DETAILED ACTION
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Claims 1-16 are present for examination
This action is Non-Final rejection.
Priority
Acknowledgment is made of applicant’s claim for domestic priority for provisional application 63/286,566 filed on 12/07/2021.
Information Disclosure Statement
The IDS filed on 02/22/2023 is reviewed and see the attached file for consideration.
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 2 and 3 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.
As of claim 2 , “MapNet” is considered indefinite since according to the specification on Para 65 “An ANN model, entitled as MapNet” and para 74, “MapNet can be found in the code for MapNet at GitHub repository” but the file was committed in GitHub after domestic priority date. Therefore under broadest interpretation, the examiner interpreted MapNet as a neural network for the rest of the claims. While claim 3 is also dependent on claim 2, so it is also rejected in the same scope.
Claim Rejections - 35 USC § 103
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
Claims 1, 2, 5,and 7 are rejected under 35 U.S.C. 103 as being unpatentable over, Heng 1 et al. "Universal machine learning for topology optimization." Computer Methods in Applied Mechanics and Engineering 375 (2020) in the view of Garcia, Xavier, and Adrian Rodriguez-Herrera. "Machine learning applied in the multi-scale 3D stress modelling." arXiv preprint arXiv:2008.11244 (2020) further in the view of Lee, Seunghye, et al. "CNN-based image recognition for topology optimization." Knowledge-Based Systems 198 (2020):
As of claim 1 Heng 1 teaches A computer-implemented method for computing a mechanical field of a structure ( section 4.1, Topology optimization is an iterative process which often involves hundreds of steps. Every time we come up with a new design, we need to solve for the structural response of the current design to compute the sensitivity information. For large-scale topology optimization, this procedure is computationally intensive. A large amount of history data (e.g., design variables, their corresponding sensitivities, and displacement solutions) are generated during topology optimization, of which we typically do not make full use).
modelling the structure by a fine-scale structure, wherein the fine-scale structure is obtained by dividing the structure into a plurality of fine-scale elements (section 3, In this section, the topology optimization formulation for the classical compliance-minimization problem is briefly reviewed. Throughout, we assume that the design domain is discretized by a finite element mesh and adopt the standard density-based approach [4,33], where the material distribution is characterized by an element-wise constant function. For a given finite element mesh with N nodes and M elements, we denote f ∈ RdN×1 as the applied global force vector).
applying a finite element method (FEM) to the coarse scale structure to calculate a coarse-scale mechanical field of the structure (section 4.1, algorithm 1, step 4-7, Filter design variables: z(k) = Pz(k); Assemble the global stiffness matrix KC on the coarse-scale mesh based on (14); Solve the state equation on coarse-scale mesh: uC = (KC)−1fC; Evaluate the strain vector εC,(k) on the coarse-scale mesh based on uC).
fragmenting the coarse-scale mechanical field into a plurality of fragments, whereby an individual fragment has a fragment boundary on the coarse-scale mechanical field such that the individual fragment is a local portion of the coarse-scale mechanical field within the fragment boundary( section 2, we devise a tailored two-scale topology optimization formulation, which allows for the training of machine learning models based on local features of the topology optimization.
PNG
media_image1.png
400
1141
media_image1.png
Greyscale
For a given coarse-scale finite element with a total of nG integration points, as illustrated in Fig. 2, we divide it into a total of nG sub-regions and each sub-region is associated with one of its integration point).
combining the respective local portions of the fine scale mechanical field to generate the fine-scale mechanical field;( section2 and algorithm 1 step 21-23, The proposed framework is capable of handling 3D large-scale design of a wide range of problem sizes while achieving significant speedup. For example, we demonstrate that the proposed framework can achieve close to an order of magnitude speedup in a 3D design problem with more than 1 million design variables.21. Use the machine learning model to predict ˜G (k) based on the input z(k) and εC; 22,Compute the predicted sensitivity as ˜G(k) = P⊤˜G(k). 23, Update z(k+1) using ˜G(k) based on (8)).
Heng 1 also teaches wherein a first number of respective coarse-scale elements in the plurality of coarse-scale elements is less than a second number of respective fine-scale elements in the plurality of fine-scale elements; (section 4.3.1, Assuming that the number of elements of the coarse-scale mesh is much smaller than that of the fine-scale mesh, the time spent in solving the state equation on the coarse-scale mesh will be negligible. The setup of the coarse-scale and fine-scale meshes is illustrated in Fig. 2. We note that although the illustration is in 2D, the numerical examples in this work also consider 3D problems.) but Heng 1 does not explicitly teach coarsening the fine-scale structure to yield a coarse-scale structure such that the coarse-scale structure is composed of a plurality of coarse-scale elements.
While Garcia teaches coarsening the fine-scale structure to yield a coarse-scale structure such that the coarse-scale structure is composed of a plurality of coarse-scale elements(
PNG
media_image2.png
514
1081
media_image2.png
Greyscale
as on fig 1 high resolution model is upscaled to low resolution and one element or sample of coarse scale is used to compute coarse properties)
using an artificial neural network (ANN) to map the local portion of the coarse-scale mechanical field to a corresponding local portion of the fine-scale mechanical field, whereby respective local portions of the fine-scale mechanical field for the plurality of fragments are computed( abstract, Fig 8, This paper proposes a methodology to estimate stress in the subsurface by a hybrid method combining finite element modeling and neural networks… section 3.1, If this is
the case, then we argue that there can be a transformation f that within an acceptable error margin would map the solutions of the partial differential equations in-between scales:
PNG
media_image3.png
56
521
media_image3.png
Greyscale
The term f(
PNG
media_image4.png
12
13
media_image4.png
Greyscale
*; xi) in 4 represents such a generic transformation (or function). The terms
PNG
media_image4.png
12
13
media_image4.png
Greyscale
ij and
PNG
media_image4.png
12
13
media_image4.png
Greyscale
*ij are respectively the solutions in the fine-scale and the coarse scale. We propose to use techniques of machine learning to obtain f and the workflow to implement this idea). The examiner interpreted ANN as neural network, the mechanical field as stress and it map stress of the coarse scale to the fine scale as shown on equation 4 and Fig 8.
setting the generated fine-scale mechanical field as the mechanical field of the structure, thereby allowing the fine-scale mechanical field with a higher accuracy than the coarse-scale mechanical field to be used as the mechanical field without a need to use the FEM to directly compute the entire fine-scale mechanical field from the fine-scale structure for computation cost saving (Introduction, Given the relatively large volumes of interest, the computational cost in this kind of problems can render these workflows impractical. One alternative to cope with the computational over head is to make use of upscaling techniques. These aim at solving the relevant equations in a coarse and manage able resolution. Previous to simulation, the properties of the physical system are represented as effective properties at the coarse scale while attempting to capture as much as possible the characteristic behavior of the fine scale).
Garcia is considered to be analogous with Heng 1 and the claim invention, since they focus on multi scale analysis of large structure for optimal solution. Therefore it would be obvious to try for a person of ordinary skill in the art to use artificial neural network in Garcia teaching of using neural network and coarsening(upscaling) fine scale into coarse scale on Heng 1 model to apply fragmentation on the low resolution to compute mechanical field.
The motivation would have been to minimize the computation by computing stress on small size elements since as the size of the model increases, or as more resolution is needed, the number of cells in the model increases and the computational cost increases linearly ( Garcia, section 2.2).
The modified model of Heng 1 and Garcia do not explicitly teach computing a fine-scale mechanical field of the structure from the coarse-scale mechanical field, wherein the computing of the fine-scale mechanical field from the coarse-scale mechanical field comprises:
While Lee teaches computing a fine-scale mechanical field of the structure from the coarse-scale mechanical field (section 3.2, The number of links between the input nodes and the first CNN hidden layers can easily go to the order of millions. Moreover, finely discretized meshes require significant computational time for the whole domain. Therefore, in the proposed method, instead of training the CNN model by using the target mesh information, dataset of lower resolution was used to train the CNN structure. Finally, to predict the compliance information of the original resolution)
Lee is considered to be analogous to the modified model and the claim invention since they focus on multi scale analysis of large structure. Therefore it would be obvious for a person of ordinary skill in the art to compute a mechanical field of original resolution of the whole structure from a coarse structure and apple fragmenting the coarse scale as the modified model teaches.
The motivation would have been to can eliminate the step of finite element analysis and accelerate topology optimization processes by the training CNN structure using coarse elements, after the training process, compliance information of domains composed with finer elements can be predicted ( Lee Conclusion).
As of claim 2, the modified model of Heng 1, Garcia and Lee teach all the limitations of claim 1, and Heng 1 also teaches wherein the ANN is implemented as MapNet( Section 4.2 Notice that our proposed framework is independent of any specific implementation of the machine learning module. Thus, other machine learning and deep machine learning models, such as Support Vector Machine (SVM) [42,43], Convolutional Neural Networks (CNNs) [44,45] and Residual Networks (ResNets) [46], and hybridized models (e.g., Principal Component Analysis (PCA) [47,48] and DNN), can be directly applied in the proposed framework). The examiner interpreted MapNet as DNN , since DNN is used in the prior art for the same functional use (mapping) as MapNet and it teaches in claim 1 with the combined model. .
As of claim 5, the combined model of Heng 1, Garcia and Lee teach all the limitations of claim 1, and Lee also teaches wherein in coarsening the fine-scale structure to yield the coarse-scale structure, the coarse-scale structure is obtained by scaling down a fine scale density field to give a coarse-scale density field (section 3.1 and Fig 2, All the material density information of each element obtained in the previous iteration, are fed into the proposed method. However, because the topology information has been discretized by a fine mesh, the resizing process is needed to convert the fine mesh domain into a coarse mesh domain. These resizing processes are due to the trained CNN model by using coarse mesh domains )
PNG
media_image5.png
339
791
media_image5.png
Greyscale
wherein the fine-scale structure defines the fine-scale density field, and wherein the coarse-scale density field defines the coarse-scale structure (section 3.2, As shown in Fig. 3, the topology image is the input of the CNN model; the material density information of each discretized cell can be an input node. The number of links between the input nodes and the first CNN hidden layers can easily go to the order of millions. Moreover, finely discretized meshes require significant computational time for the whole domain. Therefore, in the proposed method, instead of training the CNN model by using the target mesh information, dataset of lower resolution was used to train the CNN structure. Finally, to predict the compliance information of the original resolution).
As of claim 7, the modified model of Heng 1, Garcia and Lee teach all the limitations of claim 1, and Heng 1 also teaches wherein in fragmenting the coarse-scale mechanical field into the plurality of fragments, fragment overlapping among respective fragments in the plurality of fragments is absent (section 4.3.1 and FIG 3, every element in the coarse-scale mesh contains the same number of elements in the fine-scale mesh. Thus, we introduce a parameter called block size NB to quantify how many fine-scale elements are contained on each side of the coarse-scale element. For example, the illustration in Fig. 2 has a block size of NB = 5, meaning every element in the coarse-scale mesh constrains 5 × 5 = 25 fine-scale elements)
PNG
media_image6.png
353
1393
media_image6.png
Greyscale
Claim 3 is rejected under 35 U.S.C. 103 as being unpatentable over Chi, Heng 1 et al. "Universal machine learning for topology optimization." Computer Methods in Applied Mechanics and Engineering 375 (2021) in the view of Garcia, Xavier, and Adrian Rodriguez-Herrera. "Machine learning applied in the multi-scale 3D stress modelling." arXiv preprint arXiv:2008.11244 (2020) in the view of Lee, Seunghye, et al. "CNN-based image recognition for topology optimization." Knowledge-Based Systems 198 (2020), in the view of Yu, Yonggyun, et al. "Deep learning for determining a near-optimal topological design without any iteration." Structural and Multidisciplinary Optimization 59.3 (2019), further in the view of Heng 2 chi (WO2020160099A1).
As of claim 3, the modified model of Heng 1, Garcia and Lee teaches all the limitations of claim 2 but they do not explicitly teach wherein the MapNet comprises plural convolutional layers, plural deconvolutional layers and a residual block, wherein each of the convolutional and deconvolutional layers has a filter size of3x3, a stride of2x2 except for a last layer of the MapNet, and an activation function of RELU, wherein the last layer of the MapNet has a lxl stride, wherein inputs of the MapNet are the local portion of the coarse-scale mechanical field, and a corresponding local portion of a fine-scale density field within the fragment boundary, the fine-scale density field being defined by the fine-scale structure, and wherein at each deconvolutional layer, a density field with the same scale down sampled from the fine-scale density field is added.
While Yu teaches wherein the MapNet comprises plural convolutional layers, plural deconvolutional layers and a residual block, wherein each of the convolutional and deconvolutional layers has a filter size of3x3, a stride of2x2 except for a last layer of the MapNet, and an activation function of RELU, wherein the last layer of the MapNet has a lxl stride,(section 3.2, fig. 7 and 9,
PNG
media_image7.png
507
995
media_image7.png
Greyscale
Mass fraction, which is a scalar value and does not represent spatial information, was
directly inputted to the layer of latent variables. The encoder and decoder network except the output layer used a ReLU activation (Nair and Hinton 2010),).
Yu is considered to be analogous to the modified model and the claim invention since they focus on multi scale analysis for large structure. Therefor it will be obvious to try for a person of ordinary skill in the art to combine different convolutional, deconvolutional (up sampling) and residual size and stride value on the modified model to map the local portion of the coarse scale mechanical field to the corresponding local portion of the fine scale mechanical field .
The motivation would have been a convolutional neural network (CNN)-based encoder and decoder network is trained using the training dataset generated at low resolution and this proposed method can determine a near-optimal structure in terms of pixel values and compliance with negligible computational time (Yu, abstract)
The modified model does not explicitly teach , wherein inputs of the MapNet are the local portion of the coarse-scale mechanical field, and a corresponding local portion of a fine-scale density field within the fragment boundary, the fine-scale density field being defined by the fine-scale structure, and wherein at each deconvolutional layer, a density field with the same scale down sampled from the fine-scale density field is added.
While Heng 2 teaches wherein inputs of the MapNet are the local portion of the coarse-scale mechanical field, and a corresponding local portion of a fine-scale density field within the fragment boundary, the fine-scale density field being defined by the fine-scale structure, and wherein at each deconvolutional layer, a density field with the same scale down sampled from the fine-scale density field is added(Para 23- 24, Strain information on the coarse-scale mesh, together with the filtered design variables on the fine-scale mesh, are used as inputs to the machine learning-based model. In embodiments of this disclosure, machine learning module 115 employs fully- connected Deep Neural Networks (DNNs) as the universal function approximator that takes the input from the two-scale topology optimization module 111 and predicts the sensitivities of the compliance function)
Heng 2 is considered to be analogous to the modified model and the claim invention, since they focus on multi scale analysis for large structure. Therefore it would be obvious to try for a person of ordinary skill in the art to use the local portion of coarse scale and fine scale density as input in to the Neural network of the modified model to compute the mechanical field of the structure.
The motivation would have been A machine learning-based topology optimization framework provides a general approach which greatly accelerates the design process of large-scale problems in 3D ( Heng 2 para 13).
Claim 4 is rejected under 35 U.S.C. 103 as being unpatentable over Chi, Heng 1 et al. "Universal machine learning for topology optimization." Computer Methods in Applied Mechanics and Engineering 375 (2021) in the view of Garcia, Xavier, and Adrian Rodriguez-Herrera. "Machine learning applied in the multi-scale 3D stress modelling." arXiv preprint arXiv:2008.11244 (2020) in the view of Lee, Seunghye, et al. "CNN-based image recognition for topology optimization." Knowledge-Based Systems 198 (2020), further in the view of Heng 2 chi (WO2020160099A1).
As of claim 4 , the combined model of Heng 1, Garcia and Lee teach all the limitations of claim 1, but they do not explicitly teach wherein in mapping the local portion of the coarse-scale mechanical field to the corresponding local portion of the fine-scale mechanical field. The ANN predicts the corresponding local portion of the fine scale mechanical field according to a fine-scale density field and the local portion of the coarse-scale mechanical field, wherein the fine-scale structure defines the fine-scale density field.
While Heng 2 teaches wherein in mapping the local portion of the coarse-scale mechanical field to the corresponding local portion of the fine-scale mechanical field(Para 25 Coarse-scale mapping module 112 generates the coarse-scale mesh 302 based on fine-scale mesh 301, which is generated by fine-scale mapping module 114. For example, fine-scale elements 301 are mapped to coarse-scale mesh element 302 divided into sectors 302a, 302b, 302c, 302d according to shading of corresponding quadrant clusters of the fine-scale mesh elements 301, where the shading represents state variable values (e.g., strain) computed by the topology optimization module 111 for the current optimization step).The ANN predicts the corresponding local portion of the fine scale mechanical field according to a fine-scale density field and the local portion of the coarse-scale mechanical field, wherein the fine-scale structure defines the fine-scale density field (para 18, A large amount of historical data (e.g., design variables, their corresponding sensitivities, and displacement solutions) is generated during topology optimization, but typically, not all of the historical data is fully explored and used. In view of this, a universal machine learning approach is proposed herein to learn the mapping between the current design and their corresponding sensitivities from historical data. Once the machine learning model is trained, it can be employed in the later optimization steps to directly predict the sensitivities based on the current design without solving the state equations).
Heng 2 is considered to be analogous to the modified model and the claim invention, since they focus on multi scale analysis for large structure. Therefore it would be obvious to try for a person of ordinary skill in the art to map the mechanical fields using neural network of the modified model to compute the mechanical field of the structure.
The motivation would have been A machine learning-based topology optimization framework provides a general approach which greatly accelerates the design process of large-scale problems in 3D ( Heng 2 para 13).
Claim 6 is rejected under 35 U.S.C. 103 as being unpatentable over, Heng 1, et al. "Universal machine learning for topology optimization." Computer Methods in Applied Mechanics and Engineering 375 (2021) in the view of Garcia, Xavier, and Adrian Rodriguez-Herrera. "Machine learning applied in the multi-scale 3D stress modelling." arXiv preprint arXiv:2008.11244 (2020)in the view of Lee, Seunghye, et al. "CNN-based image recognition for topology optimization." Knowledge-Based Systems 198 (2020), further in the view of Huang, Junbin, and Klaus-Jürgen Bathe. "Overlapping finite element meshes in AMORE." Advances in Engineering Software 144 (2020).
As of claim 6, the modified model of Heng 1, Garcia and Lee teach all the limitations of claim 1, but they do not explicitly teach the presence of overlapping among respective fragments during fragmentation.
While Huang teaches wherein in fragmenting the coarse-scale mechanical field into the plurality of fragments, fragment overlapping among respective fragments in the plurality of fragments is present (section 2 and fig 1, the analysis domain is first divided into several subdomains, each of which is then meshed independently. If the domain decomposition is reasonable, each subdomain is of regular shape and can be given a regular conforming mesh. Any interpolation technique may be chosen for a mesh as long as the interpolation is compatible. Good candidates include the isoperimetric interpolations [1], the finite elements enriched by interpolation covers [21], and the overlapping finite elements [14–20].
PNG
media_image8.png
264
493
media_image8.png
Greyscale
Huang is considered to be analogous to the modified model and the claim invention, since they focus on multi scale analysis of large-scale structure. Therefore it would be obvious to apply Huang’s teaching of overlapping mesh on the modified model in order to compute a mechanical field for the structure.
The motivation would have been to create global interpolation with desired compatibility and accuracy by using the method of overlapping finite element meshes to couple these (local) fields (interpolated independently over each subdomain)(Huang, section 2)
Claims 8, 9, 11,14 and 16 are rejected under 35 U.S.C. 103 as being unpatentable over Heng 1, et al. "Universal machine learning for topology optimization." Computer Methods in Applied Mechanics and Engineering 375 (2021) in the view of Garcia, Xavier, and Adrian Rodriguez-Herrera. "Machine learning applied in the multi-scale 3D stress modelling." arXiv preprint arXiv:2008.11244 (2020), in the view of Lee, Seunghye, et al. "CNN-based image recognition for topology optimization." Knowledge-Based Systems 198 (2020), further in the view of Qian, Chao, and Wenjing Ye. "Accelerating gradient-based topology optimization design with dual-model neural networks." arXiv e-prints (2020).
As of claim 8, the modified model of Heng 1, Garcia and Lee also teach (b) computing a mechanical field of the candidate structure according to the method of claim 1( the combined model of Heng1 , Garcia and Lee teach computing of mechanical field, refer back to claim 1), but the modified model do not explicitly teach the rest limitations of claim 8.
Qian teaches A computer-implemented method for performing topology optimization (TO) of a structure according to a design requirement, the design requirement being specified as minimizing or maximizing an objective function subjected to one or more constraints, the method comprising the steps of (Abstract, Topology optimization (TO) is a common technique used in free-form designs. However, conventional TO-based design approaches suffer from high computational cost due to the need for repetitive forward calculations and/or sensitivity analysis, which are typically done using high-dimensional simulations such as Finite Element Analysis (FEA). In this work, neural networks are used as efficient surrogate models for forward and sensitivity calculations in order to greatly accelerate the design process of topology optimization).
A, selecting a candidate structure for testing whether the candidate structure satisfies the design requirement ( Fig 8, Initial design
PNG
media_image9.png
503
619
media_image9.png
Greyscale
( c) using the computed mechanical field to evaluate the objective function, the one or more constraints, or both of the objective function and the one or more constraints (Fig 8, step 2, objective function evaluation and sensitivity analysis using the dual model neural network, and section 2, A general topology optimization formulation for density based structure design is listed as follows: minmize𝜌𝑒:𝑓(𝒖(𝜌𝑒),𝜌𝑒) Subject to: 𝐾(𝜌𝑒)𝑈=𝐹 other constraints (2) 0≤𝜌𝑒≤1, 𝑒=1,…,𝑁 where 𝑓(𝒖,𝜌𝑒) is the objective function, 𝒖 is a state field that satisfies a linear or nonlinear state equation, 𝜌𝑒, is the density distribution, which is also the design variables, 𝐾𝑈=𝐹 is the equilibrium equation that needs to be satisfied and N is the total number of element).
(d) determining whether the candidate structure satisfies the design requirement;( Fig 8, step 3 , meeting stopping criterion)
( e) if the candidate structure does not satisfy the design requirement, updating the candidate structure and repeating the steps (b)-(e), otherwise setting the candidate structure that satisfies the design requirement as the structure obtained by TO (Fig 8. When the meeting criterion is not meet, it updates the information with gradient information and it goes back to step 2).
Qian is considered to be analogous to Heng1, Garcia and Lee since they focus on multiscale analysis of large structure. Therefore it would be obvious for a person of ordinary skill in the art to integrate the mechanical filed computed by the combined model in to the Qian flow chart of topology optimization to output optimize design based on the requirement by performing updates.
The motivation would have been by using neural networks as efficient surrogate models for forward and sensitivity calculations in order to greatly accelerate the design process of topology optimization (Qian ,abstract).
As of claim 9, the modified model of Qian, Heng 1, Garcia and Lee teach all the limitations of claim 8 and Lee also teaches wherein the objective function and the one or more constraints are related to a structural compliance minimization design problem, and wherein the mechanical field is a strain energy field( section 2.1, A topology optimization problem based on the SIMP method, where the objective function is to minimize compliance can be written as follows [12]:
PNG
media_image10.png
219
744
media_image10.png
Greyscale
where C is a compliance value; U and F indicate the global displacement and force vectors, respectively. Here the notation K is the global stiffness matrix, ue and ke denote the element displacement vector and stiffness matrix, respectively)
As of claim 11, the combined model of Heng 1, Garcia and Lee teach (b) computing a mechanical field of the candidate structure according to the method of claim 2( the combined model of Heng1 , Garcia and Lee teach computing of mechanical field by MapNet, refer back to claim 2), but they do not explicitly teach the rest of the limitations of claim 11.
While Qian teaches A computer-implemented method for performing topology optimization (TO) of a structure according to a design requirement, the design requirement being specified as minimizing or maximizing an objective function subjected to one or more constraints, the method comprising the steps of (Abstract, Topology optimization (TO) is a common technique used in free-form designs. However, conventional TO-based design approaches suffer from high computational cost due to the need for repetitive forward calculations and/or sensitivity analysis, which are typically done using high-dimensional simulations such as Finite Element Analysis (FEA). In this work, neural networks are used as efficient surrogate models for forward and sensitivity calculations in order to greatly accelerate the design process of topology optimization).
A, selecting a candidate structure for testing whether the candidate structure satisfies the design requirement ( Fig 8, Initial design
PNG
media_image9.png
503
619
media_image9.png
Greyscale
( c) using the computed mechanical field to evaluate the objective function, the one or more constraints, or both of the objective function and the one or more constraints (Fig 8, step 2, objective function evaluation and sensitivity analysis using the dual model neural network, and section 2, A general topology optimization formulation for density based structure design is listed as follows: minmize𝜌𝑒:𝑓(𝒖(𝜌𝑒),𝜌𝑒) Subject to: 𝐾(𝜌𝑒)𝑈=𝐹 other constraints (2) 0≤𝜌𝑒≤1, 𝑒=1,…,𝑁 where 𝑓(𝒖,𝜌𝑒) is the objective function, 𝒖 is a state field that satisfies a linear or nonlinear state equation, 𝜌𝑒, is the density distribution, which is also the design variables, 𝐾𝑈=𝐹 is the equilibrium equation that needs to be satisfied and N is the total number of element).
(d) determining whether the candidate structure satisfies the design requirement;( Fig 8, step 3 , meeting stopping criterion)
( e) if the candidate structure does not satisfy the design requirement, updating the candidate structure and repeating the steps (b)-(e), otherwise setting the candidate structure that satisfies the design requirement as the structure obtained by TO (Fig 8. When the meeting criterion is not meet, it updates the information with gradient information and it goes back to step 2).
Qian is considered to be analogous to Heng1, Garcia and lee since they focus on multiscale analysis of large structure. Therefore it would be obvious for a person of ordinary skill in the art to integrate the mechanical filed computed by the combined model in to the Qian flow chart of topology optimization to output optimize design based on the requirement by performing updates.
The motivation would have been by using neural networks as efficient surrogate models for forward and sensitivity calculations in order to greatly accelerate the design process of topology optimization (Qian ,abstract).
As of claim 14, the modified model of Heng 1, Garcia and Lee also teach (b) computing a mechanical field of the candidate structure according to the method of claim 5( the modified model of Heng1 , Garcia and Lee teach computing of mechanical field by method 5, refer back to claim 5), but the modified model do not teach the rest limitations of claim 14.
While Qian teaches A computer-implemented method for performing topology optimization (TO) of a structure according to a design requirement, the design requirement being specified as minimizing or maximizing an objective function subjected to one or more constraints, the method comprising the steps of (Abstract, Topology optimization (TO) is a common technique used in free-form designs. However, conventional TO-based design approaches suffer from high computational cost due to the need for repetitive forward calculations and/or sensitivity analysis, which are typically done using high-dimensional simulations such as Finite Element Analysis (FEA). In this work, neural networks are used as efficient surrogate models for forward and sensitivity calculations in order to greatly accelerate the design process of topology optimization).
A, selecting a candidate structure for testing whether the candidate structure satisfies the design requirement ( Fig 8, Initial design
PNG
media_image9.png
503
619
media_image9.png
Greyscale
( c) using the computed mechanical field to evaluate the objective function, the one or more constraints, or both of the objective function and the one or more constraints (Fig 8, step 2, objective function evaluation and sensitivity analysis using the dual model neural network, and section 2, A general topology optimization formulation for density based structure design is listed as follows: minmize𝜌𝑒:𝑓(𝒖(𝜌𝑒),𝜌𝑒) Subject to: 𝐾(𝜌𝑒)𝑈=𝐹 other constraints (2) 0≤𝜌𝑒≤1, 𝑒=1,…,𝑁 where 𝑓(𝒖,𝜌𝑒) is the objective function, 𝒖 is a state field that satisfies a linear or nonlinear state equation, 𝜌𝑒, is the density distribution, which is also the design variables, 𝐾𝑈=𝐹 is the equilibrium equation that needs to be satisfied and N is the total number of element).
(d) determining whether the candidate structure satisfies the design requirement;( Fig 8, step 3 , meeting stopping criterion)
( e) if the candidate structure does not satisfy the design requirement, updating the candidate structure and repeating the steps (b)-(e), otherwise setting the candidate structure that satisfies the design requirement as the structure obtained by TO (Fig 8. When the meeting criterion is not meet, it updates the information with gradient information and it goes back to step 2).
Qian is considered to be analogous to Heng1, Garcia and lee since they focus on multiscale analysis of large structure. Therefore it would be obvious for a person of ordinary skill in the art to integrate the mechanical filed computed by the combined model in to the Qian flow chart of topology optimization to output optimize design based on the requirement by performing update as well.
The motivation would have been by using neural networks as efficient surrogate models for forward and sensitivity calculations in order to greatly accelerate the design process of topology optimization (Qian ,abstract).
As of claim 16, the modified model of Heng 1, Garcia and Lee also teach (b) computing a mechanical field of the candidate structure according to the method of claim 7( the combined model of Heng1 , Garcia and Lee teach computing of mechanical field by the method of claim 7 , refer back to claim 7), but the modified model do not explicitly teach the rest limitations of claim 16.
While Qian teaches A computer-implemented method for performing topology optimization (TO) of a structure according to a design requirement, the design requirement being specified as minimizing or maximizing an objective function subjected to one or more constraints, the method comprising the steps of (Abstract, Topology optimization (TO) is a common technique used in free-form designs. However, conventional TO-based design approaches suffer from high computational cost due to the need for repetitive forward calculations and/or sensitivity analysis, which are typically done using high-dimensional simulations such as Finite Element Analysis (FEA). In this work, neural networks are used as efficient surrogate models for forward and sensitivity calculations in order to greatly accelerate the design process of topology optimization).
A), selecting a candidate structure for testing whether the candidate structure satisfies the design requirement ( Fig 8, Initial design
PNG
media_image9.png
503
619
media_image9.png
Greyscale
( c) using the computed mechanical field to evaluate the objective function, the one or more constraints, or both of the objective function and the one or more constraints (Fig 8, step 2, objective function evaluation and sensitivity analysis using the dual model neural network, and section 2, A general topology optimization formulation for density based structure design is listed as follows: minmize𝜌𝑒:𝑓(𝒖(𝜌𝑒),𝜌𝑒) Subject to: 𝐾(𝜌𝑒)𝑈=𝐹 other constraints (2) 0≤𝜌𝑒≤1, 𝑒=1,…,𝑁 where 𝑓(𝒖,𝜌𝑒) is the objective function, 𝒖 is a state field that satisfies a linear or nonlinear state equation, 𝜌𝑒, is the density distribution, which is also the design variables, 𝐾𝑈=𝐹 is the equilibrium equation that needs to be satisfied and N is the total number of element).
(d) determining whether the candidate structure satisfies the design requirement;( Fig 8, step 3 , meeting stopping criterion)
( e) if the candidate structure does not satisfy the design requirement, updating the candidate structure and repeating the steps (b)-(e), otherwise setting the candidate structure that satisfies the design requirement as the structure obtained by TO (Fig 8. When the meeting criterion is not meet, it updates the information with gradient information and it goes back to step 2).
Qian is considered to be analogous to Heng1, Garcia and Lee since they focus on multiscale analysis of large structure. Therefore it would be obvious for a person of ordinary skill in the art to integrate the mechanical filed computed by the combined model in to the Qian flow chart of topology optimization to output optimize design based on the requirement by performing update as well.
The motivation would have been by using neural networks as efficient surrogate models for forward and sensitivity calculations in order to greatly accelerate the design process of topology optimization (Qian ,abstract).
Claim 10 is rejected under 35 U.S.C. 103 as being unpatentable over, Heng 1, et al. "Universal machine learning for topology optimization." Computer Methods in Applied Mechanics and Engineering 375 (2021) in the view of Garcia, Xavier, and Adrian Rodriguez-Herrera. "Machine learning applied in the multi-scale 3D stress modelling." arXiv preprint arXiv:2008.11244 (2020) further in the view of Lee, Seunghye, et al. "CNN-based image recognition for topology optimization." Knowledge-Based Systems 198 (2020), in the view of Qian, Chao, and Wenjing Ye. "Accelerating gradient-based topology optimization design with dual-model neural networks." arXiv e-prints (2020), further in the view of Qu, Xueyong, et al. "Thermal topology optimization in optistruct software." 17th AIA A/ISSMO Multidisciplinary analysis and optimization conference. 2016.
As of claim 10, the modified model of Heng 1, Garcia, Lee and Qian teach all the limitations of claim 8, but they do not explicitly teach wherein the objective function and the one or more constraints are related to a thermal compliance minimization design problem, and wherein the mechanical field is a temperature field.
While Qu teaches wherein the objective function and the one or more constraints are related to a thermal compliance minimization design problem, and wherein the mechanical field is a temperature field ( Section 5, Thermal compliance can be used as objective function and/or constraint. It can be used together with all existing responses such as volume, displacement, etc. Thermal compliance[5] is defined as = 1/ 2 ´T P = 1/ 2 ´T [K + H]T c TCOMP T T (6) where [Kc] is the conductivity matrix, [H] is the convection matrix, {T} is the unknown temperature. When thermal compliance is minimized, temperature at grids where power is applied is minimized, which typically is highest in the structure).
Qu is analogous with modified model and the claim invention since it focuses on topology optimization. Therefore it would be obvious for a person of ordinary skill in the art to compute temperature field as a mechanical field on the modified model.
The motivation would have been thermal compliance optimization is shown to be significantly faster than previously implemented formulation of minimizing the maximum temperature of the entire structure (Qu, conclusion).
Claim 12 is rejected under 35 U.S.C. 103 as being unpatentable over Chi, Heng 1 et al. "Universal machine learning for topology optimization." Computer Methods in Applied Mechanics and Engineering 375 (2021) in the view of Garcia, Xavier, and Adrian Rodriguez-Herrera. "Machine learning applied in the multi-scale 3D stress modelling." arXiv preprint arXiv:2008.11244 (2020) in the view of Lee, Seunghye, et al. "CNN-based image recognition for topology optimization." Knowledge-Based Systems 198 (2020), in the view of Yu, Yonggyun, et al. "Deep learning for determining a near-optimal topological design without any iteration." Structural and Multidisciplinary Optimization 59.3 (2019), in the view of Heng 2 chi (WO2020160099A1) further in the view of Qian, Chao, and Wenjing Ye. "Accelerating gradient-based topology optimization design with dual-model neural networks." arXiv e-prints (2020).
As of claim 12, The modified model of Heng 1, Garcia ,Lee, Yu and Heng 2 also teaches (b) computing a mechanical field of the candidate structure according to the method of claim 3 ( the combined model of Heng 1, Garcia ,Lee, Yu and Heng 2 teach all the limitations of claim 3, refer back to claim 3), but the modified model do no explicitly teach the rest of claim 12.
While Qian teaches A computer-implemented method for performing topology optimization (TO) of a structure according to a design requirement, the design requirement being specified as minimizing or maximizing an objective function subjected to one or more constraints, the method comprising the steps of (Abstract, Topology optimization (TO) is a common technique used in free-form designs. However, conventional TO-based design approaches suffer from high computational cost due to the need for repetitive forward calculations and/or sensitivity analysis, which are typically done using high-dimensional simulations such as Finite Element Analysis (FEA). In this work, neural networks are used as efficient surrogate models for forward and sensitivity calculations in order to greatly accelerate the design process of topology optimization).
A, selecting a candidate structure for testing whether the candidate structure satisfies the design requirement ( Fig 8, Initial design
PNG
media_image9.png
503
619
media_image9.png
Greyscale
( c) using the computed mechanical field to evaluate the objective function, the one or more constraints, or both of the objective function and the one or more constraints (Fig 8, step 2, objective function evaluation and sensitivity analysis using the dual model neural network, and section 2, A general topology optimization formulation for density based structure design is listed as follows: minmize𝜌𝑒:𝑓(𝒖(𝜌𝑒),𝜌𝑒) Subject to: 𝐾(𝜌𝑒)𝑈=𝐹 other constraints (2) 0≤𝜌𝑒≤1, 𝑒=1,…,𝑁 where 𝑓(𝒖,𝜌𝑒) is the objective function, 𝒖 is a state field that satisfies a linear or nonlinear state equation, 𝜌𝑒, is the density distribution, which is also the design variables, 𝐾𝑈=𝐹 is the equilibrium equation that needs to be satisfied and N is the total number of element).
(d) determining whether the candidate structure satisfies the design requirement;( Fig 8, step 3 , meeting stopping criterion)
( e) if the candidate structure does not satisfy the design requirement, updating the candidate structure and repeating the steps (b)-(e), otherwise setting the candidate structure that satisfies the design requirement as the structure obtained by TO (Fig 8. When the meeting criterion is not meet, it updates the information with gradient information and it goes back to step 2).
Qian is considered to be analogous to Heng 1, Garcia ,Lee, Yu and Heng 2 since they focus on multiscale analysis of large structure. Therefore it would be obvious for a person of ordinary skill in the art to integrate the mechanical filed computed by the combined model in to the Qian flow chart of topology optimization to output optimize design based on the requirement by performing updates.
The motivation would have been by using neural networks as efficient surrogate models for forward and sensitivity calculations in order to greatly accelerate the design process of topology optimization (Qian ,abstract).
Claim 13 is rejected under 35 U.S.C. 103 as being unpatentable over Chi, Heng 1 et al. "Universal machine learning for topology optimization." Computer Methods in Applied Mechanics and Engineering 375 (2021) in the view of Garcia, Xavier, and Adrian Rodriguez-Herrera. "Machine learning applied in the multi-scale 3D stress modelling." arXiv preprint arXiv:2008.11244 (2020) in the view of Lee, Seunghye, et al. "CNN-based image recognition for topology optimization." Knowledge-Based Systems 198 (2020), in the view of Heng 2 chi (WO2020160099A1), further in the view of Qian, Chao, and Wenjing Ye. "Accelerating gradient-based topology optimization design with dual-model neural networks." arXiv e-prints (2020).
As of claim 13, the modified model of Heng 1, Garcia, Lee and Heng 2, also teaches (b) computing a mechanical field of the candidate structure according to the method of claim 4( the combined model of Heng 1, Garcia, Lee and Heng 2 teach computing of mechanical field by mapping the local portion of the coarse-scale mechanical field to the corresponding local portion of the fine-scale mechanical field, , refer back to claim 4), but the modified model do not explicitly teach the rest of claim 13.
While Qian teaches A computer-implemented method for performing topology optimization (TO) of a structure according to a design requirement, the design requirement being specified as minimizing or maximizing an objective function subjected to one or more constraints, the method comprising the steps of (Abstract, Topology optimization (TO) is a common technique used in free-form designs. However, conventional TO-based design approaches suffer from high computational cost due to the need for repetitive forward calculations and/or sensitivity analysis, which are typically done using high-dimensional simulations such as Finite Element Analysis (FEA). In this work, neural networks are used as efficient surrogate models for forward and sensitivity calculations in order to greatly accelerate the design process of topology optimization).
A, selecting a candidate structure for testing whether the candidate structure satisfies the design requirement ( Fig 8, Initial design
PNG
media_image9.png
503
619
media_image9.png
Greyscale
( c) using the computed mechanical field to evaluate the objective function, the one or more constraints, or both of the objective function and the one or more constraints (Fig 8, step 2, objective function evaluation and sensitivity analysis using the dual model neural network, and section 2, A general topology optimization formulation for density based structure design is listed as follows: minmize𝜌𝑒:𝑓(𝒖(𝜌𝑒),𝜌𝑒) Subject to: 𝐾(𝜌𝑒)𝑈=𝐹 other constraints (2) 0≤𝜌𝑒≤1, 𝑒=1,…,𝑁 where 𝑓(𝒖,𝜌𝑒) is the objective function, 𝒖 is a state field that satisfies a linear or nonlinear state equation, 𝜌𝑒, is the density distribution, which is also the design variables, 𝐾𝑈=𝐹 is the equilibrium equation that needs to be satisfied and N is the total number of element).
(d) determining whether the candidate structure satisfies the design requirement;( Fig 8, step 3 , meeting stopping criterion)
( e) if the candidate structure does not satisfy the design requirement, updating the candidate structure and repeating the steps (b)-(e), otherwise setting the candidate structure that satisfies the design requirement as the structure obtained by TO (Fig 8. When the meeting criterion is not meet, it updates the information with gradient information and it goes back to step 2).
Qian is considered to be analogous to Heng 1, Garcia, Lee and Heng 2 since they focus on multiscale analysis of large structure. Therefore it would be obvious for a person of ordinary skill in the art to integrate the mechanical filed computed by the combined model in to the Qian flow chart of topology optimization to output optimize design based on the requirement by performing update as well.
The motivation would have been by using neural networks as efficient surrogate models for forward and sensitivity calculations in order to greatly accelerate the design process of topology optimization (Qian ,abstract).
Claim 15 is rejected under 35 U.S.C. 103 as being unpatentable over, Heng 1 et al. "Universal machine learning for topology optimization." Computer Methods in Applied Mechanics and Engineering 375 (2021) in the view of Garcia, Xavier, and Adrian Rodriguez-Herrera. "Machine learning applied in the multi-scale 3D stress modelling." arXiv preprint arXiv:2008.11244 (2020)in the view of Lee, Seunghye, et al. "CNN-based image recognition for topology optimization." Knowledge-Based Systems 198 (2020), in the view of Huang, Junbin, and Klaus-Jürgen Bathe. "Overlapping finite element meshes in AMORE." Advances in Engineering Software 144 (2020) further in the view of Qian, Chao, and Wenjing Ye. "Accelerating gradient-based topology optimization design with dual-model neural networks." arXiv e-prints (2020).
As of claim 15, The modified model of Heng 1, Garcia, Lee and Huang also teach (b) computing a mechanical field of the candidate structure according to the method of claim 6 ( the combined model of Heng1 , Garcia, Lee and Huang teach computing of mechanical field by method of claim 6 , refer back to claim 6), but the modified model do not explicitly teach the rest limitations of claim 15.
While Qian teaches A computer-implemented method for performing topology optimization (TO) of a structure according to a design requirement, the design requirement being specified as minimizing or maximizing an objective function subjected to one or more constraints, the method comprising the steps of (Abstract, Topology optimization (TO) is a common technique used in free-form designs. However, conventional TO-based design approaches suffer from high computational cost due to the need for repetitive forward calculations and/or sensitivity analysis, which are typically done using high-dimensional simulations such as Finite Element Analysis (FEA). In this work, neural networks are used as efficient surrogate models for forward and sensitivity calculations in order to greatly accelerate the design process of topology optimization).
A), selecting a candidate structure for testing whether the candidate structure satisfies the design requirement ( Fig 8, Initial design
PNG
media_image9.png
503
619
media_image9.png
Greyscale
( c) using the computed mechanical field to evaluate the objective function, the one or more constraints, or both of the objective function and the one or more constraints (Fig 8, step 2, objective function evaluation and sensitivity analysis using the dual model neural network, and section 2, A general topology optimization formulation for density based structure design is listed as follows: minmize𝜌𝑒:𝑓(𝒖(𝜌𝑒),𝜌𝑒) Subject to: 𝐾(𝜌𝑒)𝑈=𝐹 other constraints (2) 0≤𝜌𝑒≤1, 𝑒=1,…,𝑁 where 𝑓(𝒖,𝜌𝑒) is the objective function, 𝒖 is a state field that satisfies a linear or nonlinear state equation, 𝜌𝑒, is the density distribution, which is also the design variables, 𝐾𝑈=𝐹 is the equilibrium equation that needs to be satisfied and N is the total number of element).
(d) determining whether the candidate structure satisfies the design requirement;( Fig 8, step 3 , meeting stopping criterion)
( e) if the candidate structure does not satisfy the design requirement, updating the candidate structure and repeating the steps (b)-(e), otherwise setting the candidate structure that satisfies the design requirement as the structure obtained by TO (Fig 8. When the meeting criterion is not meet, it updates the information with gradient information and it goes back to step 2).
Qian is considered to be analogous to Heng 1, Garcia, Lee and Huang since they focus on multiscale analysis of large structure. Therefore it would be obvious for a person of ordinary skill in the art to integrate the mechanical filed computed by the combined model in to the Qian flow chart of topology optimization to output optimize design based on the requirement by performing update as well.
The motivation would have been by using neural networks as efficient surrogate models for forward and sensitivity calculations in order to greatly accelerate the design process of topology optimization (Qian ,abstract).
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure,
Goel; Tushar (US-20100262406-A1, Date Published, 2010-10-14) this invention also teaches similar concept with the claim invention since it teaches a way of improvement for topology optimization for engineering product.
Lunati; Ivan Fabrizio (US-8594986-B2, Date Published, 2013-11-26) is also similar with the claim invention since this application also teaches multi scale finite volume method to compute pressure in the primary coarse-scale cells using the computed pressure in the dual coarse-scale cells.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to ABRHAM A. TAMIRU whose telephone number is (571)272-6987. The examiner can normally be reached Monday - Friday 8:00am - 5:00pm.
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, Ryan Pitaro can be reached at 571 272 4071. 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.
/A.A.T./Examiner, Art Unit 2188
/RYAN F PITARO/Supervisory Patent Examiner, Art Unit 2188