A method and system for predicting unbalance error of a dynamic balancing machine based on error correction

Abstract

The application provides a dynamic balancing machine weight removal integrated prediction method and system based on error correction, and belongs to the technical field of mechanical error analysis. The dynamic balancing machine weight removal integrated prediction method and system based on error correction comprises the following steps: obtaining operation data and brake disc data of a dynamic balancing machine; inputting the data into a trained error prediction model for prediction; adding the weight removal determined by a dynamic balancing machine preliminary inspection station and the model prediction result to obtain the final weight removal of the dynamic balancing machine weight removal system; wherein the weight removal determined by the dynamic balancing machine preliminary inspection station is obtained by the existing system of the dynamic balancing machine; and the historical operation data and brake disc data of the dynamic balancing machine are used as the training data set of the error prediction model. The application improves the weight removal precision of the dynamic balancing machine and the stability of the brake disc in the high-speed rotation process.

Description

Technical Field

[0001] This invention relates to the field of weight loss error prediction technology during the balancing process of brake disc imbalance, and particularly to a method and system for predicting weight loss error in dynamic balancing machines based on error correction. Background Technology

[0002] With the rapid development of the automotive industry, vehicle safety has attracted increasing attention. Among the many factors affecting vehicle safety, braking performance plays a significant role. In the development of automotive braking systems, brake configurations have continuously evolved, gradually shifting from the traditional "front disc, rear hub" design to a "front disc, rear disc" configuration. The use of brake discs has skyrocketed, and their quality severely impacts vehicle performance, driving safety, and reliability.

[0003] Due to factors such as material quality, processing technology, and processing environment during the production of brake discs, uneven mass distribution can occur within the brake disc itself. This uneven mass distribution causes vibrations during the high-speed rotation of the brake disc, which are hazardous vibrations that can seriously affect vehicle driving safety and the personal safety of occupants.

[0004] The root cause of this harmful vibration is that during vehicle operation, the brake disc rotates at high speed along with the wheels. Due to its uneven mass distribution, centrifugal force is generated, leading to an imbalance of forces within the brake disc and causing vibration. This phenomenon is called dynamic imbalance, and the uneven mass is called the imbalance amount. The problem of imbalance is mainly addressed through balancing detection and correction processes. Balance detection refers to detecting the amplitude and phase of the imbalance in the brake disc, while correction mainly refers to achieving balance by reducing weight after the imbalance detection is completed.

[0005] Currently, most research on improving the calibration accuracy of dynamic balancing machines focuses on improving the detection technology of imbalance, hoping to improve the balancing accuracy of the brake disc through more accurate imbalance detection technology. However, the milling weight removal system of the dynamic balancing machine is also affected by unavoidable factors such as insufficient or excessive tool feed and insufficient friction provided by the clamping mechanism during milling, which can cause the brake disc to slip. This is a problem that detection technology cannot solve. Summary of the Invention

[0006] To address the aforementioned issues, this invention proposes a method and system for predicting weight reduction errors in dynamic balancing machines based on error correction. The method utilizes a random forest model to predict weight reduction errors and compensates for milling errors that the detection system cannot make up for by adjusting the tool feed and milling angle based on the prediction results.

[0007] In the implementation method, the following technical solution is adopted:

[0008] A method for predicting the weight removal error of a dynamic balancing machine based on error correction, comprising:

[0009] The data acquisition module acquires the operating data of the dynamic balancing machine and the brake disc data;

[0010] The model prediction module inputs the above data into the trained correction deweighting error prediction model for prediction.

[0011] The data output module adds the predicted results output by the model to the weight determined by the initial inspection station of the dynamic balancing machine to obtain the final weight of the brake disc.

[0012] The training dataset for the weight reduction error prediction model consists of the historical operating data of the dynamic balancing machine and the brake disc data. The historical operating data of the dynamic balancing machine and the brake disc data are input into the weight reduction error prediction model, and the predicted weight reduction error value is output.

[0013] In other embodiments, the following technical solutions are adopted:

[0014] A terminal device includes a memory and a processor. The memory is used to store multiple instructions, and the processor is used to implement the instructions, which are adapted to be loaded and executed by the processor to perform the aforementioned error correction-based dynamic balancing machine weight reduction error prediction method.

[0015] In other embodiments, the following technical solutions are adopted:

[0016] A computer-readable storage medium for storing instructions, which are executed by a processor of a terminal device, of the aforementioned method for predicting the weight loss error of a dynamic balancing machine based on error correction.

[0017] Compared with the prior art, the advantages of the present invention are:

[0018] 1. This invention utilizes the random forest model in the ensemble learning algorithm to establish a deweighting error prediction model, which overcomes the shortcoming that the correction error caused by the deweighting station of the dynamic balancing machine cannot be compensated.

[0019] 2. This invention uses weight loss error prediction to correct the weight loss calculated at the dynamic balancing machine inspection station, thereby improving the weight loss accuracy of the dynamic balancing machine and enhancing the stability of the brake disc during high-speed rotation. Attached Figure Description

[0020] Figure 1 This is a flowchart of the deweighting error prediction process based on error correction in an embodiment of the present invention;

[0021] Figure 2This is a graph showing the correlation between the operating characteristics of the dynamic balancing machine, the parameter characteristics of the brake disc, and the weight removal error in an embodiment of the present invention.

[0022] Figure 3 This represents the overall model deweighting error prediction result in the embodiments of the present invention; Detailed Implementation

[0023] It should be noted that the following descriptions are exemplary and intended to provide further explanation of this application. Unless explicitly stated otherwise, all technical and scientific terms used in this invention have the same meaning as understood by one of ordinary skill in the art to which this application pertains.

[0024] Example 1

[0025] According to an embodiment of the present invention, a method for predicting the weight removal error of a dynamic balancing machine based on error correction is disclosed, comprising the following steps:

[0026] (1) Obtain real-time dynamic balancing machine operation data and brake disc data;

[0027] Specifically, dynamic balancing machine operating data and brake disc data refer to all data related to milling weight removal, including but not limited to: the dynamic balancing machine's milling power, milling depth, milling angle, maximum outer radius of the brake disc, density, and other data.

[0028] (2) Input the above data into the weight loss error prediction model for prediction;

[0029] Specifically, the system collects operating data and brake disc data determined by the dynamic balancing machine inspection station, and extracts data such as milling depth, milling angle, and maximum outer radius of the brake disc.

[0030] This example proposes a weight reduction error prediction model for dynamic balancing machines based on error correction. The method utilizes a random forest to establish the weight reduction error prediction model, and then adds the model's prediction results to the weight reduction determined at the dynamic balancing machine's inspection station to obtain the final weight reduction. This method improves the correction accuracy of the weight reduction station on the dynamic balancing machine and enhances the stability of the brake disc during high-speed rotation.

[0031] The detailed method is as follows:

[0032] (1) Data preparation

[0033] To determine the input to the prediction model, the correlation coefficient ρ is first used. xy Analyze the correlation between factors such as milling width, brake disc diameter, and weight removal, ρ xy The larger x is, the more x represents i and y i The stronger the correlation between two features, the higher ρ xy Defined as:

[0034]

[0035] in, and These are the average values ​​of the two features, and N is the number of test samples.

[0036] Table 1 shows the relationship between the correlation coefficient and the degree of correlation between two features. The degree of correlation between two features can be determined based on the correlation coefficient.

[0037] Table 1. Relationship between correlation coefficient and correlation strength

[0038]

[0039] Figure 2 The graph shows the relationship between four features—milling width, milling power, balancing time, and rotor diameter—and the weight removal error. It can be seen from the graph that milling width, milling power, and balancing time are strongly correlated. Therefore, this example selects three features—milling width, milling power, and balancing time—as module inputs.

[0040] (2) Deweighting error prediction model based on random forest

[0041] 1) Model Introduction

[0042] Random forest regression is a combined model consisting of a set of regression decision subtrees. Utilizing the idea of ​​ensemble learning, the mean of all decision subtrees is taken as the prediction result of the random forest regression. Its data expression is as follows:

[0043]

[0044] Where T represents the number of trees, x represents the independent variable, and θ t Let h(x,θ) be a random variable that follows an independent and identically distributed distribution. t ) is the output based on x and θ.

[0045] The Random Forest algorithm introduces the ideas of Bagging and random subspaces to overcome the shortcomings of decision trees, such as low accuracy and easy overfitting.

[0046] ①Bagging idea

[0047] Multiple samples are randomly selected with replacement from the original sample as training samples, and the size of each training sample is required to be equal to the size of the original sample. A decision subtree is constructed for each sample, and the average of the predictions of each decision subtree is taken as the final prediction result.

[0048] For a given sample, in a random sampling of N samples, the probability that each sample is not selected is:

[0049]

[0050] When N approaches infinity, we get...

[0051]

[0052] Therefore, in each round of sampling in Bagging, approximately 36.8% of the data is not collected; the uncollected data is referred to as out-of-bag data.

[0053] The Bagging idea not only allows for the randomization of more decision subtrees, but also ensures that each subtree is independent of the others.

[0054] ② The idea of ​​random subspaces

[0055] In the process of constructing the decision subtree, each split node of the decision subtree randomly extracts a feature subspace from the total feature space as a candidate feature set for the decision subtree split node, and selects the optimal feature from the candidate feature set for splitting.

[0056] This method ensures that the nodes between trees and the feature subsets between different nodes in each tree are different, and also guarantees the diversity and randomness of decision subtrees. It improves the randomness of node splitting in the random forest algorithm.

[0057] The algorithm flow of Random Forest is as follows:

[0058] Step 1: Using the Bagging concept, randomly generate a subset of samples;

[0059] Step 2: Using the idea of ​​random subspace, randomly select f features, split the decision subtree, and construct a single decision subtree;

[0060] Step 3: Repeat steps 1 and 2 to construct multiple decision subtrees, allowing each subtree to grow freely without pruning, thus forming a forest;

[0061] Step 4: Take the average of the predictions from multiple decision subtrees as the final prediction result.

[0062] Random forest is an ensemble learning model that uses single decision trees as basic classification predictors. It combines two major stochastic ideas: random subspaces and ensemble learning theory, to randomly select split attribute values ​​and training samples. It overcomes the overfitting problem of decision trees and is insensitive to noise and outliers, making it an effective prediction method.

[0063] 2) Model parameter optimization

[0064] The main performance parameters of the random forest model are three: the number of subtrees in the forest (tress_num), the maximum depth of the trees (max_depth), and the number of features randomly selected from each subtree (max_features). This embodiment will use a 10x grid search to optimize the model parameters of the random forest.

[0065] 3) The predicted results output by the model are added to the weight determined by the dynamic balancing machine inspection station to obtain the final weight.

[0066] Specifically, the operating parameters of the dynamic balancing machine and the brake disc are collected in real time, input into the random forest model, and the weight reduction error prediction result is obtained. Finally, the model prediction result is added to the weight reduction determined by the dynamic balancing machine inspection station to obtain the final weight reduction.

[0067] In this embodiment, the relative error (RE) for weight removal is defined as follows:

[0068]

[0069] in, To eliminate the actual weight error, P i This represents the weight prediction error.

[0070] The model prediction results are as follows:

[0071] This example uses a machine tool on a production line of a brake disc balancing company in Jilin Province to predict the weight reduction error of a brake disc about to undergo balancing. The dataset consists of historical balancing data from this production line machine tool in 2020-2021, which is randomly divided into training and test sets.

[0072] Figure 3 The results show the prediction of weight loss error for 30 brake discs using this dynamic balancing machine tool. The prediction results obtained by using the error correction-based dynamic balancing machine weight loss error prediction model proposed in this embodiment are close to the actual weight loss error. This result verifies that the prediction model proposed in this embodiment has good prediction performance.

[0073] In summary, the prediction model proposed in this embodiment can improve the calibration accuracy of the dynamic balancing machine. This embodiment fully utilizes ensemble learning to integrate multiple weakly supervised models to obtain a comprehensive strongly supervised model. The model prediction result is then added to the weight reduction determined at the dynamic balancing machine's inspection station to obtain the final weight reduction. Case studies illustrate the effectiveness of this prediction method. Analysis of the results shows that the model proposed in this embodiment can improve the calibration accuracy of the dynamic balancing machine and enhance the stability of the brake disc during high-speed rotation.

[0074] Example 2

[0075] According to an embodiment of the present invention, an embodiment of a terminal device is disclosed, which includes a memory and a processor. The memory is used to store instructions, and the processor is used to implement various instructions, which are adapted to be loaded and executed by the processor to perform the error correction-based dynamic balancing machine weight reduction error prediction method described in Embodiment 1.

[0076] The above examples are only used to illustrate the technical solutions of this application, and are not intended to limit them; the above examples are used to help understand the method and core ideas of this invention; at the same time, those skilled in the art will have certain modifications and substitutions in the specific implementation methods and application scope; and these substitutions or modifications do not cause the essence of the corresponding technical solutions to deviate from the protection scope of the technical solutions of the embodiments of this invention.

Claims

1. A method for predicting unbalance error of an error correction based balancing machine, characterized in that, The application relates to a dynamic balancing machine and a dynamic balancing method. First, operation data and brake disc data of a dynamic balancing machine are acquired, then the data are input into a trained weight error prediction model for prediction, and finally, the prediction result of the model is added to the weight error determined by the dynamic balancing machine initial inspection station as the final weight error. The weight error prediction model is constructed by using a random forest model, and a training data set is historical operation data and brake disc data of the dynamic balancing machine; the historical operation data and brake disc data of the dynamic balancing machine are input into the weight error prediction model, and a weight error prediction value is output.

2. A method for predicting the unbalance error of an error correction based balancing machine according to claim 1, characterized in that, When the weight error prediction model is constructed, binary coding is adopted, the number of model trees, the maximum depth of the trees and the number of randomly selected features of each decision tree are selected as parameters, and cross validation is adopted to find the optimal combination of the three parameters of the random forest model.

3. A terminal device comprising a memory and a processor, the memory to store instructions: the processor to implement the various stored instructions, characterized in that, The instructions are loaded and executed by the processor to implement the dynamic balancing machine weight error prediction method based on error correction according to any one of claims 1-2.

4. A computer-readable storage medium storing instructions, wherein the instructions, when executed by a processor, cause the processor to perform operations comprising: The instructions are executed by the processor of the terminal device to implement the dynamic balancing machine weight error prediction method based on error correction according to any one of claims 1-2.

Images