Robot structure design optimization simulation method based on finite element model

Abstract

The invention relates to the technical field of finite element simulation, in particular to a robot structure design optimization simulation method based on a finite element model, and the method comprises the steps: obtaining a stress vector of each division unit and a stress vector of each node when a robot is in a static state and performs motion simulation each time; determining a first grid density characteristic value according to a stress stability index similarity condition between each division unit and other division units around the division unit and a stress vector similarity condition under the same motion simulation; determining a second grid density characteristic value by combining the stress stability index of each division unit according to the stress vector difference condition between each division unit and each node during each motion simulation; through the first grid density characteristic value and the second grid density characteristic value, the grid density of each structure area is adjusted, the optimal grid density of each structure area is obtained, the accuracy of a grid division result is effectively improved, and reliable finite element simulation on the robot is facilitated.

Description

Technical Field

[0001] The present invention relates to the technical field of finite element simulation, and particularly to a simulation method for optimizing the robot structure design based on a finite element model. Background Art

[0002] Finite element simulation can analyze in detail the mechanical properties such as stress, strain, and deformation of a robot structure under different working conditions, find out the weak links in the robot structure design, improve the design quality, reduce the R & D cost, and shorten the R & D cycle. Finite element mesh generation is an important link in the process of optimizing the parameters of robot finite element simulation, which directly affects the accuracy of subsequent numerical calculations. Moreover, finite element simulation needs to be performed on each element later, and the mesh generation will also affect the amount of calculation. Therefore, reasonable segmentation can achieve accurate finite element simulation while reducing the amount of calculation and improving the calculation speed.

[0003] The mesh density is a key parameter for finite element mesh generation. When importing a robot model into finite element analysis after it is designed, a fixed mesh density is currently used. The fixed mesh density cannot represent the structural changes that should occur during the dynamic process of the robot, and different structures in the robot require different mesh densities. Therefore, the traditional method for determining the mesh density results in low accuracy of the mesh generation, which is not conducive to reliable finite element simulation of the robot. Summary of the Invention

[0004] In order to solve the technical problem that the traditional method for determining the mesh density of finite element simulation results in low accuracy of the mesh generation, the purpose of the present invention is to provide a simulation method for optimizing the robot structure design based on a finite element model. The specific technical solution adopted is as follows: Obtain each divided element in each structural area of the robot in a static state, perform several motion simulations on each structural area, and obtain the stress vector of each divided element and the stress vector of each node during each motion simulation; According to each stress component in the stress vector of each divided element, analyze the fluctuation of the stress components in the same dimension during multiple motion simulations, and determine the stress stability index of each divided element; According to the similarity of the stress stability index between each divided element and other surrounding divided elements and the similarity of the stress vectors under the same motion simulation, determine the first mesh density eigenvalue of each divided element; According to the difference of the stress vectors between each divided element and each node during each motion simulation, and in combination with the stress stability index of each divided element, determine the second mesh density eigenvalue of each divided element; Adjust the mesh density of each structural area through the first mesh density eigenvalue and the second mesh density eigenvalue to obtain the optimal mesh density of each structural area.

[0005] Further, analyzing the stress component fluctuations in the same dimension during multiple motion simulations according to each stress component in the stress vectors of the respective division units, and determining the stress stability index of each division unit, including: For any division unit, obtaining the stress component sequence in the same dimension during multiple motion simulations, calculating the stability value of the stress component sequence, and obtaining the stability value of each dimension; Selecting the minimum value from all the stability values as the stress stability index of the corresponding division unit.

[0006] Further, determining the first grid density eigenvalue of each division unit according to the similarity of the stress stability indices between each division unit and other division units around it and the similarity of the stress vectors under the same motion simulation, including: Determining any division unit as the unit to be analyzed, and taking other division units in the structural area except the unit to be analyzed as the first comparison units; Calculating the first similarity between the stress vector of the unit to be analyzed and the stress vectors of each first comparison unit under the same motion simulation, and obtaining each of the first similarities corresponding to the unit to be analyzed during each motion simulation; Calculating the second similarity between the stress stability index of the unit to be analyzed and the stress stability indices of each first comparison unit, and obtaining each of the second similarities corresponding to the unit to be analyzed; For each motion simulation, fusing each of the first similarities corresponding to the unit to be analyzed and each of the second similarities corresponding to the unit to be analyzed to obtain the first grid density eigenvalue of the unit to be analyzed.

[0007] Further, obtaining each of the second similarities corresponding to the unit to be analyzed includes: Calculating the absolute value of the difference between the stress stability index of the unit to be analyzed and the stress stability indices of each first comparison unit, performing a negative correlation process on the absolute value of the difference to obtain a negative correlation value, and taking the negative correlation value as the second similarity.

[0008] Further, fusing each of the first similarities corresponding to the unit to be analyzed and each of the second similarities corresponding to the unit to be analyzed to obtain the first grid density eigenvalue of the unit to be analyzed includes: For any motion simulation, calculating the product of the first similarity and the second similarity under the same first comparison unit, performing an accumulative calculation on all the products to obtain an accumulative value, and taking the accumulative value as the stress consistency index between the unit to be analyzed and all the first comparison units during this motion simulation; Calculate the average value of all stress consistency indicators by means of the stress consistency indicator between the unit to be analyzed and all first comparison units during each motion simulation, and obtain the first mesh density eigenvalue of the unit to be analyzed.

[0009] Further, determining the second mesh density eigenvalue of each divided unit according to the stress vector difference between each divided unit and each node during each motion simulation, in combination with the stress stability indicator of each divided unit, includes: Determine any node as the node to be analyzed, and obtain the divided unit connected to the node to be analyzed as the second comparison unit; For each motion simulation, calculate the modulus of the vector difference between the stress vector of the node to be analyzed and the stress vectors of each second comparison unit, and take the mean value of the ratios of each modulus to the stress stability indicator of the corresponding second comparison unit as the density deviation of the node to be analyzed; Obtain the density deviation of each node during each motion simulation, and determine the second mesh density eigenvalue of each divided unit according to the density deviation of each node during each motion simulation.

[0010] Further, determining the second mesh density eigenvalue of each divided unit according to the density deviation of each node during each motion simulation, includes: Calculate the mean value and standard deviation of the density deviation of the node to be analyzed during each motion simulation, perform a negatively correlated normalization process on the product of the mean value and the standard deviation to obtain a normalized value, and take the normalized value as the mesh density sub-eigenvalue of the node to be analyzed; Determine the second mesh density eigenvalue of each divided unit according to the mesh density sub-eigenvalues of each node corresponding to each divided unit.

[0011] Further, determining the second mesh density eigenvalue of each divided unit according to the mesh density sub-eigenvalues of each node corresponding to each divided unit, includes: Determine any divided unit as the unit to be analyzed, and determine all other divided units in the structural area except the unit to be analyzed as the first comparison units; calculate the mean value of the mesh density sub-eigenvalues and the standard deviation of the mesh density sub-eigenvalues of all nodes corresponding to the unit to be analyzed, and the mean value of the mesh density sub-eigenvalues of all nodes corresponding to each first comparison unit; Determine the second mesh density eigenvalue of the unit to be analyzed according to the difference between the mean value of the mesh density sub-eigenvalues between the unit to be analyzed and all first comparison units, the mean value of the mesh density sub-eigenvalues of the unit to be analyzed, and the standard deviation of the mesh density sub-eigenvalues; Among them, the difference in the mean value of the grid density sub-features and the standard deviation of the grid density sub-features are negatively correlated with the second grid density feature value, and the mean value of the grid density sub-features of the unit to be analyzed is positively correlated with the second grid density feature value.

[0012] Further, adjusting the grid density of each structural region by using the first grid density feature value and the second grid density feature value to obtain the optimal grid density of each structural region includes: Determining the grid density parameter of each structural region according to the first grid density feature value and the second grid density feature value of each division unit; setting a first density parameter threshold and a second density parameter threshold, where the first density parameter threshold is less than the second density parameter threshold; For each structural region, if the grid density parameter of the structural region is less than the first density parameter threshold, perform an operation to increase the grid density of the structural region until the grid density parameter of the structural region is within the numerical range corresponding to the first density parameter threshold and the second density parameter threshold, so as to obtain the optimal grid density of the structural region; If the grid density parameter of the structural region is greater than the second density parameter threshold, perform an operation to decrease the grid density of the structural region until the grid density parameter of the structural region is within the numerical range corresponding to the first density parameter threshold and the second density parameter threshold, so as to obtain the optimal grid density of the structural region.

[0013] Further, determining the grid density parameter of each structural region according to the first grid density feature value and the second grid density feature value of each division unit includes: For each structural region, calculate the product of the first grid density feature value and the second grid density feature value of each division unit as the grid density parameter of the corresponding division unit; Take the mean value of the grid density parameters of all division units within the same structural region as the grid density parameter of the corresponding structural region.

[0014] Compared with the prior art, the present invention has the following beneficial effects: The present invention provides a simulation method for optimizing the robot structure design based on a finite element model. By analyzing the stress changes of the statically divided elements during multiple dynamic simulations of the robot, that is, determining the stress stability index of the divided elements, the influence of the irreproducibility of the simulation results on the finite element mesh division can be prevented. By determining the first mesh density eigenvalue of each divided element according to the similarity of the stress stability indices between each divided element and other surrounding divided elements and the similarity of the stress vectors under the same motion simulation, the excessive mesh density and infinite subdivision of the elements can be prevented. By analyzing the relationship between the element stress and the node stress to obtain the mesh density characteristics represented by the nodes, and by combining the element stress and the node stress to jointly reflect the mesh density characteristics, the comprehensiveness of the mesh density characteristics analysis is enhanced. By adaptively determining the mesh density for each structural area corresponding to the robot, the accuracy of the mesh division result is effectively improved, which is beneficial to reliable finite element simulation of the robot. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In order to more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0016] Figure 1 It is a flowchart for implementing a simulation method for optimizing the robot structure design based on a finite element model proposed in an embodiment of the present invention; Figure 2 It is a flowchart for implementing step S3 in an embodiment of the present invention; Figure 3 It is a flowchart for implementing step S4 in an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0017] In order to further elaborate on the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the following will, in conjunction with the accompanying drawings and preferred embodiments, describe in detail the specific implementation manners, structures, features, and effects of the technical solutions proposed according to the present invention. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. In addition, the specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0018] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the technical field to which the present invention belongs.

[0019] The application scenarios targeted by the present invention can be: On the one hand, there are differences in the division density of different structures in the robot structure. For example, in the arm structure and finger structure of the robot, the finger structure is more flexible than the arm structure and requires a more refined finite element division. On the other hand, the existing finite grid division does not consider the structural changes of the robot during motion, such as the lifting of the arm and the stretching of the fingers. Affected by these two factors, the accuracy of the existing finite grid division results is low.

[0020] To improve the accuracy of the finite grid division results, an embodiment of the present invention provides a robot structure design optimization simulation method based on a finite element model, as Figure 1 shown, including the following steps: S1. Obtain each division unit in each structural area of the robot in a static state, perform several motion simulations on each structural area, and obtain the stress vector of each division unit and the stress vector of each node during each motion simulation.

[0021] The above step S1 can be implemented through steps S11 to S12 (not shown in the figure): S11. Obtain each division unit in each structural area of the robot in a static state.

[0022] First, import the robot structure to be simulated into the finite element analysis software, divide the robot according to the structural information of the robot to obtain each structural area, such as the head, body, etc. After completing the grouping of the robot structure information, obtain the relevant material data of the robot and import the material data so that the robot structure has the mechanical properties it should have.

[0023] Among them, a structural area can be composed of multiple structural components or a part of a component. For example, the hand can be divided into a structural area, or the hand can be divided into multiple finger areas and a palm area. The specific division is determined according to the actual robot simulation requirements. After dividing the entire robot into areas, each structural area is regarded as a whole for analysis. For example, when operating through the robot's palm, the hand area needs to be refined, and the robot's motion balance can directly regard the hand as a whole.

[0024] Then, without applying any external force in the finite element software, obtain the finite element mesh division result of the robot in a static state, that is, each division unit in each structural area. The division unit is a tetrahedral structure and is formed by connecting several nodes. Specifically, the traditional division of the robot's finite element mesh relies on the structural information of the robot in the static process to achieve unit division, and the mesh divided in the static state mainly uses the material data and specific structural data of the robot to achieve division.

[0025] It is worth noting that the above mesh division results may not conform to the actual operation of the robot. For example, static simulation leads to a denser division of the body, but in fact the body of the robot can be regarded as a whole, and the mesh division is actually larger and has a smaller mesh density. Therefore, the mesh division should be combined with the actual motion status of the robot structure to determine the most appropriate mesh density for each structural area.

[0026] At this point, the finite element mesh division of each structural area of ​​the robot in a stationary state is completed, and each division unit in each structural area is obtained.

[0027] S12, performing motion simulation for each structural region for several times, and obtaining the stress vector of each divided unit and the stress vector of each node during each motion simulation.

[0028] After completing the initial mesh division of the structural area, the finite element software is used to perform motion simulation according to the motion characteristics of each structural area of ​​the robot, such as robot finger bending, body translation, arm extension, etc., and multiple simulations are performed. During the simulation process, the force condition of each division unit in the structural area during the motion process is obtained. The force condition of each division unit and each node can be represented by a vector, namely a stress vector. The elements in the stress vector represent stress components in different directions.

[0029] The number of motion simulations can be set by the implementer according to the actual situation without specific limitation, such as 10 times; there are 6 stress components in three dimensions, 3 directional stresses and 3 shear stresses.

[0030] At this point, by performing finite element simulation on the units in the divided structural area, the force conditions of each divided unit and each node can be determined, which is helpful for subsequent robot mechanical analysis.

[0031] S2, according to each stress component in the stress vector of each division unit, analyze the stress component fluctuation in the same dimension during multiple motion simulations, and determine the stress stability index of each division unit.

[0032] After multiple finite element simulations, the final simulation analysis results should be stable. For example, after multiple simulations of the robot's finger bending motion, the simulation results should be reproducible. However, when the mesh division is inaccurate, the divided units cannot accurately describe the mechanical behavior of the robot's structural motion process, the stress calculation is unstable, and the unstable unit stress causes large differences in multiple simulation results. On the contrary, if the finite element mesh division results are appropriate, the mechanical properties of the robot's motion process can be accurately captured in multiple simulations, and the unit stress results of different motion simulations are relatively stable.

[0033] For any partition unit, the above step S2 can be implemented through steps S21 to S22 (not shown in the figure): S21. Obtain the stress component sequence in the same dimension during multiple motion simulations, calculate the stable value of the stress component sequence, and obtain the stable value of each dimension.

[0034] In this embodiment, each partition unit has its corresponding stress component sequence in different dimensions, and the stability of the stress component sequence can characterize the stability of the simulation results during multiple motion simulations.

[0035] As an exemplary implementation manner, it includes: Obtain the stress components in the i-th dimension during multiple motion simulations to form a stress component sequence; calculate the standard deviation of the stress component sequence; perform negative-correlation normalization processing on the standard deviation of the stress component sequence using exp(-) to obtain a normalized value as the stable value of the i-th dimension. Wherein, i is a positive integer representing the serial number of the dimension, the value range of the normalized value is between 0 and 1, and the data property of the stress component is a scalar.

[0036] Another exemplary implementation manner includes: Calculate the average value of the stress component sequence in the i-th dimension; perform difference calculation on each stress component in the stress component sequence with the average value to obtain each difference value; perform reciprocal processing on the non-zero difference values, and take the average value of all the obtained reciprocals as the stable value of the i-th dimension.

[0037] S22. Select the minimum value from all the stable values as the stress stability index of the corresponding partition unit.

[0038] In this embodiment, based on the calculation process of the stable value of the i-th dimension, the stable value of each dimension is obtained, and then the minimum stable value is used as the stress stability index of the corresponding partition unit.

[0039] As an example, the calculation formula of the stress stability index of the partition unit can be: ; where a represents the stress stability index of the partition unit, min represents the minimum value function, exp represents the exponential function with the natural constant e as the base, exp(-) is used to implement negative-correlation normalization processing, represents the standard deviation function, represents the stress component sequence of the i-th dimension in the stress vector of the partition unit during multiple motion simulations, I represents the number of dimensions, and can be set to 6.

[0040] When the minimum stable values are all relatively large, it indicates that each stress component has high stress stability under multiple motion simulations. Furthermore, it shows that the stress calculation stability of the divided elements is good, the differences between the simulation results of different motion simulations are small, and the numerical accuracy of the initial mesh density of the divided elements may be high.

[0041] Thus, the stress stability indicators of each divided element in each structural region are obtained.

[0042] S3. According to the similarity of the stress stability indicators between each divided element and other surrounding divided elements and the similarity of the stress vectors under the same motion simulation, determine the first mesh density eigenvalue of each divided element.

[0043] Although a too-high mesh density has relatively accurate simulation results, the increase in the number of divided elements makes it take a lot of time to obtain the final robot structure simulation results. Moreover, for some non-stress concentration regions, increasing the mesh density cannot improve the simulation of the robot structure, and a slightly smaller mesh density can also obtain a relatively accurate robot structure simulation result. Among them, the non-stress concentration region can be the robot body part.

[0044] In this embodiment, for any divided element, it is necessary to compare it with the surrounding divided elements to determine whether there is a situation of too large a mesh density for this divided element, that is, to analyze the mesh density characteristics of the divided element, denoted as the first mesh density eigenvalue. Specifically, during the process of the finite element software simulating the robot structure, the stress changes continuously in space. If the mesh is divided too finely, the sizes and shapes of adjacent elements will be very small and similar, then the loads and boundary conditions received by the divided elements will not vary much within a local range. For the reason that the sizes of adjacent elements are similar, the finite element method itself is an approximate method. When the mesh density reaches a certain level, the stress calculation results of adjacent elements tend to be the same due to local similarity.

[0045] The above step S3 can be implemented through Figure 2 the steps S31 to S34 shown as follows: S31. Determine any divided element as the element to be analyzed, and regard the other divided elements in the structural region except the element to be analyzed as the first comparison elements.

[0046] For the convenience of description, select any divided element in any structural region, denoted as the element to be analyzed, and regard all the divided elements in the structural region except the element to be analyzed as the divided elements adjacent to the selected element, denoted as the first comparison elements of the element to be analyzed.

[0047] S32. Calculate the first similarity between the stress vector of the unit to be analyzed and the stress vectors of each first comparison unit under the same motion simulation, and obtain the respective first similarities corresponding to the unit to be analyzed for each motion simulation.

[0048] In this embodiment, the higher the stress similarity between adjacent divided units, the denser the mesh density of the current structural area, that is, the mesh density is too large. On the contrary, the worse the stress similarity between adjacent divided units, the greater the stress variation between the divided units, and the mesh density of the current structural area is too sparse, that is, the mesh density is too small.

[0049] Specifically, for the stress vector of the unit to be analyzed and the stress vectors of each first comparison unit under the same motion simulation, calculate the cosine similarity between the stress vector of the unit to be analyzed and the stress vector of the first comparison unit as the first similarity. The calculation process of the cosine similarity is prior art and not within the protection scope of the present invention, so it will not be elaborated in detail here.

[0050] S33. Calculate the second similarity between the stress stability index of the unit to be analyzed and the stress stability indices of each first comparison unit, and obtain the respective second similarities corresponding to the unit to be analyzed.

[0051] In this embodiment, while the stress similarity between adjacent divided units is relatively high, it also satisfies the condition that the stress stability between the unit to be analyzed and the first comparison unit is similar, which can indicate that the stress consistency between the unit to be analyzed and other divided units in the structural area is stronger. Therefore, it is necessary to determine the second similarity.

[0052] As an exemplary implementation manner, it includes: Calculate the absolute value of the difference between the stress stability index of the unit to be analyzed and the stress stability indices of each first comparison unit, perform a negative correlation process on the absolute value of the difference to obtain a negative correlation value, and use the negative correlation value as the second similarity.

[0053] Specifically, perform a reciprocal process on the absolute value of the difference, and use the obtained reciprocal value as the second similarity. In case of extreme situations where the difference between the stress stability index of the unit to be analyzed and the stress stability index of the first comparison unit is zero, no second similarity analysis is performed on this first comparison unit, or a non-zero constant, such as 0.01, is added to the denominator position of the fraction.

[0054] S34. For each motion simulation, fuse the respective first similarities corresponding to the unit to be analyzed and the respective second similarities corresponding to the unit to be analyzed to obtain the first mesh density eigenvalue of the unit to be analyzed.

[0055] In this embodiment, after obtaining the first similarity and the second similarity, the network density characteristics are quantified from two perspectives of stress similarity and stress stability similarity.

[0056] If the network density feature is too large, it indicates that the grid density is too large and the size of the divided element is too small; if the network density feature is too small, it indicates that the grid density is too small and the size of the divided element is too large.

[0057] The above step S34 can be implemented through steps S341 to S342 (not shown in the figure): S341. For any motion simulation, calculate the product of the first similarity and the second similarity under the same first comparison unit, accumulate all the products to obtain an accumulated value, and use the accumulated value as the stress consistency index between the unit to be analyzed and all the first comparison units during this motion simulation.

[0058] In this embodiment, the dimensions of the first similarity and the second similarity are the same, and both are data between 0 and 1. When performing the multiplication calculation, calculate the product of the first similarity and the second similarity of the same first comparison unit.

[0059] As an example, the calculation formula for the stress consistency index between the unit to be analyzed and all the first comparison units during any motion simulation can be: ; where represents the stress consistency index between the unit to be analyzed and all the first comparison units, K represents the number of the first comparison units, k represents the serial number of the first comparison unit, represents the stress stability index of the unit to be analyzed, represents the stress stability index of the k-th first comparison unit of the unit to be analyzed, represents the absolute value function, cos represents the cosine similarity function, represents the stress vector of the unit to be analyzed, represents the stress vector of the k-th first comparison unit of the unit to be analyzed.

[0060] represents the second similarity, which is used to characterize the stress stability similarity between the unit to be analyzed and the k-th first comparison unit. If is equal to 0, add a non-zero constant, such as 0.01, to the denominator position of the fraction; represents the first similarity, which is used to characterize the stress vector similarity between the unit to be analyzed and the k-th first comparison unit; by determining the stress stability similarity and the stress vector similarity between the unit to be analyzed and each first comparison unit, the stress consistency between the unit to be analyzed and all the divided units in its same structural region can be quantified; when the stress consistency index is too large, it indicates that the grid density of the corresponding structural region during the same motion simulation is too large, and the grid density needs to be adjusted to be smaller.

[0061] S342. Calculate the average value of all stress consistency indices by means of the stress consistency index between the unit to be analyzed and all the first comparison units during each motion simulation, so as to obtain the first mesh density eigenvalue of the unit to be analyzed.

[0062] In this embodiment, the implementation steps of the above-mentioned step S341 are adopted to obtain the stress consistency index corresponding to the unit to be analyzed during each motion simulation, obtain the stress consistency index sequence, and calculate the average value of the stress consistency index sequence as the first mesh density eigenvalue. If the unit to be analyzed has a high stress consistency during each motion simulation, it indicates that the similarity between the unit to be analyzed and other divided units in the structural area during the finite element simulation of the robot is too large, and the initial mesh density is too large, and the mesh density needs to be reduced.

[0063] Referring to the calculation process of the first mesh density eigenvalue of the unit to be analyzed, the first mesh density eigenvalue of each divided unit in each structural area can be obtained, and the determination steps of the first mesh density eigenvalue of each divided unit will not be repeated here.

[0064] So far, this embodiment has obtained the first mesh density eigenvalue of each divided unit in each structural area.

[0065] S4. Determine the second mesh density eigenvalue of each divided unit according to the stress vector difference between each divided unit and each node during each motion simulation, in combination with the stress stability index of each divided unit.

[0066] The above-mentioned step S3 considers that too large a mesh density leads to low simulation efficiency in the structural design of the robot. However, too small a mesh density will lead to low simulation accuracy in the structural design of the robot. Specifically, the stability of the divided unit depends on the position of the divided unit in the robot structure. When the divided unit appears in a special position, such as in the body structure area, it is not sensitive to the stress change during the simulation process. Although the accuracy of the mesh density division for the body structure area is low, it will not affect its high stress stability; when the divided unit appears in an area with a large stress gradient, such as the sharp corner of the robot joint or the bolt connection, if the mesh density is too small, the stress of the divided unit may underestimate the local peak stress, while the node stress is closer to the true value through extrapolation, resulting in a significant difference between the unit stress and the node stress.

[0067] In this embodiment, the difference between the unit stress and the node stress is large, indicating that there are drastic changes in the internal stress distribution during the simulation motion of the robot. For example, when the robot finger joint bends, stress concentration occurs in the contact area. At this time, if the mesh density is too small, the unit stress shows a lower average value, while the node stress may capture the local peak through extrapolation, and the mesh density needs to be increased.

[0068] The above-mentioned step S4 can be throughFigure 3 Steps S41 to S43 shown are implemented as follows: S41, determine any node as the node to be analyzed, and obtain the partitioning unit connected to the node to be analyzed as the second comparison unit.

[0069] In this embodiment, taking any node in the robot structure as an example for analysis, any node is used as the node to be analyzed, and the partitioning unit connected to the node to be analyzed is used as the comparison unit of the node to be analyzed. To distinguish it from the first comparison unit, it is denoted as the second comparison unit.

[0070] S42, for each motion simulation, calculate the modulus of the vector difference between the stress vector of the node to be analyzed and the stress vectors of each second comparison unit, and take the mean value of the ratios of each modulus to the stress stability index of the corresponding second comparison unit as the density deviation of the node to be analyzed.

[0071] In this embodiment, the density deviation of the finite element partitioning nodes can reflect whether the mesh density needs to be increased. When the density deviation is too large, it indicates that the mesh partitioning of the robot structure is too sparse, that is, the mesh density is too small, and the mesh density needs to be adjusted upwards. When the difference between the element stress and the node stress is large and the stability of the element stress is small at the same time, it indicates that the accuracy of the current finite element mesh partitioning is low and the density deviation is large.

[0072] As an example, the calculation formula for the density deviation of the node to be analyzed can be: ; where represents the density deviation of the node to be analyzed, E represents the mean function, represents the stress vector of the m-th second comparison unit of the node to be analyzed, represents the stress vector of the node to be analyzed, represents the modulus of the vector, represents the stress stability index of the m-th second comparison unit of the node to be analyzed, M represents the number of second comparison units of the node to be analyzed, and m represents the serial number of the second comparison unit.

[0073] Generally, there is no possibility of being zero. If an extreme case occurs, then add a non-zero constant, such as 0.01, at the denominator position of the fraction .

[0074] S43, obtain the density deviation of each node during each motion simulation, and determine the second mesh density eigenvalue of each partitioning unit according to the density deviation of each node during each motion simulation.

[0075] In this embodiment, after obtaining the density deviation of the node to be analyzed during any motion simulation, referring to the implementation process of steps S41 to S42 above, the density deviation of each node during each motion simulation is obtained. Subsequently, based on the density deviation of each node during each motion simulation, the second grid density eigenvalue of each divided unit is quantified.

[0076] The above step S43 can be implemented through steps S431 to S432 (not shown in the figure): S431, calculate the mean and standard deviation of the density deviation of the node to be analyzed during each motion simulation, perform a negative correlation normalization process on the product of the mean and the standard deviation to obtain a normalized value, and use the normalized value as the grid density sub-feature of the node to be analyzed.

[0077] In this embodiment, if the same node has a high density deviation during each motion simulation and the change stability of the node's density deviation is poor, it indicates that during these motion simulation processes, minor parameter changes may lead to large errors in the robot structure analysis results, resulting in the phenomenon that the node simulation results do not converge. Thus, the obtained simulation results have no reference value and seriously interfere with the subsequent simulation optimization of the robot. The reason for the high density deviation of the node under different motion simulations and the poor density stability is that the initial grid parameters are too small and the divided units are too large, resulting in complex internal stresses in a divided unit. Considering a unit with complex stresses as a whole for structural optimization may cause large errors in the simulation process. In this case, the grid density needs to be increased.

[0078] However, in order to correspond to the logical meaning of the first grid density eigenvalue, that is, if the network density eigenvalue is too large, it means that the grid density is too large and the divided unit size is too small, and the grid density needs to be adjusted downwards. After calculating the mean and standard deviation of the density deviation of the node to be analyzed, perform a negative correlation process on the product of the mean and the standard deviation to make the physical meaning of the subsequently determined second grid density eigenvalue consistent with that of the first grid density eigenvalue.

[0079] As an example, the calculation formula for the grid density sub-feature of the node to be analyzed can be: ; in the formula, represents the grid density sub-feature of the node to be analyzed, exp represents the exponential function with the natural constant as the base, exp(-) is used to implement the negative correlation normalization process, represents the mean of the density deviation of the node to be analyzed during all motion simulations, represents the standard deviation of the density deviation of the node to be analyzed during all motion simulations, represents the density deviation of the node to be analyzed during the l-th motion simulation, L represents the number of motion simulations, and l represents the serial number of different motion simulations.

[0080] S432. Determine the second grid density eigenvalue of each division unit according to the grid density sub-features of each node corresponding to each division unit.

[0081] In this embodiment, under normal circumstances, the grid density of the robot body is small. However, if the body is damaged during a certain simulation process, such as abnormal defects like tiny holes or cracks, due to the relatively large division units, the impact of the defects on the overall unit is limited, resulting in an insignificant stress change in the division unit during motion simulation. But because only the stress change of a specific node is considered, the stress change of this node is significant at this time. Therefore, it is necessary to combine the grid density sub-features of each node to analyze the grid density feature of the division unit.

[0082] First, determine any division unit as the unit to be analyzed, and determine all other division units in the structural area except the unit to be analyzed as the first comparison units; calculate the mean value and standard deviation of the grid density sub-features of all nodes corresponding to the unit to be analyzed, and the mean value of the grid density sub-features of all nodes corresponding to each first comparison unit.

[0083] For the unit to be analyzed, if the node stress consistency of the unit to be analyzed is similar to the node stress consistency of all units, it means that the node stress of the unit to be analyzed and the node stress of adjacent other units are consistent during the simulation process. The unit to be analyzed and its surrounding units can be appropriately merged to avoid the division unit being too small due to excessive grid density; on the contrary, if the difference between the node stress consistency of the unit to be analyzed and the node stress consistency of all units is large, it means that the grid density of the unit to be analyzed is small and the unit needs to be further subdivided.

[0084] Secondly, determine the second grid density eigenvalue of the unit to be analyzed according to the difference between the mean value of the grid density sub-features between the unit to be analyzed and all first comparison units, the mean value of the grid density sub-features of the unit to be analyzed, and the standard deviation of the grid density sub-features.

[0085] Among them, the difference in the mean value of the grid density sub-features, the standard deviation of the grid density sub-features, and the second grid density eigenvalue show a negative correlation, and the mean value of the grid density sub-features of the unit to be analyzed and the second grid density eigenvalue show a positive correlation.

[0086] As an example, the calculation formula for the second grid density eigenvalue of the unit to be analyzed can be: ; where p represents the second grid density eigenvalue of the unit to be analyzed, norm represents the linear normalization function, represents the mean value of the grid density sub-features of the unit to be analyzed, Represents the mean value of the grid density sub-features among all the first comparison units, Represents the difference between the grid density sub-feature mean value between the unit to be analyzed and all the first comparison units, Represents the standard deviation of the grid density sub-features of the unit to be analyzed.

[0087] It should be noted that the second grid density eigenvalue p of the unit to be analyzed is obtained from the node features, which can be used to prevent the grid density from being excessively large. The larger the second grid density eigenvalue, the more serious the phenomenon of excessive grid density in the finite element mesh division. Referring to the calculation process of the second grid density eigenvalue of the unit to be analyzed, the second grid density eigenvalue of each divided unit can be obtained.

[0088] So far, this embodiment has obtained the second grid density eigenvalues of each divided unit in each structural region.

[0089] S5. Adjust the grid density of each structural region through the first grid density eigenvalue and the second grid density eigenvalue to obtain the optimal grid density of each structural region.

[0090] In this embodiment, both the first grid density eigenvalue and the second grid density eigenvalue are used to characterize the grid density features of the divided units, which can be used to measure the rationality of the grid density setting of the divided units. After obtaining the rationality of the grid density setting, adaptively adjust the grid density of each structural region based on the setting rationality, so as to obtain the optimal grid density of each structural region.

[0091] The above step S5 can be implemented through steps S51 to S52 (not shown in the figure): S51. Determine the grid density parameters of each structural region according to the first grid density eigenvalue and the second grid density eigenvalue of each divided unit; set the first density parameter threshold and the second density parameter threshold.

[0092] Specifically, for each structural region, calculate the product of the first grid density eigenvalue and the second grid density eigenvalue of each divided unit as the grid density parameter of the corresponding divided unit; take the mean value of the grid density parameters of all the divided units within the same structural region as the grid density parameter of the corresponding structural region.

[0093] In this embodiment, the first density parameter threshold is less than the second density parameter threshold. The two density parameter thresholds are used to represent the standard of dividing the grid density. The first density parameter threshold can take an empirical value of 0.45, while the second density parameter threshold can take an empirical value of 0.55. The implementer can set the sizes of the first density parameter threshold and the second density parameter threshold according to the specific grid division standard requirements, without specific limitations.

[0094] S52. For each structural region, compare the mesh density parameter of the structural region with the first density parameter threshold and the second density parameter threshold, and adjust the mesh density of the structural region according to the comparison result.

[0095] If the mesh density parameter of the structural region is less than the first density parameter threshold, perform an operation to increase the mesh density of the structural region until the mesh density parameter of the structural region is within the numerical range corresponding to the first density parameter threshold and the second density parameter threshold, and obtain the optimal mesh density of the structural region.

[0096] If the mesh density parameter of the structural region is greater than the second density parameter threshold, perform an operation to decrease the mesh density of the structural region until the mesh density parameter of the structural region is within the numerical range corresponding to the first density parameter threshold and the second density parameter threshold, and obtain the optimal mesh density of the structural region.

[0097] It should be noted that when adjusting the mesh density of the structural region, the finite element software has different mesh density level options. When performing an operation to increase the mesh density of the structural region, increase one mesh density level; when performing an operation to decrease the mesh density of the structural region, decrease one mesh density level.

[0098] After increasing or decreasing the mesh density level, determine the mesh density parameter of the structural region after the mesh density update again, and compare it with the density parameter threshold. Continuously repeat the above operations of determining the mesh density parameter and adjusting the mesh density until the mesh density parameter of the structural region is within the numerical range corresponding to the first density parameter threshold and the second density parameter threshold, and take the adjusted mesh density at this time as the optimal mesh density of the corresponding structural region.

[0099] So far, this embodiment has completed the optimization simulation of the robot structure design using the improved finite element model.

[0100] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that: they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements for some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments of the present invention, and should all be included in the protection scope of the present invention.

Claims

1. A robot structure design optimization simulation method based on a finite element model, characterized in that, The steps include: Obtain each division unit in each structural area of the robot in a static state, conduct a number of motion simulations on each structural area, and obtain the stress vectors of each division unit and the stress vectors of each node during each motion simulation; According to each stress component in the stress vectors of each division unit, analyze the fluctuation of the stress components in the same dimension during multiple motion simulations, and determine the stress stability index of each division unit; According to the similarity of the stress stability indexes between each division unit and other surrounding division units and the similarity of the stress vectors under the same motion simulation, determine the first grid density eigenvalue of each division unit; According to the difference of the stress vectors between each division unit and each node during each motion simulation, and in combination with the stress stability index of each division unit, determine the second grid density eigenvalue of each division unit; Adjust the grid density of each structural area through the first grid density eigenvalue and the second grid density eigenvalue to obtain the optimal grid density of each structural area.

2. The robot structure design optimization simulation method based on a finite element model according to claim 1, wherein The step of according to each stress component in the stress vectors of each division unit, analyzing the fluctuation of the stress components in the same dimension during multiple motion simulations, and determining the stress stability index of each division unit includes: For any division unit, obtain the stress component sequence in the same dimension during multiple motion simulations, calculate the stability value of the stress component sequence, and obtain the stability value of each dimension; Select the minimum value from all the stability values as the stress stability index of the corresponding division unit.

3. The optimization simulation method for robot structure design based on a finite element model according to claim 1, characterized in that The step of according to the similarity of the stress stability indexes between each division unit and other surrounding division units and the similarity of the stress vectors under the same motion simulation, determining the first grid density eigenvalue of each division unit includes: Determine any division unit as the unit to be analyzed, and regard other division units in the structural area except the unit to be analyzed as the first comparison units; Calculate the first similarity between the stress vector of the unit to be analyzed and the stress vectors of each first comparison unit under the same motion simulation, and obtain each of the first similarities corresponding to the unit to be analyzed during each motion simulation; Calculate the second similarity between the stress stability index of the unit to be analyzed and the stress stability indexes of each first comparison unit, and obtain each of the second similarities corresponding to the unit to be analyzed; For each motion simulation, fuse each of the first similarities corresponding to the unit to be analyzed and each of the second similarities corresponding to the unit to be analyzed to obtain the first grid density eigenvalue of the unit to be analyzed.

4. The method for optimizing the simulation of the robot structure design based on the finite element model according to claim 3, wherein The step of obtaining each of the second similarities corresponding to the unit to be analyzed includes: Calculate the absolute value of the difference between the stress stability index of the unit to be analyzed and the stress stability indexes of each first comparison unit, perform a negative correlation process on the absolute value of the difference to obtain a negative correlation value, and use the negative correlation value as the second similarity.

5. The optimization simulation method for robot structure design based on a finite element model according to claim 4, characterized in that The step of fusing each of the first similarities corresponding to the unit to be analyzed and each of the second similarities corresponding to the unit to be analyzed to obtain the first grid density eigenvalue of the unit to be analyzed includes: For any motion simulation, calculate the product of the first similarity and the second similarity under the same first comparison unit, accumulate all the products to obtain an accumulated value, and use the accumulated value as the stress consistency index between the unit to be analyzed and all first comparison units during this motion simulation; Calculate the average value of all stress consistency indexes through the stress consistency index between the unit to be analyzed and all first comparison units during each motion simulation, and obtain the first grid density eigenvalue of the unit to be analyzed.

6. The method for optimizing the simulation of the robot structure design based on the finite element model according to claim 1, wherein Determine the second grid density eigenvalue of each divided unit according to the stress vector difference between each divided unit and each node during each motion simulation, in combination with the stress stability index of each divided unit, including: Determine any node as the node to be analyzed, and obtain the divided unit connected to the node to be analyzed as the second comparison unit; For each motion simulation, calculate the modulus of the vector difference between the stress vector of the node to be analyzed and the stress vectors of each second comparison unit, and use the average value of the ratios of each modulus to the stress stability index of the corresponding second comparison unit as the density deviation of the node to be analyzed; Obtain the density deviation of each node during each motion simulation, and determine the second grid density eigenvalue of each divided unit according to the density deviation of each node during each motion simulation.

7. The simulation method for optimizing the robot structure design based on the finite element model according to claim 6, characterized in that, Determine the second grid density eigenvalue of each divided unit according to the density deviation of each node during each motion simulation, including: Calculate the average value and standard deviation of the density deviation of the node to be analyzed during each motion simulation, perform a negative correlation normalization process on the product of the average value and the standard deviation to obtain a normalized value, and use the normalized value as the grid density sub-eigenvalue of the node to be analyzed; Determine the second grid density eigenvalue of each divided unit according to the grid density sub-eigenvalue of each node corresponding to each divided unit.

8. The simulation method for optimizing the robot structure design based on a finite element model according to claim 7, wherein, Determine the second grid density eigenvalue of each divided unit according to the grid density sub-eigenvalue of each node corresponding to each divided unit, including: Determine any divided unit as the unit to be analyzed, and determine all other divided units in the structural area except the unit to be analyzed as the first comparison unit; calculate the average value of the grid density sub-eigenvalues and the standard deviation of the grid density sub-eigenvalues of all nodes corresponding to the unit to be analyzed, and the average value of the grid density sub-eigenvalues of all nodes corresponding to each first comparison unit; Determine the second grid density eigenvalue of the unit to be analyzed according to the difference between the average value of the grid density sub-eigenvalues between the unit to be analyzed and all first comparison units, the average value of the grid density sub-eigenvalues of the unit to be analyzed, and the standard deviation of the grid density sub-eigenvalues; Among them, the difference between the average value of the grid density sub-eigenvalues, the standard deviation of the grid density sub-eigenvalues and the second grid density eigenvalue show a negative correlation relationship, and the average value of the grid density sub-eigenvalues of the unit to be analyzed and the second grid density eigenvalue show a positive correlation relationship.

9. The simulation method for optimizing the robot structure design based on the finite element model according to claim 1, wherein Adjust the grid density of each structural area through the first grid density eigenvalue and the second grid density eigenvalue to obtain the optimal grid density of each structural area, including: Determine the grid density parameter of each structural region according to the first grid density eigenvalue and the second grid density eigenvalue of each division unit; set a first density parameter threshold and a second density parameter threshold, where the first density parameter threshold is less than the second density parameter threshold; For each structural region, if the grid density parameter of the structural region is less than the first density parameter threshold, perform an operation to increase the grid density of the structural region until the grid density parameter of the structural region is within the numerical range corresponding to the first density parameter threshold and the second density parameter threshold, to obtain the optimal grid density of the structural region; If the grid density parameter of the structural region is greater than the second density parameter threshold, perform an operation to decrease the grid density of the structural region until the grid density parameter of the structural region is within the numerical range corresponding to the first density parameter threshold and the second density parameter threshold, to obtain the optimal grid density of the structural region.

10. The method for optimizing and simulating the robot structure design based on the finite element model according to claim 9, wherein The determining the grid density parameter of each structural region according to the first grid density eigenvalue and the second grid density eigenvalue of each division unit includes: For each structural region, calculate the product of the first grid density eigenvalue and the second grid density eigenvalue of each division unit as the grid density parameter of the corresponding division unit; Take the mean value of the grid density parameters of all division units within the same structural region as the grid density parameter of the corresponding structural region.

Images