressure prob-lem … Figures 7(e) and (f) illustrate an asymmetrical thermal problem and the corresponding optimized designs at different volume fractions, respectively … a set of 3D TO problems has been considered featuring varying dis-tribution of surface loading on the surface of the channels.”).
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 Iyer (PATO), 5. Results; See also Harris, abstract, [0050] and [0014]).
Claim(s) 4 and 13 are rejected under 35 U.S.C. 103 as being unpatentable over Iyer (PATO) and
Harris as applied to claims 1 and 9 above, and further in view of Han US20220075911A1 and Sakai US20040050318A1.
Claim 4, Iyer (PATO) teaches The apparatus of claim 1, wherein the failure criterion comprises (page.1, Abstract, “… parts fabricated through Laser Powder Bed Fusion (LPBF) are prone to defects such as warpage or cracking due to high residual stress values generated from the steep thermal gradients produced during the build process … Multiple crack indices are explored and using experimental validation, maximum shear strain index (MSSI) is shown to be an accurate crack index.” Page.3, “the highly localized and transient input of tremendous energy throughout a LPBF-AM process sub-jects every region of the part to multiple phase changes in rapid heating and cooling cycles. This results in high thermal gradient and subsequently residual stresses that can give rise to detrimental cracking of the part … In this work, we explored applicability of three parame-ters, … strain failure index (SFI), maximum shear strain index (MSSI) and to-tal strain energy density index (TSI), as criteria to predict crack likelihood.” Introduction, “These rapid heating and cooling cycles can lead to steep thermal gradients in the part that can result in generation of high residual stresses. These residual stresses can manifest in part deformation or if sufficiently high, result in cracking …” “Page.6, “it can be seen that the second criterion, i.e., the MSSI criterion indicates a high value for the observed crack location with good precision. As a result, MSSI was chosen as the crack index for the outcomes described in this paper.” ).
However, Iyer (PATO) and Harris fail to teach a maximum shear stress index (MSSI) and the MSSI is a function of thermal variables.
Han teaches a maximum shear stress index (MSSI) exceeding a threshold value for indicating cracking of the structural model during manufacturing, ([0022] The strength criterion of the structural materials is summarized as: θ(σ(x))≥θcrit and x∈Ω, where Ω denotes the set of all points in the structure, θ(σ(x)) denotes the strength function of the stress tensor σ(x) at point x of the stress field, θcrit is the strength of the structural material.” [0048], “… define it as θ(σ(x))=σ1−σ3 according to the theory of maximum shear stress …” [0050], “when the complex stress reaches a specific limit state, the structure will begin to fail. The strength criterion is the particular characterization of these ultimate stress states.” [0051], “… strength theory … can be used to predict the initiation of cracks on the structure … It uses the material strength as a clear determination indicator to introduce a peridynamic model to predict the cracks' initiation and propagation process on the structure.” Examiner note: the reference defines a strength criterion θ(σ(x))≥θcrit, wherein θ(σ(x)) is a strength function of the stress tensor and θcrit is the strength of the structural material. The reference further defines the same strength function as θ(σ(x))=σ1−σ3 according to maximum shear stress theory. Thus, the maximum shear stress based strength function is quantitatively evaluated against a threshold corresponding to material strength. The reference further teaches that reaching this limit state indicates structural failure and that the strength criterion is used to predict crack initiation. Accordingly, the maximum shear stress based strength function corresponds to the claimed maximum shear stress index used for indicating cracking).
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 Iyer (PATO) and Harris to incorporate the teachings of Han, and apply a maximum shear stress based strength criterion that compares the maximum shear stress based strength function with a threshold corresponding to material strength, in order to provide a clear indicator for determining structural failure and predicting crack initiation.
However, Iyer (PATO) and Harris and Han fail to teach the index is based on a maximum shear stress index.
Sakai teaches the MSSI is a function of thermal related variables ([0049], “In the embodiment according to the present invention, the entire cylindrical surface of CaF 2 crystal was isothermal, and the maximum shear stress among thermal stress that is determined by the size of CaF2 crystal and cooling rate is approximated as follows by considering parameters including thermal diffusivity, specific heat, Young's modulus, and Poisson's ratio of CaF2 single crystal: τMAX=9E−6·exp(2E−3·T)·D 2 ·R (3)).” [0050]
where τ MAX is the maximum shear stress [MPa], D is a thickness of crystal, R is a cooling rate, and T is temperature [K] of CaF2.” Examiner note: the reference defines the maximum shear stress τMAX as a function including temperature T, wherein T is the temperature of the material. Thus. The maximum shear stress equality varies as a function of a thermal variable.
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 Iyer (PATO) and Harris and Han to incorporate the teachings of Sakai, and apply determining the maximum shear stress as a function of temperature in the failure criterion, in order to account for changes in maximum shear stress caused by thermal conditions when evaluating structural failure, thereby improving the accuracy of predicting thermally induced structural failure under varying temperature conditions.
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.
Claim(s) 6 and 15 are rejected under 35 U.S.C. 103 as being unpatentable over Iyer (PATO) and
Harris as applied to claims 1 and 9 above, and further in view of Zeng (“Optimization of support structures for selective laser melting,” published on Aug. 2015”).
However, Claim 6, Iyer (PATO) and Harris fail to teach, but Zeng teaches The apparatus of claim 1, wherein the failure simulation of the structural model simulates an additive manufacturing process in the additive production compartment (Problem statement, page.4, “… A thermo-mechanical simulation tool for SLM developed at the University of Louisville is being commercialized by 3DSIM, LLC. It is able to simulate the temperature and thermal stress distribution in the SLM process in a fine-scale … and makes possible simulating the complete problem for an SLM process.” Page.5, “the 3DSIM model able to simulate the SLM process using the same scan pattern as used in an SLM machine to study the influence of various scan patterns on the temperature, stress distribution and microstructure of parts made using SLM.”Page.99, 6.1 Introduction, “Selective laser melting (SLM) is an Additive Manufacturing (AM) technology whereby a part is built by melting successive layers of metal powder. The part is typically built upon a plate in an enclosed chamber.”) to:
form the structural model layer by layer (Page.19, “The computational domain consists of five layers of 1x1x0.03mm Titanium powder deposited on the mild steel substrate … The model was also built in ANSYS and the functionality of element death and birth is applied for new layer deposition.” Page.119, “This model will be solved using a thermomechanical eigenmodal strategy layer-by-layer with a contact model to account for snapping of supports at their interface with the baseplate and/or the part. At the end of each simulation, the dimensions of the solidified struts will be periodically updated as a function of layer-by-layer buildup of the fabricated structure. In the meanwhile, it results in updated element properties from powder to solid with an updated total stiffness matrix.”);
for a position in each layer, heat a material from a first solid state to a fluid state (Page. 10, 2.3.1 SLM Thermal Phenomenon, “A high energy density laser beam is applied in SLM to melt the powder material.” Page.11, “During the heating and cooling process, the material undergoes phase changes from solid to liquid and back to solid …”); and
allow the material in the fluid state to dissipate heat and return to a second solid state (Page. 10-11, 2.3.1 SLM Thermal Phenomenon, “A high energy density laser beam is applied in SLM to melt the powder material. The melt pool solidifies quickly to form the bulk part … The laser energy is dissipated in several ways as shown in Figure 6 … Most of the remaining energy is conducted by the powder and air trapped among powder particles through the powder bed and the base plate. During the heating and cooling process, the material undergoes phase changes from solid to liquid and back to solid …”); wherein the updated version of the structural model is prevented from cracking due to thermal conditions related to phase changes (Page. Vii, “Supports are optimized and designed based on the thermal stress accumulated in parts as they are made … The support structure is designed to be withstand the thermal stress at locations where it could cause damage to the part and support structure …” Page.119, “Thermally induced stress is calculated using 3DSIM tools and optimized support parameters are calculated …The final support structure parameters are the maximal value of all scenarios and it is updated iteratively.” Page. 41, “Two optimization runs of support structure stiffening led to a reduction of thermal stress of about 72% … The experimental results clearly show that the specimens with standard block supports have a higher stress level than the optimized support designed specimens.”).
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 Iyer (PATO) and Harris to incorporate the teachings of Zeng, and apply a simulated layer by layer additive manufacturing process including melting and solidification phase changes, while updating the structural design based on thermally induced stresses so as to reduce or avoid cracking or damaging during fabrication, in order to account for the thermal stresses generated during the additive manufacturing process when evaluating and optimizing the structural design, thereby producing an updated structural design having reduced thermal stress and improved resistance to cracking or damaging during additive manufacturing.
The elements of claim 15 is substantially the same as those of claim 6. Therefore, the elements of claim 15 is rejected due to the same reasons as outlined above for claim 6.
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