Data-driven brake disc automatic production line process parameter optimization method and system

Abstract

The invention discloses a data-driven brake disc automatic production line process parameter optimization method and system, and relates to the technical field of production line machining. Firstly, production data of a brake disc automatic production line are collected in real time; then, main components in the production data are analyzed through a main component analysis method and an orthogonal analysis method, so that the scale of the production data is reduced; and then, a full-connection neural network model is utilized to establish a relationship among the production line process parameters, the product quality and the production efficiency. And finally, obtaining an optimal process parameter combination through a particle swarm optimization algorithm, and feeding back the optimal process parameter combination to a control system of the production line by using a data system. Therefore, the problems that in an existing brake disc production line, brake disc machining quality fluctuation is large, technological parameter adjustment lags behind, timeliness is poor, and expert experience is relied on are solved. The invention provides a software-level optimization solution for improving the processing quality and the production efficiency of the automatic brake disc production line.

Description

Technical Field

[0001] The present invention relates to the field of production line processing technology, and in particular to a data-driven brake disc automatic production line process parameter optimization method and system. Background Art

[0002] With the rapid development of the automobile industry, the production quality of brake discs, as the core components of automobile braking systems, is directly related to the safety and market competitiveness of automobiles. The production process of the automatic production line for brake discs is mainly divided into seven steps: 1. Rough turning of the outer circle and end face; 2. Semi-finishing turning of the outer circle and end face; 3. Finishing turning of the outer circle and end face; 4. Milling of end face grooves; 5. Drilling; 6. Deburring; 7. Inspection and warehousing. In the first four steps, the process parameters of the production line have a great influence on the production quality of brake discs, including dimensional tolerance, surface roughness and dynamic balance performance. Therefore, the optimization of process parameters is crucial to improving the production quality and efficiency of the automatic production line for brake discs.

[0003] For traditional brake disc production lines, operators are required to continuously evaluate the quality of the produced brake discs and adjust the process parameters. This not only has poor real-time performance, but also makes it difficult to find the optimal combination of process parameters, resulting in random errors and systematic errors in the brake disc production process, affecting the production quality of the brake disc.

[0004] Therefore, a data-driven brake disc automatic production line process parameter optimization method and system are needed to solve the above problems. Summary of the invention

[0005] In view of the shortcomings of the prior art, the present invention discloses a data-driven brake disc automatic production line process parameter optimization method and system, which comprehensively analyzes the product quality information of the brake disc production line through data analysis technology, and realizes real-time and dynamic adjustment of process parameters. Thus, the impact of random errors and system errors on production quality during the production process is reduced, and the production efficiency of the brake disc production line is improved. The degree of automation and intelligence is high.

[0006] The present invention adopts the following technical solutions:

[0007] A data-driven brake disc automatic production line process parameter optimization method comprises the following steps:

[0008] S1. Collect the production process data of the brake disc, which at least includes the brake disc tool processing parameters, environmental parameters, tool wear parameters and the quality parameters of the brake disc after processing;

[0009] S2. Use principal component analysis to map the collected high-dimensional production process data into a low-dimensional space, and preliminarily screen the parameters that affect the machining quality of the brake disc and tool wear through mixed level orthogonal analysis;

[0010] S3. Evaluate the linear correlation between the selected parameters and the brake disc processing quality parameters by calculating the Pearson coefficient, and perform secondary screening on the parameters that affect the brake disc processing quality and tool wear based on the calculation results;

[0011] S4. Establishing a nonlinear relationship model between production line process parameters and product quality, production efficiency and tool wear based on a fully connected neural network model, wherein the fully connected neural network model is trained and iteratively optimized based on a training set to obtain a combination result of production line process parameters to be optimized;

[0012] S5. Normalize the process parameter combination results to be optimized, and input the normalized data into the particle swarm optimization algorithm for iterative optimization to obtain the optimal production line process parameter combination;

[0013] S6. Apply the optimal production line process parameter combination in real time to the production process of the brake disc automatic production line, and perform adaptive adjustments through the closed-loop management mechanism of the entire process to achieve quality control of the brake disc production process.

[0014] Further, in S3, the parameters affecting the machining quality of the brake disc preliminarily screened out in S2 are taken as the object, and the Pearson coefficient of the preliminarily screened parameters is calculated in combination with the parameter levels selected by the mixed level orthogonal analysis to evaluate the correlation between the preliminarily screened parameters and the machining quality parameters of the brake disc. The output function expression of the Pearson coefficient is:

[0015]

[0016] In formula (1), X i and Y i Indicates the X-axis coordinate and Y-axis coordinate of the i-th data point in the scatter plot in the preliminary screening parameter data set. and are the mean X-axis coordinates and the mean Y-axis coordinates of all data points of the preliminary screening parameters in the scatter plot, and the symbol ∑ represents the sum of all data points.

[0017] Furthermore, the fully connected neural network model uses the parameters screened out secondary as input, and the brake disc processing quality and tool wear as output, and predicts the product quality, production efficiency and tool wear under different process parameter combinations by adjusting the weights and biases of the neural network.

[0018] Furthermore, in S2, the data set obtained by principal component analysis is divided into a test set and a training set. The fully connected neural network model initializes weights and biases based on the training set data, and automatically calculates the model loss function after each training. The output formula of the loss function is:

[0019]

[0020] In formula (2), y is the actual target value, is the predicted value of the fully connected neural network model, n is the number of samples, and the parameters of the fully connected neural network model are updated by calculating the change rate of the loss function in the model parameter space.

[0021] Furthermore, the working method of the particle swarm optimization algorithm comprises the following steps:

[0022] The process parameter combination of the production line to be optimized is taken as a particle, and the maximum number of iterations, the number of particles, the inertia weight, the learning factor, the initial position and speed of the particles are set;

[0023] The fitness value of each particle is calculated according to the prediction results of the fully connected neural network model, and the individual optimal position and group optimal position of the particle are updated according to the fitness calculation results, and then the historical optimal position of each particle and the historical optimal position fitness value of the particle group are recorded;

[0024] Update the position and speed of each particle based on the speed and position of the historical optimal position and the current position, then recalculate the fitness value of each particle according to the updated particle position, and update the historical optimal process parameter combination and the historical optimal position and fitness value of the entire group;

[0025] Determine whether the maximum number of iterations or the preset convergence degree has been reached. If the maximum number of iterations or the fitness threshold has been reached, the historical optimal process parameter combination is output as the final optimization result; otherwise, return to calculate the fitness value of each particle;

[0026] Output the optimal production line process parameter combination as the final optimization result.

[0027] Furthermore, the normalization process in S5 is expressed as:

[0028]

[0029] In formula (3), X normalized represents the normalized process parameter combination to be optimized, X represents the original process parameter combination to be optimized, μ represents the mean of the process parameter combination data set to be optimized, and σ represents the standard deviation of the process parameter combination data set to be optimized.

[0030] Furthermore, a data-driven brake disc automatic production line process parameter optimization system is provided, the brake disc automatic production line process parameter optimization system comprising a data system and a data communication acquisition device, the data system is bidirectionally connected to the data communication acquisition device, and the data communication acquisition device is bidirectionally connected to the production line.

[0031] Furthermore, the output end of the data communication and acquisition device is connected to a SCADA real-time database and is connected to a MES database.

[0032] Furthermore, the output end of the data system is connected to the input end of the production line control system.

[0033] The beneficial effects of the present invention are:

[0034] 1. The present invention receives production data of the production line in real time and stores and processes it, gives appropriate optimization suggestions for processing parameters according to the current changes in production data, and quickly transmits the process parameters optimized based on the data analysis algorithm to the production equipment of the production line to realize real-time control, thereby ensuring the reliability of the optimization suggestions, improving the production efficiency of the automatic production line, and reducing production costs. Compared with manual control, it has higher reliability and stronger real-time performance.

[0035] 2. The present invention establishes a nonlinear relationship model between production line process parameters and product quality, production efficiency and tool wear based on the fully connected neural network model, which can optimize the processing process of the automatic production line and improve product quality and consistency.

[0036] 3. The present invention can accurately evaluate tool wear and factors affecting tool life by analyzing and screening production process data, thereby optimizing production line process parameters, reducing tool wear and replacement frequency, and further improving production efficiency and economic benefits. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 is a flow chart of the method of the present invention;

[0038] Figure 2 It is a schematic diagram of the actual production application of the present invention;

[0039] Figure 3 It is an execution flow chart of the particle swarm optimization algorithm in the present invention;

[0040] Figure 4 This is a schematic diagram of the architecture of the production line process parameter optimization data system of the present invention;

[0041] Figure 5 This is an execution flow chart of the production line process parameter optimization data system in the present invention. DETAILED DESCRIPTION

[0042] The following will be combined with the attached embodiment of the present invention Figure 1 To Attachment Figure 5, the technical solutions in the embodiments of the present invention are clearly and completely described. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0043] The embodiment of the present invention discloses a data-driven brake disc automatic production line process parameter optimization method and system. The brake disc automatic production line process parameter optimization method includes the following steps:

[0044] S1 collects brake disc production data. The main steps in the brake disc processing process are turning the outer circle, turning the end face, and milling the end face groove. The brake disc production data is mainly divided into three parts: one is the brake disc production process data; the second is the brake disc production efficiency data; the third is the brake disc product quality data.

[0045] The brake disc production process data mainly includes various process parameters: processing procedures, spindle speed, feed speed, tool radius compensation, cutting direction, cutting depth, etc.

[0046] The brake disc production efficiency data mainly includes: the processing time of each step in the brake disc processing process, the total time to complete the processing of each product, and the service life of the tool (the number of processed products is used as the evaluation indicator).

[0047] The brake disc product quality data mainly includes: the size of each brake disc, the surface roughness of the brake disc and the dynamic balance performance of the brake disc, and calculates the variance of the average value relative to the brake disc processing standard.

[0048] S2 The present invention proposes a data dimension reduction method for brake disc production data. This method is mainly used to reduce the size of the brake disc production data set through principal component analysis and orthogonal analysis under the premise of ensuring effective data analysis. Taking various data of S1 as the object, a mixed level orthogonal analysis scheme, such as L64 (47×86), is adopted. Combined with the selected factor levels, that is, the number of values ​​of each factor in quantity or quality in the entire scheme, S3 calculates the Pearson coefficient of each factor. On this basis, its main components are analyzed.

[0049] S4 The present invention proposes a fully connected neural network model for the brake disc production line process. This model is based on S1, S2 and S3, and uses a deep learning method to establish a nonlinear relationship between the brake disc product quality, production efficiency and process parameters. This method is mainly to improve the model's learning ability, generalization ability and model accuracy for sampled data. In order to use the loss function to measure the effectiveness of the model, the data set obtained by S2 is divided into two parts: a test set and a training set.

[0050] Preferably, different initial weights and biases are set for the training set data, and the loss function of the model is automatically calculated after each training, and the gradient of the loss function with respect to the model parameters is calculated, and the parameters of the model are updated using the calculated gradient. After a certain number of cyclic training, a neural network is generated whose loss function no longer decreases significantly. The neural network model is established by taking the production line process parameters, including the processing steps, spindle speed, cutting depth, feed speed, etc., as the input of the neural network, and taking the brake disc size error, surface roughness error, dynamic balance performance, brake disc processing efficiency, tool wear, etc. as the output.

[0051] After S5 establishes the neural network model through S4, it uses the particle swarm algorithm (PSO) to find the best combination of tool process parameters. First, all process parameter combinations are normalized. Secondly, considering that production efficiency and product quality are more important, the weighting factor of the former is 1 / 3 and the weighting factor of the latter is 2 / 3. The number of particles and the number of iterations of the particle swarm algorithm are set, and the control authority is adjusted to facilitate global search. Thus, the optimal process parameter combination is trained.

[0052] In the S6 stage, applying the optimal production line process parameter combination in real time to the production process of the brake disc automatic production line and making adaptive adjustments through the closed-loop management mechanism of the whole process is a key step in achieving quality control of the brake disc production process. This process requires not only real-time and accuracy, but also the ability to continuously monitor the entire life cycle of the production line and the scalability to adapt to different brake disc production processes.

[0053] The closed-loop management mechanism of the entire process feeds back various data generated during the production process (such as processing parameters, environmental parameters, product quality parameters, etc.) to the data system in real time, forming a complete production data chain. The data system analyzes the feedback data in real time, evaluates the gap between the current production status and the expected goals, and identifies potential problems or optimization space. Based on the data analysis results, the data system automatically triggers the adjustment mechanism to fine-tune or re-optimize the production line process parameters to adapt to changes in production conditions or improve production efficiency. Through continuous data collection, analysis and adjustment, a closed-loop management mechanism is formed to promote continuous improvement and optimization of the production process.

[0054] The data system is seamlessly integrated with the brake disc automatic production line control system, and the optimized process parameters are transmitted to the production line control system in real time through the API interface or other data exchange methods. After receiving the optimal process parameters, the production line control system automatically adjusts the settings of each production link, including tool processing parameters, environmental control parameters, etc., to ensure that the production process is strictly carried out in accordance with the optimized parameters. At the same time, the data system needs to monitor the operating status of the production line in real time, including the execution of process parameters, equipment operating status, product quality feedback, etc., to ensure the stability and controllability of the production process.

[0055] The present invention proposes a data system for optimizing the process parameters of a brake disc production line. This system is mainly used to realize decision-making and real-time control of the brake disc production process and to realize dynamic adjustment of the process parameters of the production line. The brake disc production data is collected based on the data communication acquisition equipment and uploaded to the MES service platform. Through the data analysis algorithm in the system, the optimal process parameters of the brake disc processing process are calculated in real time, and the calculation results are sent to the production line control system through the data communication acquisition equipment.

[0056] The data system has a fast response speed and calculation speed, and can process the incoming sensor data in real time and give optimization suggestions through intelligent algorithms to achieve dynamic decision-making of process parameters during the brake disc processing.

[0057] The data communication acquisition equipment is used to collect data related to the processing and analysis of brake discs. It is deployed on the brake disc automated production line. The data collected by the equipment includes three types of data. The first is to collect data from the PCS or SCADA control unit of production equipment such as CNC lathes and CNC milling machines. The second is that the data system collects data through the equipment control unit. The third is the real-time collection of the operation data of the production execution system MES. The entire acquisition equipment sends the data to the data system at a specified frequency.

[0058] The principle of the method is as follows Figure 1 As shown, the method includes collecting production data of a brake disc production line; reducing the size of a production data set; calculating the Pearson coefficient of each factor in the production process data; establishing a nonlinear relationship between brake disc product quality, production efficiency and process parameters through a fully connected neural network model; normalizing the above process parameter combination results, and training the optimal process parameter combination through a particle swarm algorithm to achieve better processing results.

[0059] The following will be combined Figures 2 to 5 A preferred embodiment of a data-driven brake disc automatic production line process parameter optimization method and system is described in detail.

[0060] like Figure 2In a brake disc automatic production line, the processing tools and workpieces are put in place, and the current sensor, voltage sensor, pressure sensor, liquid level sensor, rotary transformer, induction synchronizer, speed sensor and data acquisition device in the device are installed on the production equipment of the brake disc automatic production line.

[0061] The workpieces are processed according to the technological process: rough turning of the outer circle and end face, semi-finishing turning of the outer circle and end face, finishing turning of the outer circle and end face, milling of end face grooves, drilling, deburring, inspection and warehousing. The production data of the brake disc is collected, including spindle speed, tool speed, cutting depth, feed speed, processing time, tool life, brake disc size, surface roughness of the brake disc and dynamic balance performance of the brake disc.

[0062] A data dimensionality reduction method for brake disc production data is proposed. This method is mainly used to reduce the size of the brake disc production data set through principal component analysis and orthogonal analysis while ensuring the effectiveness of data analysis. The purpose of principal component analysis is to extract meaningful features from the acquired production line raw data for neural network input in subsequent steps. First, the raw data is standardized to ensure that all features have similar scales. This usually includes mean centering (subtracting the mean) and standardization (dividing by the standard deviation):

[0063]

[0064] Where: X standardized is the standardized data, X is the original data, μ is the mean of each feature, and σ is the standard deviation of each feature.

[0065] Next, we calculate the covariance matrix of the normalized data, which measures the correlation between features. The elements of the covariance matrix C are ij Represents the covariance between feature i and feature j:

[0066]

[0067] Where: n is the number of samples, X i is the i-th sample, is the mean vector of all samples.

[0068] Then perform eigenvalue decomposition on the covariance matrix C to obtain eigenvectors and eigenvalues. Eigenvector v i Corresponding to the eigenvalue λ i , where i = 1, 2, ..., p, p is the characteristic dimension of the data. The eigenvector describes the main direction of change in the data, and the eigenvalue indicates the size of the variance in the corresponding direction.

[0069] When choosing the number of principal components to retain, the size of the eigenvalues ​​needs to be considered. The principal components whose variance contribution in the production line production data is greater than the 95% threshold will be retained.

[0070] Project the data onto the first k principal components selected to obtain the reduced dimension data. This can be achieved by the following formula:

[0071] Y=X·V k

[0072] Among them: Y is the data matrix after dimension reduction, X is the data matrix after standardization, V k is a matrix containing the first k eigenvectors.

[0073] Taking the above-mentioned dimension-reduced data Y as the object, a mixed level orthogonal analysis method, such as L64 (47×86), is used to calculate the Pearson coefficient of each factor by combining the selected factor levels. First, the mean of each pair of features is calculated, which will be used to calculate the correlation coefficient. For the i-th feature, its mean (average value) can be calculated as follows:

[0074]

[0075] in: is the mean of the ith feature. ij is the value of the i-th feature on the j-th sample.

[0076] Next, calculate the Pearson correlation coefficient between each pair of features. The Pearson correlation coefficient r is defined by the following formula:

[0077]

[0078] Where: r ij is the Pearson correlation coefficient between the i-th feature and the j-th feature. ik is the value of the i-th feature on the k-th sample. and are the means of the i-th feature and the j-th feature, respectively. Repeat this step to calculate the correlation coefficient between k(k-1) / 2 pairs of features, where k is the number of features.

[0079] After calculating the correlation coefficients, we can interpret them to understand the linear correlation between the features after dimensionality reduction. The correlation coefficient ranges from -1 to 1 and has the following meanings:

[0080] r = 1 indicates a perfect positive correlation, r = -1 indicates a perfect negative correlation, and r = 0 indicates no correlation. Positive correlation means that when one feature increases, the other feature also increases, while negative correlation means that when one feature increases, the other feature decreases. Using this coefficient as a reference, the main components are analyzed on this basis.

[0081] On the basis of the above, the deep learning method is used to establish the nonlinear relationship between the brake disc processing parameters and the brake disc product quality and tool wear. The deep learning method uses a fully connected neural network model, a fully connected neural network, which contains:

[0082] Input layer: There are n features, represented as x 1 ,x 2 ,…,x n .

[0083] L hidden layers: Layer 1 (where l = 1, 2, ..., L) contains m (l) A neuron.

[0084] Output layer: contains k neurons, where k is the dimension of the output. In the present invention, k=3.

[0085] The forward propagation process of a fully connected neural network is as follows:

[0086] 1. Calculation from input layer to hidden layer

[0087] For the first hidden layer l = 1, the weighted sum of the inputs of the hidden layer neurons is calculated, and then a nonlinear transformation is performed through the activation function f:

[0088]

[0089] in, represents the weighted input of the jth neuron in the first hidden layer. represents the weight connecting the i-th feature of the input layer to the j-th neuron of the first hidden layer. represents the bias of the jth neuron in the first hidden layer. Represents the output of the jth neuron in the first hidden layer (the value after the activation function).

[0090] 2. Hidden layer to hidden layer calculation

[0091] This method calculates the weighted sum of the inputs of the hidden layer neurons for the lth hidden layer (where l = 2, 3, ..., L), and then performs a nonlinear transformation through the activation function f:

[0092]

[0093] in, represents the weighted input of the jth neuron in the lth hidden layer. represents the weight connecting the i-th neuron in the l-1th hidden layer to the j-th neuron in the l-th hidden layer. represents the bias of the jth neuron in the lth hidden layer. Represents the output of the jth neuron in the lth hidden layer (the value after the activation function).

[0094] 3. Calculation from the last hidden layer to the output layer

[0095] For the output layer, the weighted sum of the inputs to the output layer neurons is calculated, and then the appropriate activation function is applied:

[0096]

[0097] in, represents the weighted input of the jth neuron in the output layer. represents the weight connecting the i-th neuron in the last hidden layer to the j-th neuron in the output layer. Represents the bias of the jth neuron in the output layer. a_j^((L+1)) represents the output of the jth neuron in the output layer (the value after the activation function).

[0098] The above process is mainly to improve the model's learning ability, generalization ability and model accuracy for sampled data. Since the loss function is a standard for measuring model stability, the data set obtained after orthogonal analysis is divided into two parts: the test set and the training set. Different initial weights and biases are set for the training set data. After each training, the loss function of the model in the test set is automatically calculated and compared with the loss function of the previous model. A network with a smaller change in loss function is retained. After a certain number of cyclic training, a neural network with a loss function that no longer decreases significantly is generated.

[0099] The neural network model is established with spindle speed, tool speed, cutting depth, feed rate, etc. as the input of the neural network, and processing efficiency, tool life, brake disc product quality, etc. as the output. The process of establishing a fully connected neural network includes the following key steps. First, select the architecture of the network, determine the number of neurons in the input layer, the number of layers in the middle hidden layer and the number of neurons in each layer, and the number of neurons in the output layer. These choices depend on the complexity of the specific problem and the task to be solved. Secondly, initialize the weights (the strength of the connection between the connected neurons) and the bias (the offset of each neuron) of the network. Next, define the activation functions, which are applied on each neuron to introduce nonlinearity. In this invention, ReLU will be used as the activation function:

[0100] f(x)=max(0,x)

[0101] Wherein, x represents the input value, and f(x) represents the output of the activation function. ReLU as an activation function can alleviate the gradient vanishing problem of the model. Then, a loss function is determined to measure the difference between the network output and the true value. In order to solve the regression problem of the process optimization parameters, the mean square error is used as the loss function. Next, the gradient descent algorithm is selected as the optimization algorithm to update the weights and reduce the value of the loss function. Specifically, the gradient descent algorithm used in the present invention is the stochastic gradient descent algorithm (SGD). In each iteration, a sample in the training data set is randomly selected to calculate the gradient and update the parameters:

[0102]

[0103] Among them, θ is the parameter vector to be optimized, η is the learning rate, J(θ,x i ,y i ) is for sample x i and label y i The loss function is is the gradient of the objective function with respect to the parameters θ, based on only a single example.

[0104] After establishing the neural network model through the above steps, the particle swarm algorithm is used to find the best combination of process parameters to improve production efficiency and product quality. Figure 3 The flowchart of particle swarm optimization is shown in Figure 2. Particle swarm optimization (PSO) is an optimization algorithm based on swarm intelligence, which is inspired by natural phenomena such as collective behavior of bird flocks or fish schools. The implementation process of PSO is as follows: First, a group of "particles" are randomly initialized, each particle represents a potential solution, and these solutions have a specific position X in the solution space. i These positions can be understood as search points in the solution space, and each particle is looking for the best solution. Next, a random velocity V is assigned to each particle. i , used to move in the solution space. This speed affects the direction and rate of exploration of the particle in the solution space. In each iteration, each particle considers its own historical optimal solution and the global optimal solution. The historical optimal solution refers to the best solution found by the particle in the past iteration. The global optimal solution is the particle with the best fitness value f(X i ). Each particle adjusts its own speed and position according to the velocity formula:

[0105] V i (t+1)=w·V i (t)+c 1 ·r 1 ·(Pbest i -X i (t))+c2 ·r 2 ·(Gbest-X i (t))

[0107] The speed formula consists of two parts: one is the historical experience of the particle Gbest (i.e. its own historical optimal position), and the other is the historical experience of the entire group Pbest i (i.e. the global optimal position). These factors jointly affect the movement direction and speed of the particles in the solution space. Through continuous iteration, the particle swarm will gradually gather towards the position of the global optimal solution. The movement of each particle in the search space is affected by its own experience and group collaboration, which enables the PSO algorithm to effectively find the global optimal solution. This embodiment uses a particle swarm algorithm combined with a deep learning model to calculate the optimal tool process parameters for brake disc processing.

[0108] The results of the above process parameter combinations are normalized. Considering that production efficiency and product quality are more important, the weighting factor of the former is 1 / 3 and the weighting factor of the latter is 2 / 3. The number of particles and the number of iterations of the particle swarm algorithm are set, and the control authority is adjusted to facilitate global search. Thus, the optimal process parameter combination is trained.

[0109] The brake disc production data is collected based on the data communication acquisition equipment and uploaded to the MES service platform. The optimal process parameters of the brake disc processing process are calculated in real time through the data analysis algorithm in the system, and the calculation results are sent to the production line control system through the data communication acquisition equipment.

[0110] In addition to introducing the brake disc automatic production line process parameter optimization method, this embodiment uses a brake disc automatic production line process parameter optimization data system, the main equipment includes current sensor, voltage sensor, pressure sensor, liquid level sensor, rotary transformer, induction synchronizer, speed sensor, CNC milling machine, CNC lathe, data communication acquisition equipment and data processing platform. This system is mainly used for real-time control and auxiliary decision-making of the processing parameters of the brake disc production line, and realizes the dynamic adjustment of the brake disc processing parameters.

[0111] The architecture of the data system is as follows Figure 4 As shown in the figure, the system is divided into three layers, namely the physical layer, the data layer and the model layer. The physical layer is the actual production line and includes sensors arranged in the actual production line. The data system obtains the real-time status of the production line from the physical layer; the database in the data layer can store and transmit the data transmitted from the physical layer and the model layer. The model layer will perform real-time analysis of the production line status based on the optimization algorithm and deep learning neural network, and provide the optimized process parameters.

[0112] The workflow of the data system is as follows Figure 5 As shown in the figure, real-time data is obtained through data communication acquisition equipment at the physical layer and uploaded to the MES service platform. In the data layer, time domain analysis and frequency domain analysis are used for feature extraction. The extracted features are uploaded to the data preprocessing module for principal component analysis and horizontal mixed orthogonal analysis. Then, the relationship between production line process parameters and product quality and production efficiency is established using a fully connected neural network model. Finally, the optimal combination of process parameters is obtained through the particle swarm optimization algorithm and fed back to the control system at the physical layer.

[0113] The data communication acquisition equipment is used to collect data related to the process parameter analysis of the brake disc production line. It is deployed in the physical layer of the data system. The data collected by the equipment includes three types of data. The first is to collect data from the PCS or SCADA control unit of production equipment such as CNC lathes and CNC milling machines. The second is that the data system collects data through the equipment control unit. The third is the real-time collection of the operation data of the production execution system MES. Data communication acquisition sends data to the data layer of the data system at a specified frequency.

[0114] So far, the technical solutions of the present invention have been described in conjunction with the preferred embodiments shown in the accompanying drawings. However, it is easy for those skilled in the art to understand that the protection scope of the present invention is obviously not limited to these specific embodiments. Without departing from the principle of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will fall within the protection scope of the present invention.

Claims

1. A data-driven brake disc automatic production line process parameter optimization method, characterized in that: The following steps are involved: S1. Collect the production process data of the brake disc, which at least includes the brake disc tool processing parameters, environmental parameters, tool wear parameters and the quality parameters of the brake disc after processing; S2. Use principal component analysis to map the collected high-dimensional production process data into a low-dimensional space, and preliminarily screen the parameters that affect the machining quality of the brake disc and tool wear through mixed level orthogonal analysis; S3. Evaluate the linear correlation between the selected parameters and the brake disc processing quality parameters by calculating the Pearson coefficient, and perform secondary screening on the parameters that affect the brake disc processing quality and tool wear based on the calculation results; S4. Establishing a nonlinear relationship model between production line process parameters and product quality, production efficiency and tool wear based on a fully connected neural network model, wherein the fully connected neural network model is trained and iteratively optimized based on a training set to obtain a combination result of production line process parameters to be optimized; S5. Normalize the process parameter combination results to be optimized, and input the normalized data into the particle swarm optimization algorithm for iterative optimization to obtain the optimal production line process parameter combination; S6. Apply the optimal production line process parameter combination in real time to the production process of the brake disc automatic production line, and perform adaptive adjustments through the closed-loop management mechanism of the entire process to achieve quality control of the brake disc production process.

2. A data-driven brake disc automatic production line process parameter optimization method according to claim 1, characterized in that: In S3, the parameters affecting the machining quality of the brake disc preliminarily screened out in S2 are taken as the object, and the Pearson coefficient of the preliminarily screened parameters is calculated in combination with the parameter levels selected by the mixed level orthogonal analysis to evaluate the correlation between the preliminarily screened parameters and the machining quality parameters of the brake disc. The output function expression of the Pearson coefficient is: In formula (1), X i and Y i Indicates the X-axis coordinate and Y-axis coordinate of the i-th data point in the scatter plot in the preliminary screening parameter data set. and are the mean X-axis coordinates and the mean Y-axis coordinates of all data points of the preliminary screening parameters in the scatter plot, and the symbol ∑ represents the sum of all data points.

3. The data-driven brake disc automatic production line process parameter optimization method according to claim 1 is characterized in that: The fully connected neural network model uses the parameters screened out secondary as input, and the brake disc processing quality and tool wear as output, and predicts the product quality, production efficiency and tool wear under different process parameter combinations by adjusting the weights and biases of the neural network.

4. The data-driven brake disc automatic production line process parameter optimization method according to claim 1 is characterized in that: In S2, the data set obtained by principal component analysis is divided into a test set and a training set. The fully connected neural network model initializes weights and biases based on the training set data, and automatically calculates the model loss function after each training. The output formula of the loss function is: In formula (2), y is the actual target value, is the predicted value of the fully connected neural network model, n is the number of samples, and the parameters of the fully connected neural network model are updated by calculating the change rate of the loss function in the model parameter space.

5. The data-driven brake disc automatic production line process parameter optimization method according to claim 1 is characterized in that: The working method of the particle swarm optimization algorithm comprises the following steps: The process parameter combination of the production line to be optimized is taken as a particle, and the maximum number of iterations, the number of particles, the inertia weight, the learning factor, the initial position and speed of the particles are set; The fitness value of each particle is calculated according to the prediction results of the fully connected neural network model, and the individual optimal position and group optimal position of the particle are updated according to the fitness calculation results, and then the historical optimal position of each particle and the historical optimal position fitness value of the particle group are recorded; Update the position and speed of each particle based on the speed and position of the historical optimal position and the current position, then recalculate the fitness value of each particle according to the updated particle position, and update the historical optimal process parameter combination and the historical optimal position and fitness value of the entire group; Determine whether the maximum number of iterations or the preset convergence degree has been reached. If the maximum number of iterations or the fitness threshold has been reached, the historical optimal process parameter combination is output as the final optimization result; otherwise, return to calculate the fitness value of each particle; Output the optimal production line process parameter combination as the final optimization result.

6. The data-driven brake disc automatic production line process parameter optimization method according to claim 1 is characterized in that: The normalization process in S5 is expressed as: In formula (3), X normalized represents the normalized process parameter combination to be optimized, X represents the original process parameter combination to be optimized, μ represents the mean of the process parameter combination data set to be optimized, and σ represents the standard deviation of the process parameter combination data set to be optimized.

7. A data-driven brake disc automatic production line process parameter optimization system, applied to a data-driven brake disc automatic production line process parameter optimization method according to any one of claims 1 to 6, characterized in that: The process parameter optimization system for the automatic production line of brake discs includes a data system and a data communication and acquisition device. The data system is bidirectionally connected to the data communication and acquisition device, and the data communication and acquisition device is bidirectionally connected to the production line.

8. The data-driven brake disc automatic production line process parameter optimization system according to claim 7, characterized in that: The output end of the data communication and acquisition equipment is connected to a SCADA real-time database and is also connected to an MES database.

9. The data-driven brake disc automatic production line process parameter optimization system according to claim 7, characterized in that: The output end of the data system is connected to the input end of the production line control system.

Images