Design method of high-speed angular contact ball bearing

Through the combination of orthogonal test method and BP neural network, the internal structural parameters of angular contact ball bearings are optimized, the problem of poor heat dissipation effect is solved, more efficient heat dissipation and more rigorous design are achieved, and the reliability of bearings is improved. sex.

CN118940425APending Publication Date: 2024-11-12AVIC HARBIN BEARING CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202410928044.4
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-11
Publication Date
2024-11-12

AI Technical Summary

Technical Problem

The heat dissipation effect of existing angular contact ball bearings is poor during the design process, which may cause the bearing to burn and the empirical parameters are not rigorous.

Method used

The factor table is established using the orthogonal experiment method, the calorific value is calculated using simulation software, and the calorific value prediction model is obtained through BP neural network training, and the internal structural parameters of the minimum calorific value are found in combination with the genetic algorithm to optimize the bearing design.

Benefits of technology

It effectively reduces the temperature of high-speed angular contact ball bearings, improves heat dissipation efficiency, provides more rigorous parameter design, and improves the reliability of the bearings.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118940425A_ABST
    Figure CN118940425A_ABST
Patent Text Reader

Abstract

The invention discloses a high-speed angular contact ball bearing design method and belongs to the field of high-speed ball bearing design. The problem that in the high-speed working process of an existing angular contact ball bearing, the calorific value is large is solved. According to the method, the orthogonal factor level table is utilized, and simulation software is adopted to carry out simulation calculation on the calorific value of the bearing when factors are at different levels; establishing a BP neural network, and training the BP neural network by using the orthogonal factor table and the corresponding heating value result of the simulation calculation; obtaining a calorific value prediction model; taking the orthogonal test factor table as the input of a calorific value prediction model, obtaining a calorific value prediction result, and searching a factor level when the calorific value is minimum by using a genetic algorithm; and acquiring parameters of the high-speed angular contact ball bearing. The method is suitable for high-speed angular contact ball bearing design.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention belongs to the field of high-speed ball bearing design. Background Art

[0002] Rolling bearings are widely used as basic components in mechanical equipment and are one of the key components that determine the safe and reliable operation of mechanical equipment. From household appliances and civil equipment to rail transportation, precision machine tools, and aerospace, their normal operation cannot be separated from rolling bearings.

[0003] With the development and progress of science and technology in today's society, mechanical equipment is increasingly tending towards high performance, high speed, high load and complexity. Rolling bearings are required to have high speed, high precision, high reliability and high durability. Angular contact ball bearings are widely used in aircraft engines and precision machine tool spindles due to their high manufacturing precision, high limit speed, high load capacity and small space occupation.

[0004] During the operation of rolling bearings, a large amount of heat will be generated due to the friction between internal parts and the effect of external loads. When the heat dissipation rate is slower than the heat generation rate, more heat will accumulate in the bearing. If the bearing generates a large amount of heat in a short period of time and cannot be dissipated in time, the bearing may burn out. Therefore, it is of great significance to analyze the heat generation of the bearing. In the design process of angular contact ball bearings, parameters such as clearance, contact angle, and groove curvature radius are usually based on experience. These parameters have a great influence on the heat generation of high-speed angular contact ball bearings. It is obviously not rigorous to rely only on experience. Summary of the invention

[0005] The present invention aims to solve the problem of poor heat dissipation effect in the existing angular contact ball bearing design process, and now provides a high-speed angular contact ball bearing design method.

[0006] The invention provides a method for designing a high-speed angular contact ball bearing, the method comprising:

[0007] Step 1: Based on the known factors affecting the heat generation of high-speed angular contact ball bearings, an orthogonal factor table is established using the orthogonal test method;

[0008] Step 2: using the orthogonal factor table and simulation software to simulate and calculate the heat generation of the bearing at different factor levels;

[0009] Step 3: Establish a BP neural network, and train the BP neural network using the orthogonal factor table and the calorific value result of the corresponding simulation calculation; obtain the trained BP neural network and obtain the calorific value prediction model;

[0010] Step 4: Use the orthogonal test factor table as the input of the calorific value prediction model to obtain the calorific value prediction result;

[0011] Step 5: Using the orthogonal test factor table as the population individuals in the genetic algorithm, and using the prediction results as the individual fitness values;

[0012] Step 6: Use the roulette method to select N individuals with the smallest fitness value from the population to form a new population, and then perform a crossover mutation operation on the new population to obtain a transformed population of N individuals;

[0013] Step 7: Take the transformed population of N individuals as the input of the trained neural network, and obtain the N corresponding heat values; determine whether the maximum number of iterations M is reached, and if so, obtain the parameters of the individual corresponding to the minimum heat value as the high-speed angular contact ball bearing parameters; otherwise, return to step 6 and continue to iterate until the maximum number of iterations M is reached, and obtain the parameters of the individual corresponding to the minimum heat value as the high-speed angular contact ball bearing parameters. The M=100.

[0014] Furthermore, in the present invention, in step 3, the process of training the BP neural network using the influencing factor table and the corresponding heating value is as follows:

[0015] Step 3: Initialize the BP neural network according to the influencing factor table sequence;

[0016] Step 3.2: According to the input vector X, the connection weights ω between the input layer and the stable layer are calculated. ij And the hidden layer threshold a, calculate the hidden layer output;

[0017] Step 3. According to the hidden layer output, connect the weight ω jk and output layer threshold b, calculate the BP neural network prediction output;

[0018] Step 34: Calculate the network prediction error based on the network prediction output and heat value;

[0019] Step 35: Use network prediction error to update network connection weight ω ij and ω jk ;

[0020] Step 36: Update the network node thresholds a and b according to the network prediction error;

[0021] Step 37: Determine whether the algorithm iteration is finished. If not, return to step 32; otherwise, complete the training of the BP neural network.

[0022] Furthermore, in step 31, the specific process of initializing the BP neural network is as follows:

[0023] Using the influencing factor table X and the corresponding heat value Y, determine the number of network input layer nodes n, the number of hidden layer nodes l, and the number of output layer nodes m; and calculate the connection weights ω between the input layer, the hidden layer, and the stable layer. ijThe connection between the hidden layer and the output layer neurons ω jk , the hidden layer threshold a, the output layer threshold b are randomly initialized, and the learning rate and neuron activation function are given.

[0024] Furthermore, in the present invention, in step 32, the calculation method of the hidden layer output is:

[0025]

[0026] In the formula, H j is the jth hidden layer output, l is the number of hidden layer nodes; f is the hidden layer activation function, a j is the threshold of the jth hidden layer; x i Represents the i-th orthogonal experimental level, i=1,2...,n, j=1,2,…,l.

[0027] Furthermore, in the present invention, in step 33, the specific method for calculating the BP neural network prediction output is:

[0028]

[0029] In the formula, O k represents the predicted calorific value at the kth orthogonal test level, k = 1, 2..., m, b k represents the output layer threshold at the kth orthogonal test level.

[0030] Further, in the present invention, in steps 3 and 4, the method for calculating the network prediction error is:

[0031] e k =Y k -O k

[0032] Y k Represents the calculated calorific value at the kth orthogonal test level.

[0033] Furthermore, in the present invention, in step 35, the network connection weight ω is updated using the network prediction error. ij and ω jk The method is:

[0034]

[0035] ω jk =ω jk +ηH j e k

[0036] Where η is the learning rate.

[0037] Furthermore, in the present invention, in step 36, the method for updating the network node thresholds a and b according to the network prediction error is:

[0038]

[0039] b k =b k +e k

[0040] Among them, b k Represents the hidden layer threshold at the kth orthogonal test level.

[0041] The high-speed angular contact ball bearing design method of the present invention first determines the internal structural parameters that affect the heat generation of the high-speed angular contact ball bearing. In order to select suitable representative test factors, an orthogonal test factor level table is selected based on orthogonal design. The heat generation under each orthogonal test is calculated based on bearing professional simulation software. The BP neural network is trained based on the orthogonal test factor level table and its corresponding heat generation. After the BP neural network training is completed, a genetic algorithm is used to find the internal structural parameters of the bearing when the heat generation of the bearing is the smallest, which provides a reference for the optimal bearing internal structural parameter design. The parameters determined by the method of the present invention effectively reduce the temperature of the high-speed angular contact ball bearing. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 It is a schematic diagram of the structure of the BP neural network described in the present invention;

[0043] Figure 2 The present invention is a flow chart of the method. DETAILED DESCRIPTION

[0044] The following will be combined with the accompanying drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work belong to the scope of protection of the present invention. It should be noted that the embodiments of the present invention and the features in the embodiments can be combined with each other without conflict.

[0045] Specific implementation method 1: refer to Figure 1 and Figure 2 Specifically describing this embodiment, a high-speed angular contact ball bearing design method described in this embodiment includes:

[0046] Step 1: Based on the known factors affecting the heat generation of high-speed angular contact ball bearings, an orthogonal factor table is established using the orthogonal test method;

[0047] Step 2: using the orthogonal factor table and simulation software to simulate and calculate the heat generation of the bearing at different factor levels;

[0048] Step 3: Establish a BP neural network, and train the BP neural network using the orthogonal factor table and the calorific value result of the corresponding simulation calculation; obtain the trained BP neural network and obtain the calorific value prediction model;

[0049] Step 4: Use the orthogonal test factor table as the input of the calorific value prediction model to obtain the calorific value prediction result;

[0050] Step 5: Using the orthogonal test factor table as the population individuals in the genetic algorithm, and using the prediction results as the individual fitness values;

[0051] Step 6: Use the roulette method to select N individuals with the smallest fitness value from the population to form a new population, and then perform a crossover mutation operation on the new population to obtain a transformed population of N individuals;

[0052] Step 7: Take the transformed population of N individuals as the input of the trained neural network, and obtain the N corresponding heat values; determine whether the maximum number of iterations M is reached, and if so, obtain the parameters of the individual corresponding to the minimum heat value as the high-speed angular contact ball bearing parameters; otherwise, return to step 6 and continue to iterate until the maximum number of iterations M is reached, and obtain the parameters of the individual corresponding to the minimum heat value as the high-speed angular contact ball bearing parameters. The M=100.

[0053] Further, in this embodiment, in step 3, the process of training the BP neural network using the influencing factor table and the corresponding heating value is:

[0054] Step 3: Initialize the BP neural network according to the influencing factor table sequence;

[0055] Step 3.2: According to the input vector X, the connection weights ω between the input layer and the stable layer are calculated. ij And the hidden layer threshold a, calculate the hidden layer output;

[0056] Step 3. According to the hidden layer output, connect the weight ω jk and output layer threshold b, calculate the BP neural network prediction output;

[0057] Step 34: Calculate the network prediction error based on the network prediction output and heat value;

[0058] Step 35: Use network prediction error to update network connection weight ω ij and ω jk ;

[0059] Step 36: Update the network node thresholds a and b according to the network prediction error;

[0060] Step 37: Determine whether the algorithm iteration is finished. If not, return to step 32; otherwise, complete the training of the BP neural network.

[0061] Furthermore, in step 31, the specific process of initializing the BP neural network is as follows:

[0062] The influencing factor table X and the corresponding heat value Y are used to determine the number of network input layer nodes n, the number of hidden layer nodes l, and the number of output layer nodes m; and the connection weights ω between the input layer, the hidden layer, and the output layer are calculated. ij The connection between the hidden layer and the output layer neurons ω jk , the hidden layer threshold a, the output layer threshold b are initialized, and the learning rate and neuron activation function are given.

[0063] In this implementation, when the number of factors in the orthogonal factor table is 16, the number of input layer nodes n and the number of output layer nodes m are both 16.

[0064] Furthermore, in this implementation, in step 32, the calculation method of the hidden layer output is:

[0065]

[0066] In the formula, H j is the jth hidden layer output, l is the number of hidden layer nodes; f is the hidden layer activation function, a j is the threshold of the jth hidden layer; x i Represents the i-th orthogonal experimental level, i=1,2...,n.

[0067] Furthermore, in this implementation, in step 33, the specific method for calculating the BP neural network prediction output is:

[0068]

[0069] In the formula, O k represents the predicted calorific value at the kth orthogonal test level, k = 1, 2..., m, b k represents the output layer threshold at the kth orthogonal test level.

[0070] Further, in this implementation, in steps 3 and 4, the method for calculating the network prediction error is:

[0071] e k =Y k -O k

[0072] Among them, Y kRepresents the calculated calorific value at the kth orthogonal test level.

[0073] Furthermore, in this embodiment, in steps 3 and 5, the network connection weight ω is updated using the network prediction error. ij and ω jk The method is:

[0074]

[0075] ω jk =ω jk +ηH j e k

[0076] Where η is the learning rate.

[0077] Furthermore, in this embodiment, in step 36, the method for updating the network node thresholds a and b according to the network prediction error is:

[0078]

[0079] b k =b k +e k

[0080] Among them, b k Represents the hidden layer threshold at the kth orthogonal test level.

[0081] The main idea of ​​the high-speed angular contact ball bearing design method proposed in this embodiment is to solve the problem that the internal structural parameters of high-speed angular contact ball bearings usually rely on experience, train a BP neural network based on the orthogonal test analysis table and the calculation results of bearing professional simulation software, and after the BP neural network training is completed, use a genetic algorithm to find the internal structural parameters of the bearing when the heat generation of the bearing is the minimum, and then use the genetic algorithm to obtain the internal structural parameters of the bearing when the heat generation is the minimum.

[0082] The present invention uses the nonlinear fitting ability of the BP neural network and the nonlinear optimization ability of the genetic algorithm to find the internal structural parameters when the heat generation of the bearing is the minimum. The present invention uses 16 orthogonal levels as the BP neural network input, and the heat generation calculated by the 16 orthogonal level simulation is used as the BP neural network output to train the BP neural network. The trained BP neural network can predict the heat generation of the bearing with different internal parameters. The genetic algorithm extreme value optimization uses the predicted value result of the trained BP neural network as the individual fitness value, and finds the minimum value of the heat generation and its corresponding input value through selection, crossover and mutation operations.

[0083] The BP neural network of the present invention has 5 input parameters (A outer groove radius (mm) B guide clearance (mm) C inner groove radius (mm) D radial clearance (mm) E half of the gasket thickness (mm)) and 1 output parameter (heat generation). The genetic algorithm individuals are encoded with real numbers, and each individual is a real number string. Since the BP neural network has 5 input parameters, the individual length is 5. The individual fitness value is the BP neural network prediction value. The smaller the fitness value, the better the individual. The minimum value of the heat generation and its corresponding input value are found through selection, crossover and mutation operations.

[0084] The selection operation uses the roulette method to select individuals with good fitness from the population to form a new population. The crossover operation randomly selects an individual from the population and mutates it with a certain probability to obtain a new individual. The mutation operation randomly selects an individual from the population and mutates it with a certain probability to obtain a new individual. Iterate to find the optimal individual.

[0085] Table 1 Orthogonal factor level table

[0086]

[0087]

[0088] Specifically, the design method of high-speed angular contact ball bearings is as follows: Figure 1 As shown, the following steps are included:

[0089] The present invention is verified by taking a certain engine main shaft bearing as an example. After simulation analysis of the influence of single factor heat generation on the bearing, it is found that the factors affecting the heat generation of the bearing are: outer groove radius, inner groove radius, guide clearance, etc. Based on orthogonal test, the factor level table is determined as L16(4 5 ). Based on the bearing professional simulation software and the orthogonal test factor level table, the heat generation of the bearing under each test condition is calculated. During the calculation process, it is assumed that the bearing installation, speed and other conditions are constant. The orthogonal test factor level table and its test results are shown in the table below.

[0090] Table 2 Orthogonal table and test results

[0091]

[0092] For this case, according to the orthogonal table, there are 5 input parameters and 1 output parameter, and the BP neural network structure is determined to be 5-10-1.

[0093] In the genetic algorithm, individuals are encoded with real numbers. Since the optimization function has 5 input parameters, the length of the individual is 5. The individual fitness value is the BP neural network prediction value. The smaller the fitness value, the better the individual. The number of iterations of the genetic algorithm is 100, the population size is 20, the crossover probability is 0.4, the mutation probability is 0.2, and the iterative optimization is performed. The optimization result is [4.90610.38845.0576 0.0861 0.1026]. Under this condition, the heat generated by the bearing is 3376.9W, which is used as the optimal design parameter.

[0094] Although the present invention is described herein with reference to specific embodiments, it should be understood that these embodiments are merely examples of the principles and applications of the present invention. It should therefore be understood that many modifications may be made to the exemplary embodiments and that other arrangements may be devised without departing from the spirit and scope of the present invention as defined by the appended claims. It should be understood that the various dependent claims and features described herein may be combined in a manner different from that described in the original claims. It should also be understood that the features described in conjunction with a single embodiment may be used in other described embodiments.

Claims

1. A design method for high-speed angular contact ball bearings. It is characterized in that The method includes: Step 1: Based on the known factors affecting the heat generation of high-speed angular contact ball bearings, an orthogonal factor table is established using the orthogonal test method; Step 2: using the orthogonal factor table and simulation software to simulate and calculate the heat generation of the bearing at different factor levels; Step 3: Establish a BP neural network, and train the BP neural network using the orthogonal factor table and the calorific value result of the corresponding simulation calculation; obtain the trained BP neural network and obtain the calorific value prediction model; Step 4: Use the orthogonal test factor table as the input of the calorific value prediction model to obtain the calorific value prediction result; Step 5: Using the orthogonal test factor table as the population individuals in the genetic algorithm, and using the prediction results as the individual fitness values; Step 6: Use the roulette method to select N individuals with the smallest fitness value from the population to form a new population, and then perform a crossover mutation operation on the new population to obtain a transformed population of N individuals; Step 7. Use the transformed population of N individuals as the input of the trained neural network to obtain N corresponding heating values; determine whether the maximum number of iterations M is reached. If so, obtain the parameters of the individual corresponding to the minimum heating value as the parameters of the high-speed angular contact ball bearing; otherwise, return to step 6 and continue to iterate until the maximum number of iterations M is reached, and obtain the parameters of the individual corresponding to the minimum heating value as the parameters of the high-speed angular contact ball bearing.

2. A high-speed angular contact ball bearing design method according to claim 1, It is characterized in that In step 3, the process of training the BP neural network using the influencing factor table and the corresponding heating value is as follows: Step 3: Initialize the BP neural network according to the influencing factor table sequence; Step 3.2: According to the input vector X, the connection weights ω between the input layer and the stable layer are calculated. ij And the hidden layer threshold a, calculate the hidden layer output; Step 3. According to the hidden layer output, connect the weight ω jk and output layer threshold b, calculate the BP neural network prediction output; Step 34: Calculate the network prediction error based on the network prediction output and heat value; Step 35: Use network prediction error to update network connection weight ω ij and ω jk ; Step 36: Update the network node thresholds a and b according to the network prediction error; Step 37: Determine whether the algorithm iteration is finished. If not, return to step 32; Otherwise, the training of the BP neural network is completed.

3. A high-speed angular contact ball bearing design method according to claim 2, It is characterized in that In step 31, the specific process of initializing the BP neural network is as follows: The influencing factor table X and the corresponding heat value Y are used to determine the number of network input layer nodes n, the number of hidden layer nodes l, and the number of output layer nodes m; and the connection weights ω between the input layer, the hidden layer, and the output layer are calculated. ij The connection between the hidden layer and the output layer neurons ω jk , the hidden layer threshold a, the output layer threshold b are initialized, and the learning rate and neuron activation function are given.

4. A high-speed angular contact ball bearing design method according to claim 2 or 3, It is characterized in that In step 32, the calculation method of the hidden layer output is: In the formula, H j is the jth hidden layer output, l is the number of hidden layer nodes; f is the hidden layer activation function, a j is the threshold of the jth hidden layer; x i Represents the i-th orthogonal experimental level, i=1,2...,n.

5. A high-speed angular contact ball bearing design method according to claim 4, It is characterized in that In step 33, the specific method for calculating the BP neural network prediction output is: In the formula, O k represents the predicted calorific value at the kth orthogonal test level, k = 1, 2..., m, b k represents the output layer threshold at the kth orthogonal test level.

6. A high-speed angular contact ball bearing design method according to claim 5, It is characterized in that In steps 3 and 4, the method for calculating the network prediction error is: and k =And k -EITHER k Among them, Y k represents the calculated calorific value at the kth orthogonal test level, e k Represents the network prediction error at the kth orthogonal test level.

7. A high-speed angular contact ball bearing design method according to claim 6, It is characterized in that In step 3 and 5, the network prediction error is used to update the network connection weight ω ij and ω jk The method is: oh jk =ω jk +ηH j e k Where η is the learning rate.

8. A high-speed angular contact ball bearing design method according to claim 7, It is characterized in that In step 36, the method for updating the network node thresholds a and b according to the network prediction error is: b k =b k +e k Among them, a k represents the hidden layer threshold at the kth orthogonal test level, b k Represents the output layer threshold at the kth orthogonal test level.