CN-122021144-A - NRBO-ELM algorithm-based gearbox parallelism prediction method

CN122021144ACN 122021144 ACN122021144 ACN 122021144ACN-122021144-A

Abstract

The invention belongs to the field of monitoring, and particularly relates to a gearbox parallelism prediction method based on NRBO-ELM algorithm. The invention aims to solve the problem that the existing gearbox parallelism prediction method has low prediction precision. The invention discloses a gearbox parallelism prediction method based on NRBO-ELM algorithm. The method comprises the step S1 of calculating distribution data of the bending stress of the tooth root along the tooth width direction of the shafting in a deflection state by a finite element method. The method comprises the steps of carrying out linear normalization processing on data, dividing the data into a training data set and a testing data set, creating an Extreme learning machine (Extreme LEARNING MACHINE, ELM) model, initializing internal parameters of the model, carrying out optimization on the internal parameters of the Extreme learning machine model (ELM) by utilizing a Newton-Raphson-Based optimization algorithm (NRBO) through a Newton-Raphson iteration method, searching an optimal parameter combination of the ELM prediction model, and solving the problem of low prediction precision of the existing gearbox parallelism prediction method.

Inventors

GUO FENG
WANG TIANHAO

Assignees

中国船舶集团有限公司第七〇三研究所

Dates

Publication Date: 20260512
Application Date: 20260121

Claims (10)

1. A gearbox parallelism prediction method based on NRBO-ELM algorithm is characterized by comprising the following steps: 1. Constructing a gear box parallelism prediction model, and obtaining a trained gear box parallelism prediction model; 2. Acquiring stress distribution data of a gear box to be predicted, and inputting the stress distribution data of the gear box to be predicted into a trained gear box parallelism prediction model to obtain the parallelism of the gear box to be predicted; The method comprises the steps of constructing a gear box parallelism prediction model to obtain a trained gear box parallelism prediction model, wherein the specific process is as follows: step one, obtaining a training set; and training the constructed gear box parallelism prediction model according to a training set to obtain a trained gear box parallelism prediction model.
2. A gearbox parallelism prediction method based on NRBO-ELM algorithm according to claim 1, wherein, The specific process of acquiring the training set in the first step is as follows: the method comprises the steps of constructing a finite element model of the gear box, wherein the concrete process comprises the following steps: Extracting the shafting space position of each pair of meshing gears in the gear box, and setting the shafting space position of each pair of meshing gears in finite element software according to the extracted shafting space position of each pair of meshing gears in the gear box; And step two, performing finite element simulation calculation according to the constructed gear box finite element model, and constructing a training set.
3. The method for predicting the parallelism of the gear box based on NRBO-ELM algorithm according to claim 2, wherein the specific process of setting the shafting space positions of a pair of meshing gears in finite element software according to the extracted shafting space positions of a certain pair of meshing gears in the gear box in one-to-one mode is as follows: according to the extracted shafting space position of a certain pair of meshing gears in the gear box, a floating bearing position and three fixed bearing positions are set in finite element software, In the first step, finite element simulation calculation is carried out according to a finite element model of a constructed gear box, and a training set is constructed, wherein the specific process is as follows: Setting N floating bearing positions in finite element software to obtain N shafting inclined states, wherein N is a positive integer; Step two and two, calculating the distribution data Y of tooth root bending stress of the meshing gear in the gear box along the tooth width direction under N shafting deflection states of the shafting of the gear box respectively through finite element software, Wherein, the distribution data of tooth root bending stress of the meshing gear in the gear box in the n-th shafting deflection state along the tooth width direction is expressed as , ; ; Normalizing the distribution data of the tooth root bending stress of the meshing gears in the gear boxes along the tooth width direction under N shafting deflection states to obtain the distribution data of the tooth root bending stress of the meshing gears in the N normalized gear boxes along the tooth width direction ; Wherein, the distribution data of tooth root bending stress of the meshing gear in the gear box in the n-th shafting deflection state along the tooth width direction Normalization processing is carried out to obtain distribution data of tooth root bending stress of the meshing gear in the gear box in the n-th shafting deflection state after normalization in the tooth width direction The specific process is as follows: In the formula, Representation of The i-th element of (a); Representation of Is a maximum value among the elements of (a), Representation of Minimum of elements in (2); Representation of I is a positive integer; Step two and four, distributing data of tooth root bending stress of the meshing gears in the N normalized gear boxes in the direction along the tooth width Scrambling to obtain scrambled data The scrambled data And (5) dividing according to the proportion to obtain a training set.
4. The method for predicting the parallelism of the gear box based on NRBO-ELM algorithm according to claim 3, wherein the model for predicting the parallelism of the gear box in the second step is an extreme learning machine model; training the constructed gear box parallelism prediction model according to a training set to obtain a trained gear box parallelism prediction model, wherein the specific process is as follows: step two, constructing an extreme learning machine model and determining internal parameters to be optimized in the extreme learning machine model; inputting normalized stress distribution data in a training set into an initialized extreme learning machine model, and outputting a prediction result by the extreme learning machine model; And thirdly, optimizing and training the internal parameters to be optimized of the extreme learning machine model by using NRBO algorithm according to the input and output of the extreme learning machine model, and stopping training when the training times reach the maximum, so as to obtain the trained extreme learning machine model.
5. The method for predicting the parallelism of a gear box based on NRBO-ELM algorithm as set forth in claim 4, wherein the extreme learning machine in the second step comprises an input layer, a hidden layer and an output layer; the specific process for constructing the extreme learning machine model in the step two is as follows: setting M input layer nodes of the input layer; setting L hidden layer nodes in the hidden layer; setting K output layer nodes in the output layer, wherein M, L and K are positive integers; The internal parameters to be optimized in the extreme learning machine model comprise a first type of optimized parameter set and a second type of optimized parameter set; The first type of optimization parameters are weights from an input layer node to a hidden layer node, wherein the weights from an i node to a j hidden layer node of the input layer are as follows , The second class of optimization parameters are bias values of hidden layer nodes, wherein the bias value of the j-th hidden layer node is 。
6. The method for predicting the parallelism of a gear box based on NRBO-ELM algorithm according to claim 5, wherein in the second and third steps, according to the input and output of the extreme learning machine model, the internal parameters to be optimized of the extreme learning machine model are optimally trained by using NRBO algorithm, and when the training times reach the maximum, the training is stopped, and the trained extreme learning machine model is obtained, wherein the specific process is as follows: Setting NRBO algorithm population particle number N p , maximum iteration number Max_IT and decision coefficient DF, initializing NRBO algorithm population position, and representing NRBO algorithm population position as , Step two, calculating the fitness value of all particles in the population, and selecting an optimal particle position vector and a worst particle position vector according to the fitness value of all particles; step two, three, namely according to the optimal particle position vector sum Worst particle position vector Updating all particles in the population; Calculating the fitness value of all particles in the updated population, and updating the optimal particle position vector according to the fitness value of all particles in the updated population And worst particle position vector Obtaining updated optimal particle position vector And an updated worst particle position vector ; Returning to the second, third and third step when the iteration number does not reach the maximum iteration number; stopping iteration when the iteration number reaches the maximum iteration number, and updating the updated optimal particle position vector As the trained optimal particle position vector, entering a second, third and fifth step; and step two, three and five, taking the extreme learning machine model corresponding to the trained optimal particle position as the trained extreme learning machine model.
7. The method for predicting parallelism of gear box based on NRBO-ELM algorithm as set forth in claim 6, wherein the step of setting up the number N p of population particles of NRBO algorithm, the maximum iteration number Max_IT, the decision coefficient DF and initializing the population position of NRBO algorithm, wherein the population position of NRBO algorithm is expressed as Expressed by the formula: In the formula, Represent the first The position vector of each particle is expressed as: In the formula, Representation of Middle (f) Position parameters, each representing an internal parameter to be optimized in the extreme learning machine model; representing the number of location parameters; Calculating the third in the population in the second, third and fourth steps The fitness value of each particle is expressed as Expressed by the formula: In the formula, The true value is represented by a value that is true, Represent the first And the predicted value output by the extreme learning machine model corresponding to each particle.
8. The method for predicting parallelism of a gearbox based on NRBO-ELM algorithm as set forth in claim 7, wherein said steps two, three and three are based on optimal particle position vector And worst particle position vector For the first group of The individual particles are updated, and the specific process is as follows: setting a random variable rand, wherein the range of the random variable rand is between (1, dim); If the random variable rand is not smaller than the decision coefficient DF, entering a step two, three or two; if the random number rand is smaller than the decision factor DF, the method proceeds to the second, third and third steps: step two, three and two according to the optimal particle position vector And worst particle position vector Using first update rules to first in population Updating the individual particles; step two, three and three according to the optimal particle position vector And worst particle position vector Using a second update rule to a first one of the populations The individual particles are updated.
9. The method for predicting parallelism of a gearbox based on NRBO-ELM algorithm as set forth in claim 8, wherein the step of performing two-three-two is based on an optimal particle position vector And worst particle position vector Using first update rules to first in population The individual particles are updated once, expressed as: in the formula, IT represents an iteration number variable, wherein IT is a positive integer; represents the IT+1st iteration time The position vector of each particle, namely the updated position vector; IT represents an iteration number variable; Representing a first intermediate position variable, Representing a second intermediate position variable, Representing a third intermediate position variable, expressed by the formula: In the formula, Represents the IT time of iteration The position vectors of the individual particles, a, b represent random variables, r 2 represent random variables, Represents a first intermediate variable, Represents a second intermediate variable, Representing a third intermediate variable, formulated as: in the formula, mean () represents an average value, A fourth intermediate variable is represented by the formula, Representing a fifth intermediate variable, formulated as: 。
10. The method for predicting parallelism of a gearbox based on NRBO-ELM algorithm as set forth in claim 9, wherein the steps two, three and three are based on optimal particle position vectors And worst particle position vector Using a second update rule to a first one of the populations The individual particles are updated, and the specific process is as follows: construction of random numbers Said random number In the range between (0, 1), When random number Less than 0.5, for the first population The updating of individual particles is formulated as: In the formula, 、、 Are random numbers between (0, 1); When random number When the ratio is not more than 0.5, for the first group of The updating of individual particles is formulated as: 。

Description

NRBO-ELM algorithm-based gearbox parallelism prediction method Technical Field The invention belongs to the field of monitoring, and particularly relates to a gearbox parallelism prediction method based on NRBO-ELM algorithm. Background Dynamic balance is used as a core technology for detecting precision of rotary machinery, and the main purpose of the dynamic balance is to reduce vibration and unbalanced force by adjusting mass distribution of a rotary part, so that stability, reliability and operation efficiency of equipment are improved. In the field of gear drives, in particular gearboxes, the axis parallelism has a direct and important influence on the dynamic balancing effect. Axis parallelism refers to the degree of deviation between the centerlines of the multiple axes. Ideally, the axes should be kept parallel, however, in the actual manufacturing and assembly process, due to factors such as machining precision, assembly precision, installation errors and the like, the parallelism of the axes often cannot completely meet the design requirements, and when the parallelism error of the axes is large, serious dynamic unbalance phenomenon is caused. The axis parallelism with higher precision is not only a technical requirement, but also a key factor for ensuring the normal operation and long-term stability of a mechanical system. Although axis parallelism is critical to the dynamic balance effect, its accurate measurement still faces some challenges. The traditional measuring method may be comprehensively influenced by various factors such as the precision of measuring equipment, the operation complexity, the time cost, the environmental conditions and the like, and the convenience and the high precision are often not compatible. Therefore, the establishment and application of the efficient and accurate axis parallelism prediction model become particularly important, the axis parallelism can be predicted and optimized in the design and manufacturing stages, the influence of the axis parallelism on dynamic balance can be accurately estimated, and the reliability and performance of a mechanical system can be remarkably improved. Disclosure of Invention The invention aims to solve the problem that the existing gearbox parallelism prediction method has low prediction precision. The invention discloses a gearbox parallelism prediction method based on NRBO-ELM algorithm. Comprising the following steps: 1. Constructing a gear box parallelism prediction model, and obtaining a trained gear box parallelism prediction model; 2. Acquiring stress distribution data of a gear box to be predicted, and inputting the stress distribution data of the gear box to be predicted into a trained gear box parallelism prediction model to obtain the parallelism of the gear box to be predicted; The method comprises the steps of constructing a gear box parallelism prediction model to obtain a trained gear box parallelism prediction model, wherein the specific process is as follows: step one, obtaining a training set; and training the constructed gear box parallelism prediction model according to a training set to obtain a trained gear box parallelism prediction model. Further, the specific process of obtaining the training set in the first step is as follows: the method comprises the steps of constructing a finite element model of the gear box, wherein the concrete process comprises the following steps: Extracting the shafting space position of each pair of meshing gears in the gear box, and setting the shafting space position of each pair of meshing gears in finite element software according to the extracted shafting space position of each pair of meshing gears in the gear box; step one, performing finite element simulation calculation according to a constructed gear box finite element model, and constructing a training set; the data of the training set is the distribution data of the bending stress of the tooth root of the gear box gear along the tooth width direction under the shafting deflection state The gear case parallelism prediction model in the second step is an Extreme learning machine model (Extreme LEARNING MACHINE, ELM) Training the constructed gear box parallelism prediction model according to a training set to obtain a trained gear box parallelism prediction model, wherein the specific process is as follows: step two, constructing an extreme learning machine model and determining internal parameters to be optimized in the extreme learning machine model; inputting normalized stress distribution data in a training set into an initialized extreme learning machine model, and outputting a prediction result by the extreme learning machine model; Thirdly, according to the input and output of the extreme learning machine model, carrying out optimization training on internal parameters to be optimized of the extreme learning machine model by using NRBO algorithm Newton-Lapherson iteration method and a trap escape algorithm, an