Artificial limb foot plate constant rigidity reverse design method based on deep neural network

Abstract

The invention discloses a deep neural network-based artificial limb foot plate constant stiffness reverse design method. The method comprises the following steps of: 1, parameterized script modeling; step 2, simulation output; step 3, constructing a data set; 4, constructing a deep neural network; 5, training a prediction model; step 6, testing and evaluating the model; 7, reversely outputting a design scheme; wherein in the step 3, a data set is established according to a simulation result; wherein in the step 4, a deep neural network is constructed by using a data set; wherein in the step 6, the network model is tested and evaluated; according to the systematic design method, the design efficiency can be remarkably improved, reverse structural parameter prediction is carried out on the prediction model through the deep neural network, meso-scale refined design can be carried out, the method is more reliable and accurate, network parameters can be optimized in real time in the design process for different design objects, and the design efficiency is improved. And the network model performance and design accuracy are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of prosthetic design, and specifically to a reverse design method for a prosthetic foot plate with a fixed stiffness based on a deep neural network. Background Art

[0002] The stiffness of a prosthetic foot plate is closely related to its energy storage characteristics and comfort, and is an important performance index. In actual application scenarios, the energy storage characteristics and comfort of the prosthetic foot plate must meet a given range. Good energy storage characteristics and wearing and walking comfort require the stiffness characteristic curve of the prosthetic foot plate to conform to specific standards. Existing design methods for the stiffness characteristics of prosthetic foot plates have the following defects: First, the existing designs lack systematic prosthetic design and prediction methods, and the design efficiency is low. Second, the current stiffness design methods for prosthetic foot plates are relatively rough and cannot perform refined design at the mesoscopic scale, and are not reliable enough. Third, the existing design methods are difficult to adapt to different design objects, and a large number of tests and reconstructions are required to complete the specific stiffness design. Summary of the Invention

[0003] The purpose of the present invention is to provide a reverse design method for a prosthetic foot plate with a fixed stiffness based on a deep neural network to solve the problems raised in the above background art.

[0004] To achieve the above purpose, the present invention provides the following technical solution: A reverse design method for a prosthetic foot plate with a fixed stiffness based on a deep neural network, including Step 1, parametric script modeling; Step 2, simulation output; Step 3, constructing a data set; Step 4, constructing a deep neural network; Step 5, training a prediction model; Step 6, model testing and evaluation; Step 7, reverse outputting a design scheme;

[0005] Among them, in the above Step 1, batch structural parametric modeling is performed on the prosthetic foot plate;

[0006] Among them, in the above Step 2, batch simulations of the stiffness of the forefoot and hindfoot are sequentially performed on each model;

[0007] Among them, in the above Step 3, a data set is established according to the simulation results;

[0008] Among them, in the above Step 4, a deep neural network is constructed using the data set;

[0009] Among them, in the above Step 5, a structural parameter prediction model for the prosthetic foot plate is trained;

[0010] Among them, in the above Step 6, the network model is tested and evaluated;

[0011] Among them, in the above Step 7, the structural parameters are reverse output to obtain a design scheme.

[0012] Preferably, in the first step, the Abaqus software is used to model with a parameterized script in the Python language, and the structural parameter variables of the prosthetic foot sole are set.

[0013] Preferably, the structural parameter variables are used to change the front foot sole thickness distribution, rear foot sole thickness distribution, sandwich thickness of the mid-foot lattice sandwich structure, core angle, core radius, and core number of the model.

[0014] Preferably, in the second step, according to the prosthetic foot sole test standard, batch simulations of the front foot and rear foot stiffness are performed on each model in turn, and the front and rear foot displacement-load curves are output from the simulation results to characterize the front and rear foot stiffness.

[0015] Preferably, in the third step, a script is written to extract the foot sole modeling structural parameters and the front and rear foot stiffness characteristic curves from the batch simulation result data, and a data set is established.

[0016] Preferably, in the fourth step, a deep neural network is constructed using the data set, with the stiffness characteristic curve as the input data of the neural network model and the corresponding prosthetic foot sole structural parameters as the output of the neural network, to construct a deep neural network model consisting of an input layer, a fully connected layer, a hidden layer, and an output layer.

[0017] Preferably, in the fifth step, the network model is used to learn the stiffness-structural parameter data, and the structural parameter prediction model of the prosthetic foot sole is trained to predict the structural parameters of the foot sole according to the specified stiffness curve. During the training process, the stochastic gradient descent algorithm is used to adjust and optimize the network parameters to further improve the performance of the network model.

[0018] Preferably, in the sixth step, an independent test set is used to test and evaluate the network model and the parameter model.

[0019] Preferably, the evaluation indexes of the test and evaluation include accuracy, recall rate, etc., and a validation set is used to evaluate the performance of the network model. According to the generalization ability and performance of the evaluation model, the network parameters are adjusted to achieve the best performance.

[0020] Preferably, in the seventh step, with the stiffness curve as the input parameter, the structural parameters of the prosthetic foot sole are inversely solved using the structural parameter prediction model with a preset stiffness to obtain the design scheme.

[0021] Compared with the prior art, the beneficial effects of the present invention are as follows: Through a systematic design method, the present invention can significantly improve the design efficiency. By using a deep neural network to inversely predict the structural parameters of the prediction model, fine-scale design at the mesoscopic scale can be carried out, which is more reliable and accurate. Moreover, for different design objects, the network parameters can be optimized in real time during the design process to improve the performance of the network model and the design accuracy. Description of the Drawings

[0022] Figure 1 is the method flow chart of the present invention;

[0023] Figure 2 is the test condition for the stiffness of the front foot of the present invention;

[0024] Figure 3 is the test condition for the stiffness of the rear foot of the present invention;

[0025] Figure 4 is the curve graph of the stiffness characteristics of the prosthetic foot plate of the present invention.

[0026] In the figure, a is the thickness of the sole of the rear foot of the rear foot, b is the thickness of the sole of the front foot, c is the number of cores, d is the core angle, and e is the core diameter. Detailed implementation manners

[0027] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0028] Please refer to Figures 1-4 , an embodiment provided by the present invention: a reverse design method for a prosthetic foot plate with a fixed stiffness based on a deep neural network, including Step 1, parametric script modeling; Step 2, simulation output; Step 3, constructing a data set; Step 4, constructing a deep neural network; Step 5, training a prediction model; Step 6, model testing and evaluation; Step 7, reverse output of the design scheme;

[0029] Among them, in the above Step 1, batch structural parametric modeling is performed on the prosthetic foot plate. By using the Abaqus software to apply the python language parametric script for modeling, the structural parameter variables of the prosthetic foot plate are set. The structural parameter variables are used to change the front foot sole thickness distribution, rear foot sole thickness distribution, sandwich thickness of the mid-foot lattice sandwich structure, core angle, core radius, and the number of cores of the model;

[0030] Among them, in the above Step 2, batch simulations of the stiffness of the front foot and the rear foot are sequentially performed on each model. According to the prosthetic foot plate test standard, batch simulations of the stiffness of the front foot and the rear foot are sequentially performed on each model, and the front and rear foot displacement load curves are output from the simulation results to characterize the front and rear foot stiffness;

[0031] Among them, in the above Step 3, a data set is established according to the simulation results. A script is written to extract the foot plate modeling structural parameters and the front and rear foot stiffness characteristic curves from the batch simulation result data, and a data set is established;

[0032] In the above step 4, a deep neural network is constructed using the data set, with the stiffness characteristic curve as the input data of the neural network model and the corresponding prosthetic foot sole structure parameters as the output of the neural network. A deep neural network model consisting of an input layer, a fully connected layer, a hidden layer, and an output layer is constructed.

[0033] In the above step 5, the structure parameter prediction model of the prosthetic foot sole is trained. The network model is used to learn the stiffness-structure parameter data, and the structure parameter prediction model of the prosthetic foot sole is trained to predict the structure parameters of the foot sole according to the specified stiffness curve. During the training process, the stochastic gradient descent algorithm is used to adjust and optimize the network parameters to further improve the performance of the network model.

[0034] In the above step 6, an independent test set is used to test and evaluate the network model and the parameter model. The evaluation indexes of the test and evaluation include accuracy, recall rate, etc., and a validation set is used to evaluate the performance of the network model. According to the generalization ability and performance of the evaluation model, the network parameters are adjusted to achieve the best performance.

[0035] In the above step 7, the stiffness curve is used as the input parameter, and the structure parameter prediction model is used to inversely solve the structure parameters of the prosthetic foot sole with the preset stiffness to obtain the design scheme.

[0036] Based on the above, the advantages of the present invention are as follows: Through a systematic design method, the present invention significantly improves the design efficiency of the prosthetic foot sole. By using a deep neural network to inversely predict the structure parameters of the prediction model, fine-scale design at the mesoscopic scale can be carried out. The gap between the model prediction and the real model is small, and the design is more reliable and accurate. For different design objects, the network parameters can be evaluated and optimized in real time during the design process to improve the performance of the network model and the design accuracy, having the advantage of strong adaptability.

[0037] For those skilled in the art, it is obvious that the present invention is not limited to the details of the above exemplary embodiments, and can be implemented in other specific forms without departing from the spirit or basic characteristics of the present invention. Therefore, from any point of view, the embodiments should be regarded as exemplary and non-limiting. The scope of the present invention is defined by the appended claims rather than the above description. Therefore, all changes falling within the meaning and scope of the equivalent elements of the claims are intended to be included in the present invention. Any reference signs in the claims should not be regarded as limiting the claimed rights.

Claims

1. A reverse design method for a prosthetic foot sole with a fixed stiffness based on a deep neural network, including Step 1, parametric script modeling; Step 2, simulation output; Step 3, constructing a data set; Step 4, constructing a deep neural network; Step 5, training a prediction model; Step 6, model testing and evaluation; Step 7, reverse output of the design scheme; It is characterized in that: In Step 1 above, batch structural parametric modeling is performed on the prosthetic foot sole; In Step 2 above, batch simulations of the stiffness of the forefoot and hindfoot are sequentially performed on each model; In Step 3 above, a data set is established according to the simulation results; In Step 4 above, a deep neural network is constructed using the data set; In Step 5 above, a structural parameter prediction model for the prosthetic foot sole is trained; In Step 6 above, the network model is tested and evaluated; In Step 7 above, the structural parameters are reversely output to obtain the design scheme.

2. A reverse design method for a prosthetic foot sole with a fixed stiffness based on a deep neural network according to Claim 1, It is characterized in that: In Step 1, parametric script modeling is performed using the Python language through the Abaqus software, and the structural parameter variables of the prosthetic foot sole are set.

3. A reverse design method for a prosthetic foot sole with a fixed stiffness based on a deep neural network according to Claim 2, It is characterized in that: The structural parameter variables are used to change the forefoot sole thickness distribution, hindfoot sole thickness distribution, sandwich thickness of the midfoot lattice sandwich structure, core angle, core radius, and core number of the model.

4. A reverse design method for a prosthetic foot sole with a fixed stiffness based on a deep neural network according to Claim 1, It is characterized in that: In Step 2, according to the prosthetic foot sole test standard, batch simulations of the stiffness of the forefoot and hindfoot are sequentially performed on each model, and the forefoot and hindfoot displacement-load curves are output from the simulation results to characterize the forefoot and hindfoot stiffness.

5. A reverse design method for a prosthetic foot sole with a fixed stiffness based on a deep neural network according to Claim 1, It is characterized in that: In Step 3, a script is written to extract the foot sole modeling structural parameters and the forefoot and hindfoot stiffness characteristic curves from the batch simulation result data, and a data set is established.

6. A reverse design method for a prosthetic foot sole with a fixed stiffness based on a deep neural network according to Claim 1, It is characterized in that: In Step 4, a deep neural network is constructed using the data set, the stiffness characteristic curve is used as the input data of the neural network model, and the corresponding prosthetic foot sole structural parameters are used as the output of the neural network to construct a deep neural network model composed of an input layer, a fully connected layer, a hidden layer, and an output layer.

7. A reverse design method for a prosthetic foot sole with a fixed stiffness based on a deep neural network according to Claim 1, It is characterized in that: In Step 5, the network model is used to learn the stiffness-structural parameter data, and a structural parameter prediction model for the prosthetic foot sole is trained to predict the structural parameters of the foot sole according to the specified stiffness curve. During the training process, the random gradient descent algorithm is used to adjust and optimize the network parameters to further improve the performance of the network model.

8. A method for inverse design of a prosthetic foot sole with a fixed stiffness based on a deep neural network according to claim 1, characterized in that: in step six, an independent test set is used to test and evaluate the network model and the parameter model.

9. A method for inverse design of a prosthetic foot sole with a fixed stiffness based on a deep neural network according to claim 8, characterized in that: the evaluation indicators of the test evaluation include accuracy, recall, etc., and a validation set is used to evaluate the performance of the network model. According to the generalization ability and performance of the evaluation model, the network parameters are adjusted to achieve the best performance.

10. A method for inverse design of a prosthetic foot sole with a fixed stiffness based on a deep neural network according to claim 1, characterized in that: in step seven, the stiffness curve is used as an input parameter, and the structural parameters of the prosthetic foot sole are inversely solved by using the structural parameter prediction model with a preset stiffness to obtain a design scheme.

Images