A multi-objective optimization design method and system for permanent magnet motor

Abstract

The present invention discloses a method and system for multi-objective optimization design of permanent magnet motors, and belongs to the field of permanent magnet motors. The present invention adopts the Taguchi experimental method under the condition of a given number of optimization parameters and the number of levels of each parameter, which can greatly reduce the number of simulations (experiments) required, that is, there is no need to adopt the "one scheme one simulation" analysis mode, which effectively shortens the time required for the motor optimization process; on this basis, the RBF network is used to establish a nonlinear relationship between the motor structure parameters and its performance indicators, and the discrete solution domain is converted into a continuous solution domain, which can ensure the accuracy of subsequent optimization. The present invention uses the fuzzy optimal and worst method to obtain the weight coefficients of each performance indicator, making it more reasonable to convert the multi-objective problem into a single-objective problem for analysis and optimization.

Description

Technical Field

[0001] The present invention belongs to the field of permanent magnet motors, and more specifically, relates to a multi-objective optimization design method and system for permanent magnet motors. Background Art

[0002] With the improvement of permanent magnet materials and the rapid development of power electronics and control technology, permanent magnet motors have been widely used in electric propulsion, servo systems, wind power generation and other fields, showing their own characteristics of high power factor, high efficiency and high power density. As the application occasions of permanent magnet motors increase, their motor structure becomes more and more complex, and the requirements for their performance indicators become higher and higher. How to accurately and quickly optimize the optimal design that meets the corresponding performance indicator requirements during the design of the motor body is the direction that motor designers have been working hard for.

[0003] One of the important components of motor optimization design is the optimization of the motor body structural parameters, such as the inner and outer diameters of the motor stator and rotor, slot size, pole slot matching, permanent magnet shape and arrangement, etc. This optimization design process corresponds to a typical multi-objective optimization problem. The strong nonlinearity of the electromagnetic field inside the motor brings great difficulties to modeling and analysis. Usually, motor designers rely on commercial finite element software to optimize the design of motors. Under the premise of good segmentation, finite element analysis can obtain sufficient accuracy. However, finite element analysis is time-consuming, especially when analyzing schemes for multiple structural parameter matching. If the "one scheme one simulation" analysis mode is adopted, the optimization process of the motor will take a very long time. At the same time, the above process is usually carried out in a discrete search space, and it is impossible to find the structural parameter matching that best meets the design requirements in a continuous search space.

[0004] In order to optimize the structural parameters of the motor in a continuous search space, commercial finite element software usually includes some heuristic algorithm plug-ins to assist in optimization, such as Maxwell's genetic algorithm plug-in and ANSYS Workbench's optislang module. These optimizers have not gotten rid of the "one solution, one simulation" model. The multiple calculations of the evaluation function during the algorithm operation are very time-consuming and the optimization cycle is too long. More importantly, the optimizer usually uses the method of weighting the multi-objective optimization problem into a single-objective optimization problem to set the evaluation function, but there is no basis or method to follow for selecting the weights of each optimization goal in the weighting process.

[0005] In summary, it is necessary to propose a multi-objective optimization design method for permanent magnet motors that can significantly shorten the optimization cycle and solve the problem of weight setting in the multi-objective optimization process. Summary of the invention

[0006] In view of the above defects or improvement needs of the prior art, the present invention provides a permanent magnet motor multi-objective optimization design method and system, which aims to provide a permanent magnet motor multi-objective optimization design method that can significantly shorten the optimization cycle and solve the problem of weight setting in the multi-objective optimization process.

[0007] To achieve the above object, the present invention provides a multi-objective optimization design method for a permanent magnet motor, comprising:

[0008] S1. Complete the preliminary design of the motor, determine the target performance to be optimized, the motor structural parameters involved in the optimization process, and their range of variation and number of levels;

[0009] S2. Use the Taguchi method combined with finite element simulation to obtain the target performance under different parameter combinations, and calculate the signal-to-noise ratio corresponding to each target performance according to the optimization requirements of each target performance;

[0010] S3. Taking each set of parameters as input data and the signal-to-noise ratio of the corresponding target performance as output data, the nonlinear relationship between the input data and the output data is modeled using an RBF neural network;

[0011] S4. Using the constructed model to predict the signal-to-noise ratio of the target performance of the motor under the initial design scheme and corresponding to other parameters;

[0012] S5. Calculate the mass loss percentage of each target performance under the corresponding parameter matching scheme according to the predicted signal-to-noise ratio; the mass loss percentage represents the improvement of different structural parameter matching schemes compared to the initial design scheme;

[0013] S6. Calculate the weight coefficient of each target performance using the fuzzy optimal and worst method;

[0014] S7. Under different parameters, a synthetic index is constructed according to the weight coefficient of each target performance and the quality loss percentage;

[0015] S8. Find the minimum value of the composite index and take the corresponding parameter matching scheme as the optimal design result.

[0016] Furthermore, step S2 specifically includes:

[0017] Based on the principle of Taguchi method, a Taguchi experimental orthogonal table that meets the number of structural parameters and their level numbers is established;

[0018] The parameter matching schemes corresponding to the orthogonal table of the Taguchi experiment were simulated and analyzed in the finite element software, and the target performances under different parameter matching were recorded;

[0019] According to the optimization requirements of each target performance, select appropriate characteristics for the target performance among "looking at the eye", "looking at the big" and "looking at the small", and use the corresponding formula to calculate the signal-to-noise ratio corresponding to each target performance.

[0020] Furthermore, the signal-to-noise ratio (S / N) calculation formulas for the three characteristics of looking at the eye, looking at the big, and looking at the small are:

[0021]

[0022] Among them, y i represents the measured data; n is the total number of samples.

[0023] Furthermore, the calculation formula of the percentage of quality loss PRQL is:

[0024]

[0025] Where η represents the signal-to-noise ratio of the new design; η0 represents the signal-to-noise ratio of the initial design.

[0026] Furthermore, the calculation formula of the composite indicator WPRQL is:

[0027]

[0028] Where: N represents the number of performance indicators; ω N Represents the weight coefficient of the Nth performance indicator; PRQL N Indicates the PRQL value of the Nth performance indicator.

[0029] Furthermore, step S6 specifically includes:

[0030] 01: Select the best and worst indicators;

[0031] 02: Evaluate the importance of the optimal index and all indexes, and get the fuzzy optimal comparison vector: in represents the fuzzy preference of the optimal indicator for indicator j;

[0032] 03: Evaluate the importance of all indicators and the worst indicator, and obtain the fuzzy worst comparison vector: in represents the fuzzy preference of indicator j for the worst indicator;

[0033] 04: Obtain the optimal fuzzy weight set in, They represent the optimal fuzzy weight of the best indicator, the optimal fuzzy weight of the worst indicator, and the optimal fuzzy weight of indicator j respectively;

[0034] 05. The optimal fuzzy weight is converted into the following R single value expression through the centroid method. Calculated by the following formula:

[0035]

[0036] l, m, u represent the lower limit, modal value and upper limit of the triangular fuzzy number respectively.

[0037] Furthermore, the specific implementation process of step S8 is:

[0038] The SAPSO algorithm is used to search for the optimal structural parameter coordination. The synthetic index WPRQL is used as the evaluation function of the SAPSO main optimization program. The optimal structural parameter coordination scheme corresponds to the minimum value of WPRQL. After finding the minimum value of WPRQL, the optimal structural parameter coordination scheme is solved in reverse.

[0039] In general, compared with the prior art, the above technical solution conceived by the present invention can achieve the following beneficial effects.

[0040] (1) The present invention adopts a method combining the Taguchi method with the RBF network, and utilizes the characteristic of the Taguchi method that the number of experiments required can be greatly reduced under the premise of a given number of optimization parameters and the number of levels of each parameter, thereby reducing the difficulty of obtaining RBF network training set data and weakening the dependence of the optimization process on finite element analysis; at the same time, the RBF network can transform the discrete optimization parameter space caused by the Taguchi method into a continuous space, ensuring that the optimization process can find the global optimal solution, avoiding the problem of local optimal solution caused by the single use of the Taguchi method.

[0041] (2) The present invention uses the fuzzy optimal and worst method to obtain the weight coefficient of each performance indicator, making it more reasonable to transform the multi-objective problem into a single-objective problem for analysis and optimization. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 It is a flow chart of the multi-objective optimization method of a permanent magnet motor provided by the present invention;

[0043] Figure 2 is a two-dimensional cross-sectional diagram of a permanent magnet vernier motor;

[0044] Figure 3(a)-Figure 3(c) is the RBF network prediction result provided by the present invention;

[0045] Figure 4 It is the WPRQL index change curve in the SAPSO algorithm process provided by the present invention. DETAILED DESCRIPTION

[0046] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention. In addition, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0047] like Figure 1 As shown, an embodiment of the present invention provides a multi-objective optimization design method for a permanent magnet motor, comprising:

[0048] S1. Complete the preliminary design of the motor, determine the target performance to be optimized, the motor structural parameters involved in the optimization process, and their range of variation and number of levels;

[0049] S2. Use the Taguchi method combined with finite element simulation to obtain the target performance under different parameter combinations, and calculate the signal-to-noise ratio corresponding to each target performance according to the optimization requirements of each target performance;

[0050] Step S2 specifically includes: according to the principle of Taguchi method, establish Taguchi experimental orthogonal table that meets the number of structural parameters and their level numbers; simulate and analyze the parameter matching schemes corresponding to Taguchi experimental orthogonal table in finite element software, and record the target performance under different parameter matching; according to the optimization requirements of each target performance, select appropriate characteristics for the target performance among "Wang Da", "Wang Xiao" and "Wang Mu", and use the corresponding formula to calculate the signal-to-noise ratio corresponding to each target performance. Among them, the signal-to-noise ratio (Signal / Noise, S / N) calculation formulas of the three characteristics of Wang Mu, Wang Da and Wang Xiao are as follows:

[0051]

[0052]

[0053]

[0054] Among them, y i represents the measured data; n is the total number of samples.

[0055] S3. Taking each group of parameters as input data and the signal-to-noise ratio of the corresponding target performance as output data, a radial basis function (RBF) neural network is used to model the nonlinear relationship between the input data and the output data;

[0056] S4. Using the constructed model to predict the signal-to-noise ratio of the target performance of the motor under the initial design scheme and corresponding to other parameters;

[0057] S5. According to the predicted signal-to-noise ratio, the mass loss percentage of each target performance under the corresponding parameter matching scheme is calculated; the mass loss percentage represents the improvement of different structural parameter matching schemes compared with the initial design scheme; the calculation formula of the mass loss percentage (Percentage Reduction of Quality Loss, PRQL) is:

[0058]

[0059] Where η represents the signal-to-noise ratio of the new design; η0 represents the signal-to-noise ratio of the initial design.

[0060] S6. Calculate the weight coefficient of each target performance using the Fuzzy Best-Worst Method (FBWM); Step S6 specifically includes:

[0061] 01: Select the best and worst indicators, that is, the most important performance indicator and the least important performance indicator;

[0062] 02: Evaluate the importance of the optimal indicator and all indicators, that is, the priority of the most important indicator compared to other indicators. The evaluation results are expressed in the form of language evaluation levels, including 5 levels, and converted into corresponding triangular fuzzy numbers through the membership function shown in Table 5. The fuzzy optimal comparison vector is: in Represents the fuzzy preference of the optimal indicator for indicator j.

[0063] Table 5. Language assessment levels and conversion rules

[0064]

[0065] 03: Evaluate the importance of all indicators and the worst indicator, that is, the priority of all indicators compared to the least important indicator. The evaluation results are presented in the same way as step 02, and the fuzzy worst comparison vector is finally obtained: in Represents the fuzzy preference of indicator j for the worst indicator.

[0066] 04: Obtain the optimal fuzzy weight set

[0067] By minimizing and to obtain the larger value of . They represent the optimal fuzzy weight of the best indicator, the optimal fuzzy weight of the worst indicator, and the optimal fuzzy weight of indicator j. They are all represented by triangular fuzzy numbers. l, m, u represent the lower limit, modal value and upper limit of a triangular fuzzy number respectively. According to the above statement, the optimal fuzzy weight is solved according to the following nonlinear constrained optimization problem:

[0068]

[0069]

[0070] The above formula can be transformed into the following form:

[0071]

[0072]

[0073] in Considering l ξ ≤m ξ ≤u ξ , assuming there is an optimal And there is k * ≤l ξ , the above formula can be transformed into:

[0074]

[0075]

[0076] 05. The weights expressed by triangular fuzzy numbers can be converted into the following R single-value expression by the centroid method, Calculated by the following formula:

[0077]

[0078] S7. Under different parameters, a composite index is constructed according to the weight coefficient and quality loss percentage of each target performance; the calculation formula of the composite index WPRQL (Weighted Percentage Reduction of Quality Loss) is:

[0079]

[0080] Where: N represents the number of performance indicators; ω N Represents the weight coefficient of the Nth performance indicator; PRQL N Indicates the PRQL value of the Nth performance indicator.

[0081] S8. Find the minimum value of the composite index and take the corresponding parameter matching scheme as the optimal design result.

[0082] The specific implementation process of step S8 is: using the SAPSO algorithm to search for the optimal structural parameter matching, using the synthetic index WPRQL as the evaluation function of the SAPSO main optimization program, the optimal structural parameter matching scheme corresponds to the minimum value of WPRQL, and after finding the minimum value of WPRQL, the optimal structural parameter matching scheme is reversely solved.

[0083] In order to more clearly illustrate the above technical solution, the embodiment of the present invention also provides the following design example:

[0084] First, the preliminary design of an 18-slot 15-pole surface-mounted permanent magnet vernier motor was completed. The initial design scheme of its structural parameters is shown in Table 1, and the two-dimensional structure diagram is shown in Figure 2 .

[0085] Table 1. Initial design of surface mount vernier motor

[0086]

[0087] Then, the optimized performance indicators are selected as larger power factor, larger average torque and smaller torque pulsation. Pole arc coefficient, split ratio, slot width, yoke thickness and permanent magnet thickness are selected as structural parameters involved in the optimization process. The range of these structural parameters and the value step are shown in Table 2:

[0088] Table 2. Variation range and number of levels of optimization parameters

[0089]

[0090] According to the Taguchi experimental design principle, this is a five-factor four-level experiment, which requires 16 independent experiments in total. The Taguchi experimental orthogonal table is shown in Table 3:

[0091] Table 3. Taguchi experiment orthogonal table

[0092]

[0093]

[0094] Under the above structural parameters, the finite element simulation software is used to obtain the average torque, torque pulsation and power factor under different parameter combinations, and their values ​​are filled into the response column of the Taguchi experiment orthogonal table, and then the respective signal-to-noise ratios are calculated. The calculation results are shown in Table 4:

[0095] Table 4. Performance index values ​​and signal-to-noise ratio under different parameter combinations

[0096]

[0097] The structural parameters in Table 3 are used as input, and the signal-to-noise ratio in Table 4 is used as output. The RBF network is used to perform nonlinear modeling of input and output, and the signal-to-noise ratio of the initial design scheme and the design scheme with other parameters can be predicted. The prediction results of the three indicators are as follows: Figure 3(a)-Figure 3(c) shown.

[0098] Now, it is necessary to determine the weight values ​​of the three performance indicators in the optimization process. For this example, the power factor is selected as the most important indicator (optimal indicator) among the three indicators, and the torque pulsation is selected as the least important indicator (worst indicator). Table 6 shows the importance evaluation results of the optimal indicator and all indicators, and Table 7 shows the importance evaluation results of all indicators and the worst indicator.

[0099] Table 6. Evaluation results of the importance of the best index and all indexes

[0100]

[0101] Table 7. Importance evaluation results of all indicators and the worst indicator

[0102]

[0103] The final weight values ​​of the three performance indicators are shown in Table 8.

[0104] Table 8. Weight values ​​of each performance indicator

[0105]

[0106] The output of the RBF network is first converted into a PRQL value, and then converted into a WPRQL value according to the weights of the three performance indicators. It is then brought into the SAPSO algorithm as input for minimization optimization. The change curve of WPRQL during the optimization process is shown in the figure below: Figure 4 As shown in the figure, after obtaining the minimum value, the optimal structural parameter matching can be solved in reverse.

[0107] It will be easily understood by those skilled in the art that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A multi-objective optimization design method for a permanent magnet motor, characterized in that: include: S1. Complete the preliminary design of the motor, determine the target performance to be optimized, the motor structural parameters involved in the optimization process, and their range of variation and number of levels; S2. Use the Taguchi method combined with finite element simulation to obtain the target performance under different parameter combinations, and calculate the signal-to-noise ratio corresponding to each target performance according to the optimization requirements of each target performance; S3. Taking each set of parameters as input data and the signal-to-noise ratio of the corresponding target performance as output data, the nonlinear relationship between the input data and the output data is modeled using an RBF neural network; S4. Using the constructed model to predict the signal-to-noise ratio of the target performance of the motor under the initial design scheme and corresponding to other parameters; S5. According to the predicted signal-to-noise ratio, the mass loss percentage of each target performance under the corresponding parameter matching scheme is calculated; the mass loss percentage represents the improvement of different structural parameter matching schemes compared with the initial design scheme; the calculation formula of the mass loss percentage PRQL is: Where η represents the signal-to-noise ratio of the new design; η0 represents the signal-to-noise ratio of the initial design; S6. Calculate the weight coefficient of each target performance using the fuzzy optimal and worst method; S7. Under different parameters, a synthetic index is constructed according to the weight coefficient of each target performance and the quality loss percentage; S8. Find the minimum value of the composite index and take the corresponding parameter matching scheme as the optimal design result.

2. A multi-objective optimization design method for a permanent magnet motor according to claim 1, characterized in that: Step S2 specifically includes: Based on the principle of Taguchi method, a Taguchi experimental orthogonal table that meets the number of structural parameters and their level numbers is established; The parameter matching schemes corresponding to the orthogonal table of the Taguchi experiment were simulated and analyzed in the finite element software, and the target performances under different parameter matching were recorded; According to the optimization requirements of each target performance, the target performance is selected from "looking at the eye", "looking at the big" and "looking at the small", and the corresponding formula is used to calculate the signal-to-noise ratio corresponding to each target performance.

3. A multi-objective optimization design method for a permanent magnet motor according to claim 2, characterized in that: The signal-to-noise ratio (S / N) calculation formulas for the three characteristics of looking at the eye, looking at the big, and looking at the small are: in, represents the average value of the measured data, y i represents the measured data; n is the total number of samples.

4. A multi-objective optimization design method for a permanent magnet motor according to any one of claims 1 to 3, characterized in that: The calculation formula of the composite indicator WPRQL is: Where: N represents the number of performance indicators; ω N Represents the weight coefficient of the Nth performance indicator; PRQL N Indicates the PRQL value of the Nth performance indicator.

5. A multi-objective optimization design method for a permanent magnet motor according to any one of claims 1 to 3, characterized in that: Step S6 specifically includes: 01: Select the best and worst indicators; 02: Evaluate the importance of the optimal index and all indexes, and get the fuzzy optimal comparison vector: in represents the fuzzy preference of the optimal indicator for indicator j; 03: Evaluate the importance of all indicators and the worst indicator, and obtain the fuzzy worst comparison vector: in represents the fuzzy preference of indicator j for the worst indicator; 04: Obtain the optimal fuzzy weight set in, represents the optimal fuzzy weight of index j; 05. The optimal fuzzy weight is converted into the following R single value expression through the centroid method. Calculated by the following formula: l, m, u represent the lower limit, modal value and upper limit of the triangular fuzzy number respectively.

6. A multi-objective optimization design method for a permanent magnet motor according to any one of claims 1 to 3, characterized in that: The specific implementation process of step S8 is: The SAPSO algorithm is used to search for the optimal structural parameter coordination. The synthetic index WPRQL is used as the evaluation function of the SAPSO main optimization program. The optimal structural parameter coordination scheme corresponds to the minimum value of WPRQL. After finding the minimum value of WPRQL, the optimal structural parameter coordination scheme is solved in reverse.

7. A multi-objective optimization design system for a permanent magnet motor, characterized in that: include: A computer readable storage medium and a processor; The computer-readable storage medium is used to store executable instructions; The processor is used to read the executable instructions stored in the computer-readable storage medium to execute the multi-objective optimization design method for a permanent magnet motor according to any one of claims 1 to 6.

Images