Gear service life prediction method and system based on big data

Abstract

The invention discloses a gear service life prediction method and system based on big data. The method comprises the steps of data acquisition, data set division, gear service life prediction model establishment, parameter optimization and gear service life prediction. The invention belongs to the field of data processing, and particularly relates to a gear service life prediction method and system based on big data, according to the scheme, a dynamic weight updating mechanism based on prediction residual errors is introduced, and high sensitivity to abnormal conditions is ensured; a smooth weight adjustment function is introduced for the conditions of abnormally short life and abnormally long life, so that the risk of missed judgment is reduced; adding prediction uncertainty penalty into the abnormal region loss item; the accuracy of gear life prediction is improved; through chaotic mapping initialization, multi-stage iteration and double-strategy updating, adaptive evolution rate and difference measurement are carried out; the method ensures that the requirement of efficient response of industrial monitoring is met in a complex industrial environment with more noise interference, and the effect of predicting the service life of the gear is improved.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and in particular to a gear life prediction method and system based on big data. Background Art

[0002] Gear life prediction methods use a series of technical means, such as data acquisition, signal processing, feature extraction, and modeling, to analyze the wear, fatigue, and damage of gears during operation, thereby predicting their service life in advance. However, general gear life prediction methods lack a dynamic feedback adjustment mechanism and are unable to distinguish the life levels of gears under different conditions. In the face of changes in gear data distribution or sparse abnormal samples, the ability to identify key gear states is poor, resulting in low gear life prediction accuracy. General gear life prediction methods cannot ensure the diversity and global coverage of parameter searches under complex gear operating conditions, and cannot respond in real time to changes in data distribution and operating conditions in the production environment, resulting in poor gear life prediction results. Summary of the Invention

[0003] In view of the above situation, in order to overcome the defects of the prior art, the present invention provides a gear life prediction method and system based on big data. In view of the fact that general gear life prediction methods lack a dynamic feedback adjustment mechanism and cannot distinguish the life levels of gears under different states, and have poor recognition capabilities for key gear states in the face of changes in gear data distribution or scarcity of abnormal samples, which leads to low accuracy in gear life prediction, this solution introduces a dynamic weight update mechanism based on prediction residuals to adapt to real-time data changes under different working conditions, avoid dilution of key abnormal data, and ensure high sensitivity to abnormal conditions; for abnormally low life and abnormally high life, a smooth weight adjustment function is introduced to reduce the risk of missed judgment; a prediction uncertainty penalty is added to the abnormal area loss term to enhance the uncertainty under high-risk working conditions. factors; thereby improving the accuracy of gear life prediction; in view of the problem that general gear life prediction methods cannot ensure the diversity and global coverage of parameter search under complex gear working conditions, and cannot respond to changes in data distribution and working conditions in the production environment in real time, which leads to poor gear life prediction results, this scheme ensures the uniform distribution of population individuals in the hyperparameter space through chaos map initialization, laying a good foundation for global search; through multi-stage iteration and dual-strategy update, a dynamic balance between global exploration and local fine search is achieved; through adaptive evolution rate and difference measurement, the search step size and direction can be adjusted in real time; the quantum mutation strategy achieves escape from local optimality through probability transition; ensures that in complex industrial environments with more noise interference, it adapts to the requirements of efficient response of industrial monitoring and improves the effect of gear life prediction.

[0004] The technical solution adopted by the present invention is as follows: The present invention provides a gear life prediction method based on big data, which includes the following steps:

[0005] Step S1: data collection;

[0006] Step S2: data set division;

[0007] Step S3: establishing a gear life prediction model;

[0008] Step S4: parameter optimization;

[0009] Step S5: Gear life prediction.

[0010] Furthermore, in step S1, the data collection is to collect historical gear life monitoring data; the historical gear life monitoring data includes physical quantity monitoring data, operating condition record data and real life; and the real life is used as a data label.

[0011] Furthermore, in step S2, the data set division is to perform feature engineering and normalization processing on the collected historical gear life monitoring data; and divide the processed data into a test set and a training set.

[0012] Furthermore, in step S3, the gear life prediction model is established based on the test set and the training set by using a neural network; specifically, the following steps are included:

[0013] Step S31: Model architecture design; the overall model uses a multi-layer feedforward neural network, including: an input layer that accepts feature vectors of the test set and the training set; a hidden layer that uses four fully connected layers, each followed by a ReLU activation function and introduces Dropout; an output layer, the last layer is a fully connected layer, whose output is a scalar representing the predicted gear life;

[0014] Step S32: Establish a weight mechanism; for gear life data, define a normal life region; define a low life region; define an abnormally long life region; calculate weights; and introduce an error feedback mechanism. After each training cycle, calculate an error index based on the prediction residual of the abnormal region samples to perform dynamic weight updates.

[0015] Step S33: defining a smooth weight adjustment function; introducing an importance factor and a smooth weight adjustment function to impose different penalties on asymmetric situations where the prediction is too low or too high;

[0016] Step S34: constructing an overall loss function; dividing the entire sample space into three parts according to the actual life level of the gear: abnormally low life, normal life, and abnormally high life, and designing loss terms for each part;

[0017] Step S35: model training; update model parameters based on gradient descent method; set the maximum number of training times, when the loss based on the training set converges or reaches the maximum number of training times, the gear life prediction model training is completed; set the prediction threshold, when the prediction accuracy of the trained gear life prediction model for the test set is higher than the prediction threshold, the gear life prediction model is established; otherwise, perform parameter optimization.

[0018] Furthermore, in step S4, the parameter optimization specifically includes the following steps:

[0019] Step S41: establishing a search space based on the hyperparameters of the gear life prediction model, generating a chaotic sequence using piecewise linear chaotic mapping, and mapping the chaotic sequence to each dimensional interval of the search space as the position initialization of the population individuals;

[0020] Step S42: Divide the total number of iterations into five stages; and use the prediction accuracy of the gear life prediction model trained based on the individual position for the test set as the individual fitness value; assume that the optimal solution is the individual with the highest fitness value;

[0021] Step S43: Position update in the exploration phase; different update formulas are used in the two exploration sub-phases;

[0022] Step S44: defining an exploration operator; adjusting the exploration step size based on the exploration operator;

[0023] Step S45: defining a reduction operator; dynamically reducing the search range by introducing a reduction operator control parameter;

[0024] Step S46: defining an adaptive evolution rate; using a randomly decaying adaptive evolution rate to control the exploration intensity;

[0025] Step S47: defining a difference measurement coefficient; measuring the difference between the current solution and the optimal solution by introducing the difference measurement coefficient;

[0026] Step S48: Development phase position update; in the two development sub-phases, two different update strategies are adopted;

[0027] Step S49: Probabilistic jump mutation strategy; In the last stage, quantum mutation based on Gaussian distribution and Monte Carlo method is introduced;

[0028] Step S410: Iteration determination; when there is an individual whose fitness value is higher than the prediction threshold, the iteration ends; otherwise, if the maximum number of iterations is reached, the population individuals are reinitialized; otherwise, the iteration continues.

[0029] Furthermore, in step S5, the gear life prediction is based on the established gear life prediction model, and physical quantity monitoring data and operating condition recording data are collected in real time. After preprocessing, they are input into the gear life prediction model, and the actual life output by the model is used as the gear life prediction result.

[0030] The gear life prediction system based on big data provided by the present invention includes a data acquisition module, a data set division module, a gear life prediction model establishment module, a parameter optimization module and a gear life prediction module;

[0031] The data acquisition module collects historical gear life monitoring data;

[0032] The data set partitioning module performs feature engineering and normalization on the collected data, and then divides it into a training set and a test set;

[0033] The gear life prediction model establishment module constructs an overall loss function based on the training set and the test set, divides the sample space into abnormally low life, normal life and abnormally high life areas, and applies dynamic weights, smoothing weight adjustment functions and uncertainty coefficients to establish a gear life prediction model;

[0034] The parameter optimization module optimizes the parameters of the gear life prediction model using piecewise linear chaotic mapping and multi-stage iterative strategy;

[0035] The gear life prediction module predicts the gear life based on the gear life prediction model based on the real-time collected physical quantity monitoring data and operating condition record data.

[0036] The beneficial effects achieved by the present invention using the above scheme are as follows:

[0037] (1) In view of the fact that general gear life prediction methods lack a dynamic feedback adjustment mechanism and cannot distinguish the life levels of gears under different states, and have poor recognition capabilities for key gear states when faced with changes in gear data distribution or scarcity of abnormal samples, which leads to low accuracy in gear life prediction, this scheme introduces a dynamic weight update mechanism based on prediction residuals to adapt to real-time data changes under different working conditions, avoid dilution of key abnormal data, and ensure high sensitivity to abnormal conditions; for abnormally low life and abnormally high life, a smooth weight adjustment function is introduced to reduce the risk of missed judgment; a prediction uncertainty penalty is added to the abnormal area loss term to enhance the ability to handle uncertainty factors under high-risk working conditions; thereby improving the accuracy of gear life prediction.

[0038] (2) In view of the problem that general gear life prediction methods cannot ensure the diversity and global coverage of parameter search under complex gear working conditions, and cannot respond to changes in data distribution and working conditions in the production environment in real time, which leads to poor gear life prediction results, this scheme ensures the uniform distribution of population individuals in the hyperparameter space through chaos mapping initialization, laying a good foundation for global search; achieves a dynamic balance between global exploration and local fine search through multi-stage iteration and dual strategy update; enables real-time adjustment of search step size and direction through adaptive evolution rate and difference measurement; quantum mutation strategy achieves escape from local optimality through probability transition; ensures that in complex industrial environments with more noise interference, it adapts to the requirements of efficient response of industrial monitoring and improves the effect of gear life prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 A schematic diagram of the process of the gear life prediction method based on big data provided by the present invention;

[0040] Figure 2 A schematic diagram of a gear life prediction system based on big data provided by the present invention;

[0041] Figure 3 Schematic diagram of the process of step S3.

[0042] The accompanying drawings are used to provide further understanding of the present invention and constitute a part of the specification. They are used to explain the present invention together with the embodiments of the present invention and do not constitute a limitation of the present invention. DETAILED DESCRIPTION

[0043] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments; based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0044] In the description of the present invention, it should be understood that terms such as "up", "down", "front", "back", "left", "right", "top", "bottom", "inside" and "outside" indicating directions or positional relationships are based on the directions or positional relationships shown in the accompanying drawings. They are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the system or element referred to must have a specific direction, be constructed and operated in a specific direction. Therefore, they should not be understood as limiting the present invention.

[0045] Example 1, see Figure 1 The present invention provides a gear life prediction method based on big data, which includes the following steps:

[0046] Step S1: Data collection: collecting historical gear life monitoring data;

[0047] Step S2: Dataset division: feature engineering and normalization are performed on the collected data, and then divided into training set and test set;

[0048] Step S3: Establishing a gear life prediction model; based on the training set and the test set, constructing an overall loss function, dividing the sample space into abnormally low life, normal life, and abnormally high life regions, and applying dynamic weights, smoothing weight adjustment functions, and uncertainty coefficients to establish a gear life prediction model;

[0049] Step S4: parameter optimization; using piecewise linear chaotic mapping and multi-stage iterative strategy to optimize the parameters of the gear life prediction model;

[0050] Step S5: Gear life prediction: Gear life prediction is performed based on the gear life prediction model based on the real-time collected physical quantity monitoring data and operating condition record data.

[0051] Example 2, see Figure 1 This embodiment is based on the above embodiment. In step S1, the historical gear life monitoring data includes physical quantity monitoring data, operating condition record data and real life; the physical quantity monitoring data includes vibration, temperature, torque and speed; the operating condition record data includes load size, torque, speed, lubrication status and ambient temperature; and the real life is used as a data label.

[0052] Example 3, see Figure 1 This embodiment is based on the above embodiment. In step S2, the data set division is to perform feature engineering and normalization processing on the collected historical gear life monitoring data; and divide the processed data into a test set and a training set.

[0053] Example 4, see Figure 1 and Figure 3 This embodiment is based on the above embodiment. In step S3, the gear life prediction model is established based on the test set and the training set, and the gear life prediction model is established using a neural network. Specifically, the following steps are included:

[0054] Step S31: Model architecture design; the overall model uses a multi-layer feedforward neural network, including: an input layer that accepts feature vectors of the test set and the training set; a hidden layer that uses four fully connected layers, each followed by a ReLU activation function and introduces Dropout; an output layer, the last layer is a fully connected layer, whose output is a scalar representing the predicted gear life;

[0055] Step S32: Establish a weighting mechanism; in the gear life prediction, the samples in the data set that are in abnormal life status often account for a relatively low proportion. By introducing a larger weight to these samples, it is ensured that they will not be diluted in the training, thereby increasing the sensitivity to abnormal conditions; for the gear life data, define the normal life area for: ;in, is the feature vector of the i-th sample; is the actual life of the gear corresponding to the sample; and Is the lifespan determination threshold; defines the low lifespan area for: ; Define areas with unusually long lifespans for: ; Calculate the weight, expressed as: ; ;in, is the sample size; and are the amplification weights for samples with low lifespan and samples with abnormally long lifespan, respectively. An error feedback mechanism is introduced. After each training cycle, an error index is calculated based on the prediction residual of samples in the abnormal area, and the weight is dynamically updated, which is expressed as: ; ; and are the amplification weights during the T-th training; is the amplification adjustment coefficient; is the mean residual in the normal region; and are the mean residuals of the areas with low life expectancy and the areas with abnormally long life expectancy, respectively; is a smoothing term;

[0056] Step S33: Define a smooth weight adjustment function. In actual gear life prediction, the deviation between the predicted value and the true value should be more carefully distinguished. For gears in the abnormally low life range, if the model prediction value is too high, it may lead to missed premature failure risk. Similarly, in the abnormally high life range, if the prediction is too low, it will also lead to underestimated risk. The importance factor a and the smooth weight adjustment function are introduced to impose different penalties for asymmetric low or high predictions to capture key abnormalities.

[0057] for , ;

[0058] for , ;in, The model input The predicted lifespan value; are the current parameters of the model; and is a smooth weight adjustment function;

[0059] Step S34: Construct an overall loss function; divide the entire sample space into three parts according to the actual life level of the gear: abnormally low life, normal life, and abnormally high life, design loss terms for each part, and give higher penalties to key abnormal conditions, thereby improving the overall prediction effect; the overall loss function J is defined as ; ; ; ;in, 、 and They are the losses of samples with low life span, normal samples and samples with abnormally long life span; is the uncertainty coefficient; It means that the sample The confidence index of the predicted value distribution obtained after multiple forward propagations;

[0060] Step S35: model training; update model parameters based on gradient descent method; set the maximum number of training times, when the loss based on the training set converges or reaches the maximum number of training times, the gear life prediction model training is completed; set the prediction threshold, when the prediction accuracy of the trained gear life prediction model for the test set is higher than the prediction threshold, the gear life prediction model is established; otherwise, perform parameter optimization.

[0061] By performing the above operations, the general gear life prediction method lacks a dynamic feedback adjustment mechanism, cannot distinguish the life levels of gears under different states, and has poor recognition ability of key gear states in the face of changes in gear data distribution or scarce abnormal samples, which leads to low accuracy of gear life prediction. This scheme introduces a dynamic weight update mechanism based on prediction residuals to adapt to real-time data changes under different working conditions, avoid the dilution of key abnormal data, and ensure high sensitivity to abnormal conditions; for abnormally low life and abnormally high life, a smooth weight adjustment function is introduced to reduce the risk of missed judgment; a prediction uncertainty penalty is added to the abnormal area loss term to enhance the ability to handle uncertainty factors under high-risk working conditions; thereby improving the accuracy of gear life prediction.

[0062] Example 5, see Figure 1 This embodiment is based on the above embodiment. In step S4, parameter optimization specifically includes the following steps:

[0063] Step S41: Establish a search space based on the hyperparameters of the gear life prediction model, use piecewise linear chaotic mapping to generate a chaotic sequence in [0,1], and map the chaotic sequence to each dimensional interval of the search space as the position initialization of the population individual; expressed as: ;

[0064] Among them, p is the control parameter; is the chaotic sequence value; n is the number of mapping steps; the generated chaotic sequence value is linearly mapped to the value interval of each variable in the optimization problem according to the component, which is expressed as: ; is the position of the jth dimension of the uth individual; 0 refers to initialization; and are the upper and lower limits of the j-th dimension respectively;

[0065] Step S42: Divide the total number of iterations into five stages, alternating with different strategies to solve the problems of premature maturity and stage imbalance; and use the prediction accuracy of the gear life prediction model trained based on the individual position on the test set as the individual fitness value; assume that the optimal solution is the individual with the highest fitness value;

[0066] Step S43: Position update in the exploration phase; in the two exploration sub-phases, different update formulas are used to take into account both global and local searches, expressed as: ;

[0067] Where t is the number of iterations; is the position of the jth dimension of the optimal solution; is the exploration operator; is the scaling factor; is the reduction operator; rand is a random number between 0 and 1; Tr is the maximum number of iterations; is the position of the random individual in the jth dimension; is the adaptive evolution rate;

[0068] Step S44: define an exploration operator; adjust the exploration step size based on the exploration operator to enhance the search capability for the area around the current optimal solution, expressed as: ;in, is the coefficient of difference measurement;

[0069] Step S45: Define the reduction operator; by introducing the reduction operator control parameter, the search range is dynamically reduced to prevent excessive step length from crossing the boundary or skipping the optimal area, which is expressed as: ;in, is the position of another random individual in the jth dimension; is a smoothing term;

[0070] Step S46: Define the adaptive evolution rate; use a randomly decaying adaptive evolution rate to control the exploration intensity and enhance the adaptive ability, expressed as: ;in, It is a random number between 0 and 1, independent of rand;

[0071] Step S47: Define a difference metric coefficient. The difference metric coefficient is introduced to measure the difference between the current solution and the optimal solution, and is used to dynamically adjust the search step size and direction, which is expressed as: ;in, is a regulating factor; is the mean of all dimensions of the u-th individual;

[0072] Step S48: Development phase position update; in the two development sub-phases, two different update strategies are adopted to balance fast approximation and fine search, expressed as: ;

[0073] Step S49: Probabilistic jump mutation strategy; In the last stage, quantum mutation based on Gaussian distribution and Monte Carlo method is introduced to further jump out of the local optimum and conduct a fine search, which is expressed as: ; ;

[0074] in, It is the quantum center point; and is an independent uniform random number in the range of 0 to 1; is a random number between 0 and 1, independent;

[0075] Step S410: Iteration determination; when there is an individual whose fitness value is higher than the prediction threshold, the iteration ends; otherwise, if the maximum number of iterations is reached, the population individuals are reinitialized; otherwise, the iteration continues.

[0076] By performing the above operations, the general gear life prediction method cannot ensure the diversity and global coverage of parameter search under complex gear working conditions, and cannot respond in real time to changes in data distribution and working conditions in the production environment, which leads to poor gear life prediction effect. This scheme ensures the uniform distribution of population individuals in the hyperparameter space through chaotic map initialization, laying a good foundation for global search; achieves a dynamic balance between global exploration and local fine search through multi-stage iteration and dual strategy update; enables real-time adjustment of search step size and direction through adaptive evolution rate and difference measurement; quantum mutation strategy achieves escape from local optimality through probability transition; ensures that in complex industrial environments with more noise interference, it adapts to the requirements of efficient response of industrial monitoring and improves the effect of gear life prediction.

[0077] Example 6, see Figure 1This embodiment is based on the above embodiment. In step S5, the gear life prediction is based on the established gear life prediction model, and physical quantity monitoring data and operating condition record data are collected in real time. After preprocessing, they are input into the gear life prediction model, and the actual life output by the model is used as the gear life prediction result.

[0078] Example 7, see Figure 2 This embodiment is based on the above embodiment. The gear life prediction system based on big data provided by the present invention includes a data acquisition module, a data set division module, a gear life prediction model establishment module, a parameter optimization module and a gear life prediction module;

[0079] The data acquisition module collects historical gear life monitoring data;

[0080] The data set partitioning module performs feature engineering and normalization on the collected data, and then divides it into a training set and a test set;

[0081] The gear life prediction model establishment module constructs an overall loss function based on the training set and the test set, divides the sample space into abnormally low life, normal life and abnormally high life areas, and applies dynamic weights, smoothing weight adjustment functions and uncertainty coefficients to establish a gear life prediction model;

[0082] The parameter optimization module optimizes the parameters of the gear life prediction model using piecewise linear chaotic mapping and multi-stage iterative strategy;

[0083] The gear life prediction module predicts the gear life based on the gear life prediction model based on the real-time collected physical quantity monitoring data and operating condition record data.

[0084] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that includes a list of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus.

[0085] While the embodiments of the present invention have been shown and described, it will be apparent to those skilled in the art that various changes, modifications, substitutions, and alterations can be made to the embodiments without departing from the principles and spirit of the invention.

[0086] The present invention and its embodiments are described above. This description is not restrictive. The drawings show only one embodiment of the present invention, and the actual structure is not limited thereto. In short, if a person skilled in the art is inspired by this and, without departing from the purpose of the present invention, designs structures and embodiments similar to this technical solution without inventiveness, they shall fall within the scope of protection of the present invention.

Claims

1. A gear life prediction method based on big data, characterized by: The method comprises the following steps: Step S1: Data collection: collecting historical gear life monitoring data; Step S2: Dataset division: feature engineering and normalization are performed on the collected data, and then divided into training set and test set; Step S3: Establishing a gear life prediction model; based on the training set and the test set, constructing an overall loss function, dividing the sample space into abnormally low life, normal life, and abnormally high life regions, and applying dynamic weights, smoothing weight adjustment functions, and uncertainty coefficients to establish a gear life prediction model; Step S4: parameter optimization; using piecewise linear chaotic mapping and multi-stage iterative strategy to optimize the parameters of the gear life prediction model; Step S5: Gear life prediction: Gear life prediction is performed based on the gear life prediction model based on the real-time collected physical quantity monitoring data and operating condition record data.

2. The gear life prediction method based on big data according to claim 1 is characterized by: In step S3, the gear life prediction model is established based on the test set and the training set, and the gear life prediction model is established using a neural network; The specific steps include: Step S31: Model architecture design; the overall model uses a multi-layer feedforward neural network, including: an input layer that accepts feature vectors of the test set and the training set; a hidden layer that uses four fully connected layers, each followed by a ReLU activation function and introduces Dropout; an output layer, the last layer is a fully connected layer, whose output is a scalar representing the predicted gear life; Step S32: establishing a weighting mechanism; Step S33: defining a smooth weight adjustment function; introducing an importance factor and a smooth weight adjustment function to impose different penalties on asymmetric situations where the prediction is too low or too high; Step S34: constructing an overall loss function; dividing the entire sample space into three parts according to the actual life level of the gear: abnormally low life, normal life, and abnormally high life, and designing loss terms for each part; Step S35: model training; update model parameters based on gradient descent method; set the maximum number of training times, when the loss based on the training set converges or reaches the maximum number of training times, the gear life prediction model training is completed; set the prediction threshold, when the prediction accuracy of the trained gear life prediction model for the test set is higher than the prediction threshold, the gear life prediction model is established; otherwise, perform parameter optimization.

3. The gear life prediction method based on big data according to claim 2 is characterized by: In step S32, the weighting mechanism is to define a normal life area, a low life area, and an abnormally long life area for the gear life data; and calculate the weights. An error feedback mechanism is introduced. After each training cycle, an error index is calculated based on the prediction residual of the abnormal area samples to perform dynamic weight updates.

4. The gear life prediction method based on big data according to claim 3 is characterized by: In step S4, the parameter optimization specifically includes the following steps: Step S41: establishing a search space based on the hyperparameters of the gear life prediction model, generating a chaotic sequence using piecewise linear chaotic mapping, and mapping the chaotic sequence to each dimensional interval of the search space as the position initialization of the population individuals; Step S42: Divide the total number of iterations into five stages; and use the prediction accuracy of the gear life prediction model trained based on the individual position for the test set as the individual fitness value; assume that the optimal solution is the individual with the highest fitness value; Step S43: Position update in the exploration phase; different update formulas are used in the two exploration sub-phases; Step S44: defining an exploration operator; adjusting the exploration step size based on the exploration operator; Step S45: defining a reduction operator; dynamically reducing the search range by introducing a reduction operator control parameter; Step S46: defining an adaptive evolution rate; using a randomly decaying adaptive evolution rate to control the exploration intensity; Step S47: defining a difference measurement coefficient; measuring the difference between the current solution and the optimal solution by introducing the difference measurement coefficient; Step S48: Development phase position update; in the two development sub-phases, two different update strategies are adopted; Step S49: Probabilistic jump mutation strategy; In the last stage, quantum mutation based on Gaussian distribution and Monte Carlo method is introduced; Step S410: Iteration determination; when there is an individual whose fitness value is higher than the prediction threshold, the iteration ends; otherwise, if the maximum number of iterations is reached, the population individuals are reinitialized; otherwise, the iteration continues.

5. The gear life prediction method based on big data according to claim 4 is characterized in that: In step S1, the historical gear life monitoring data includes physical quantity monitoring data, operating condition record data and real life; the real life is used as a data label.

6. The gear life prediction method based on big data according to claim 5, characterized in that: In step S5, the gear life prediction is based on the established gear life prediction model, and physical quantity monitoring data and operating condition recording data are collected in real time. After preprocessing, they are input into the gear life prediction model, and the actual life output by the model is used as the gear life prediction result.

7. The gear life prediction method based on big data according to claim 6, characterized in that: In step S2, the data set division is to perform feature engineering and normalization processing on the collected historical gear life monitoring data; and divide the processed data into a test set and a training set.

8. A gear life prediction system based on big data, for implementing the gear life prediction method based on big data as described in any one of claims 1 to 7, characterized in that: It includes data acquisition module, data set division module, gear life prediction model establishment module, parameter optimization module and gear life prediction module; The data acquisition module collects historical gear life monitoring data; The data set partitioning module performs feature engineering and normalization on the collected data, and then divides it into a training set and a test set; The gear life prediction model establishment module constructs an overall loss function based on the training set and the test set, divides the sample space into abnormally low life, normal life and abnormally high life areas, and applies dynamic weights, smoothing weight adjustment functions and uncertainty coefficients to establish a gear life prediction model; The parameter optimization module optimizes the parameters of the gear life prediction model using piecewise linear chaotic mapping and multi-stage iterative strategy; The gear life prediction module predicts the gear life based on the gear life prediction model based on the real-time collected physical quantity monitoring data and operating condition record data.

Images