Prediction analysis system for controlling tire toe bubble generation equipment

Abstract

The invention relates to the technical field of tire toe bubble inspection, and discloses a predictive analysis system for controlling tire toe bubble generation equipment, and the system comprises a data collection module which is used for obtaining equipment operation data; the data preprocessing module is used for carrying out cleaning, normalization, abnormal value detection and missing value filling on the data; the feature extraction module is used for extracting key features; the bubble generation prediction module is used for constructing a prediction model based on a deep neural network DNN algorithm; the equipment state monitoring module is used for monitoring the abnormal state of the equipment by using a random forest RF algorithm; and the parameter optimization module is used for optimizing equipment parameters by adopting a genetic algorithm GA so as to reduce bubbles. The model output integration module integrates output results of all the models, and the decision support module generates an equipment control strategy report according to the results. According to the method, the accuracy and practicability of bubble generation prediction are improved, and the bubble problem of the tire toe is effectively controlled.

Description

Technical Field

[0001] The invention relates to the technical field of tire toe bubble inspection, in particular to a prediction and analysis system for controlling tire toe bubble generating equipment. Background Art

[0002] During the tire manufacturing process, the problem of bubbles at the tire toe is a common and critical problem that affects product quality. The tire toe is the key part where the tire contacts the rim, and its quality directly affects the sealing, durability and safety of the tire. The generation of bubbles will not only reduce the appearance quality of the tire, but may also affect the performance of the tire during use, and even cause serious safety accidents such as tire bursting when driving at high speed.

[0003] Traditionally, controlling the occurrence of bubbles at the toe of tires mainly relies on the experience of operators and regular quality inspections. This method has many limitations. On the one hand, the experience levels of operators vary, making it difficult to accurately judge and promptly handle the bubble problem. On the other hand, regular quality inspections are often conducted after the problem has already occurred, making it impossible to achieve early warning and effective control of bubble occurrence. With the continuous development of industrial automation and intelligent technology, the application of predictive analysis systems in the industrial field is becoming more and more extensive. However, the predictive analysis system for tire toe bubble generating equipment is still imperfect. Most existing predictive analysis systems focus on the overall monitoring of the production process, but lack targeted analysis of the specific problem of tire toe bubbles. Summary of the invention

[0004] The object of the present invention is to provide a predictive analysis system for controlling a tire toe bubble generating device to solve the problems raised in the above-mentioned background technology.

[0005] To achieve the above object, the present invention provides the following technical solution: a prediction and analysis system for controlling a tire toe bubble generating device, the system comprising: Data collection module: collects the operating data of the tire toe bubble generating device; Data preprocessing module: preprocess the collected data, including data cleaning, normalization, outlier detection and missing value filling, to form a preprocessed data set; Feature extraction module: extracts features from the preprocessed data set to form a feature data set; Bubble Occurrence Prediction Module: Build a bubble occurrence prediction model based on the deep neural network (DNN) algorithm. The model is trained using feature data sets to predict the probability of bubble occurrence. Equipment status monitoring module: Build an equipment status monitoring model based on the random forest RF algorithm. This model uses the feature data in the preprocessed data set to identify abnormal status of equipment operation. Parameter optimization module: Build a parameter optimization model based on the genetic algorithm (GA) algorithm. This model uses historical operating data and equipment parameters to optimize the equipment's operating parameters to reduce the occurrence of bubbles. Model output integration module: input the characteristic data set into the bubble occurrence prediction model, equipment status monitoring model and parameter optimization model respectively to obtain the model output results; Decision support module: Generates equipment control strategy reports based on model output results, including bubble occurrence risk, equipment status assessment, and optimized operating parameters.

[0006] Preferably, the operating data includes equipment working pressure, temperature, gas flow, toe size, equipment vibration frequency, operating time and maintenance records; the feature data set includes statistical features, time domain features and frequency domain features.

[0007] Preferably, the bubble occurrence prediction model adopts a multi-layer deep neural network structure, extracts features and predicts the probability of bubble occurrence through multiple layers of fully connected layers, and its network structure includes: Input layer: receives data from the feature dataset, and the feature vector dimension is [N], where N is the number of features; Hidden layer: multiple layers of fully connected layers are used, each layer is followed by a ReLU activation function to extract deep features; Output layer: Outputs the probability of bubble occurrence and uses the Sigmoid function to map the output value to the [0,1] interval.

[0008] Preferably, the training step of the bubble generation prediction model includes: D1: Initialize the parameters of the DNN model, including weights, bias terms, and learning rates; D2: Divide the preprocessed feature data set into a training set and a validation set. The training set is used for model training, and the validation set is used for model validation. D3: Use the training set to iteratively train the DNN model. In each iteration, the model output is calculated by forward propagation, and the loss between the predicted value and the true value is calculated according to the loss function; the loss function adopts the binary cross entropy loss function; D4: Update the model parameters based on the loss value through the back propagation algorithm and gradient descent method; D5: Repeat steps D3 to D4 until the loss converges or the preset number of iterations is reached; D6: Use the validation set to validate the trained DNN model and evaluate the accuracy, precision, and recall of the model; D7: When the model performance reaches the preset standard, the model parameters are saved to obtain the trained bubble occurrence prediction model.

[0009] Preferably, the equipment status monitoring model realizes monitoring of the equipment operation status by constructing multiple decision trees and integrating their output results, and its structure includes: Input layer: receives data from feature datasets; Decision tree layer: construct multiple decision trees, each of which is split according to the input features to form decision rules; Integration layer: Integrate the output results of all decision trees and determine the final device status through a voting mechanism; Output layer: Output equipment status monitoring results, including normal operation, abnormal operation or fault warning.

[0010] Preferably, the step of training the equipment status monitoring model includes: E1: Use the feature dataset as the training set and select the device status as the label; E2: Determine the parameters of the random forest, including the number of trees, maximum depth, and minimum number of sample splits; E3: Use the training set to train the random forest model until the model converges; E4: Use the validation set to validate the random forest model and evaluate the classification accuracy and generalization ability of the model; E5: When the model performance reaches the preset standard, the model parameters are saved to obtain the trained equipment status monitoring model.

[0011] Preferably, the training step of the parameter optimization model includes: F1: Define the optimization objective function, including minimizing the probability of bubble occurrence and maximizing equipment operation efficiency; F2: Initialize the population, each individual represents a set of equipment working parameters; F3: Evaluates the fitness of each individual in the population, indicating the value of the objective function; F4: Select the individuals with the highest fitness to perform crossover and mutation operations to generate a new generation of population; F5: Repeat steps F3 to F4 until the preset number of iterations is reached or the optimal solution is found; F6: Output the best individual, indicating the optimized equipment working parameters.

[0012] Preferably, the specific implementation steps of the decision support module include: S1: Define the probability output by the bubble occurrence prediction model as P_Bubble, the status output by the equipment status monitoring model as S_Status, and the optimization parameter output by the parameter optimization model as P_Optimized; S2: Determine the risk level of bubble occurrence based on P_Bubble and the preset threshold; S3: Combined with S_Status, evaluate the current operating status of the equipment and issue an early warning if there is an abnormality; S4: According to P_Optimized, adjust the operating parameters of the equipment to reduce the risk of bubble generation; S5: Generate equipment control strategy report based on risk level, equipment status and optimization parameters; S6: Send the control strategy report to the equipment operator or control system to implement corresponding control measures.

[0013] Preferably, the risk level output by the decision support module is divided into four levels: low risk, medium risk, high risk, and extremely high risk.

[0014] Preferably, the present invention further comprises: an early warning release module, which automatically selects a corresponding early warning information release method according to the risk level output by the policy support module, including SMS notification, email notification, sound and light alarm and voice alarm.

[0015] Compared with the prior art, the present invention has the following beneficial effects: The bubble occurrence prediction module uses the deep neural network DNN algorithm to build a prediction model and uses the feature data set for training. It can accurately predict the probability of bubble occurrence and improve the prediction accuracy. The system uses the random forest RF algorithm to build an equipment status monitoring model through the equipment status monitoring module to identify abnormal equipment operation in real time and promptly discover potential bubble risks. Combined with the output results of the bubble occurrence prediction model, the system can achieve early warning of bubble occurrence, provide operators with sufficient reaction time, and take effective measures to prevent the generation of bubbles.

[0016] The parameter optimization module optimizes the working parameters of the equipment through the genetic algorithm GA algorithm, reduces the possibility of bubble occurrence, and realizes effective control of bubble occurrence.

[0017] This system is specially designed for tire toe bubble generating equipment. It is highly targeted and can accurately reflect the complex relationship between bubble generation and equipment operating parameters. The modules of the system work together to form a complete prediction and analysis process, which improves the practicality and operability of the system. The decision support module generates an equipment control strategy report based on the model output results, providing operators with clear guidance and suggestions for practical operation and application. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 This is a working principle diagram of a predictive analysis system for controlling a tire toe bubble generating device according to the present invention; Figure 2 Flowchart for training the bubble occurrence prediction model; Figure 3This is the working step diagram of the decision support module. DETAILED DESCRIPTION

[0019] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0020] See also Figure 1-3 The present invention provides a technical solution: a prediction and analysis system for controlling a tire toe bubble generating device, the system comprising: Data collection module: This module is responsible for collecting the operating data of the tire toe bubble generating device. In specific implementation, various sensors such as temperature sensors, pressure sensors, vibration sensors, etc. are installed on the equipment to collect various parameters of the equipment operation process in real time. The operating data includes equipment working pressure, temperature, gas flow, toe size, equipment vibration frequency, operating time and maintenance records.

[0021] Data preprocessing module: This module preprocesses the collected raw data to ensure the quality and consistency of the data. In specific implementation, data cleaning is first performed to remove noise and outliers in the data; then normalization is performed to convert data of different dimensions to the same dimension to facilitate subsequent feature extraction and model training; then outlier detection is performed to identify and process outliers in the data; finally, missing value filling is performed to estimate and fill in missing data to form a complete preprocessed data set. The specific method of outlier detection is: The Z-score method is used to detect outliers. The mean and standard deviation of each feature data are calculated. If the absolute value of the Z-score of a data point is greater than 3, it is determined to be an outlier. The formula is: ,in, is the value of a data point, is the mean, is the standard deviation.

[0022] Feature extraction module: This module extracts features from the preprocessed data set to form a feature data set. In specific implementation, the key features that have an impact on the occurrence of bubbles are selected based on the mechanism of bubble generation at the tire toe and the operating characteristics of the equipment. These features may include temperature fluctuations, pressure change rates, vibration spectra, etc. of the equipment. These key features are extracted from the preprocessed data set through feature extraction algorithms, such as principal component analysis (PCA) and independent component analysis (ICA), to form a feature data set. The feature data set includes statistical features, time domain features, and frequency domain features.

[0023] Bubble Occurrence Prediction Module: This module builds a bubble occurrence prediction model based on the deep neural network (DNN) algorithm. In the specific implementation, the network structure of the DNN is first designed, including the input layer, hidden layer, and output layer. The input layer receives the feature data set, the hidden layer performs nonlinear transformation through multiple layers of neurons, and the output layer gives the probability of bubble occurrence. Then, the DNN model is trained using the feature data set, and the model parameters are adjusted through the back propagation algorithm so that the model can accurately predict the probability of bubble occurrence.

[0024] Equipment status monitoring module: This module builds an equipment status monitoring model based on the random forest (RF) algorithm. In specific implementation, the RF model is first trained using the feature data in the preprocessed data set, and the abnormal status of the equipment operation is identified by building multiple decision trees and synthesizing their output results. The RF model can process high-dimensional data and accurately classify and identify abnormal status of the equipment.

[0025] Parameter optimization module: This module builds a parameter optimization model based on genetic algorithm (GA). In specific implementation, the optimization target is first defined, that is, the combination of working parameters that reduces the occurrence of bubbles. Then, using historical operating data and equipment parameters as the initial population, the optimal combination of working parameters is iteratively searched through GA selection, crossover and mutation operations. Finally, the optimized working parameters are obtained to adjust the operating status of the equipment and reduce the occurrence of bubbles.

[0026] Model output integration module: This module inputs the feature data set into the bubble occurrence prediction model, equipment status monitoring model and parameter optimization model respectively to obtain the model output results. In specific implementation, the feature data set is passed to each model through the data interface, and the output results of each model are received. These output results include the probability of bubble occurrence, the evaluation results of the equipment status and the optimized working parameters.

[0027] Decision support module: This module generates a device control strategy report based on the model output results. In specific implementation, first analyze the probability of bubble occurrence and evaluate the risk level of bubble occurrence; then, based on the evaluation results of the equipment status, determine whether the equipment needs maintenance or adjustment; finally, formulate an adjustment plan for the equipment based on the optimized working parameters. This information is integrated into the device control strategy report to provide decision support for operators.

[0028] The present invention will be further described below in conjunction with Examples 1 to 4: Example

[0029] The bubble occurrence prediction module builds a bubble occurrence prediction model based on the deep neural network (DNN) algorithm. The training steps are specifically implemented as follows: D1: Initialize the parameters of the DNN model Determine the network structure of the DNN model, including an input layer, several hidden layers, and an output layer, each layer containing several neurons.

[0030] Initialize the model parameters, including the weight matrix, bias term, and learning rate. The weight matrix is ​​initialized randomly, such as Gaussian distribution or uniform distribution; the bias term is usually initialized to 0 or a small constant; the learning rate is determined based on experience or experiments and is used to control the step size of the model parameter update.

[0031] D2: Divide the dataset The preprocessed feature dataset is divided into a training set and a validation set. The training set is used for model training, and the validation set is used for model verification and performance evaluation.

[0032] The division ratio can be determined based on the amount of data and the complexity of the model. Generally, the training set accounts for a larger proportion, such as 70%-80%, and the validation set accounts for a smaller proportion, such as 20%-30%.

[0033] D3: Iterative training of DNN model The DNN model is iteratively trained using the training set. In each iteration, forward propagation calculations are performed, starting from the input layer, and the output of neurons is calculated layer by layer until the output layer obtains the predicted value.

[0034] The loss between the predicted value and the true value is calculated according to the loss function. In this embodiment, the loss function adopts a binary cross entropy loss function, which is suitable for binary classification problems, such as whether bubbles occur or not.

[0035] The formula for the binary cross entropy loss function is: ; Where N is the number of samples, is the true label (0 or 1), is the predicted probability.

[0036] D4: Update model parameters According to the loss value, the gradient of the loss function to the output of each layer of neurons is calculated through the back propagation algorithm.

[0037] Use the gradient descent method to update the model parameters, namely the weight matrix and bias term. The update formula is: ; in are model parameters, is the learning rate, is the gradient of the loss function with respect to the parameters.

[0038] D5: Repeat the training process Repeat steps D3 to D4 until the loss converges or the preset number of iterations is reached. The judgment criteria for loss convergence can be determined based on experience or experiments, such as the loss value no longer significantly decreases or the change amplitude is less than a threshold.

[0039] D6: Verify model performance Use the validation set to validate the trained DNN model and evaluate the accuracy, precision, and recall of the model.

[0040] Accuracy indicates the proportion of samples predicted correctly; precision indicates the proportion of samples predicted to be positive that are actually positive; recall indicates the proportion of samples actually positive that are predicted to be positive.

[0041] The calculation formulas for these performance indicators are: Accuracy = (TP+TN) / (TP+TN+FP+FN), Precision = TP / (TP+FP), Recall = TP / (TP+FN), where TP is a true positive example, TN is a true negative example, FP is a false positive example, and FN is a false negative example.

[0042] D7: Save model parameters When the model performance reaches the preset standard, such as the accuracy, precision and recall rate all reach or exceed the threshold, the model parameters are saved to obtain the trained bubble occurrence prediction model.

[0043] The saved model parameters include weight matrix, bias term and network structure information, which can be used for subsequent model loading and prediction.

[0044] Through the above steps, the bubble occurrence prediction model provided in this embodiment can accurately predict the probability of tire toe bubble occurrence.

[0045] The following is a detailed description of the implementation process of the bubble occurrence prediction model, with examples, in which the input layer receives the feature data set, the hidden layer performs nonlinear transformation through multiple layers of neurons, and the output layer gives the probability of bubble occurrence.

[0046] The data collection module collects equipment operation data through various sensors installed on the tire toe bubble generating equipment, such as temperature sensors, pressure sensors, vibration sensors, etc. These data include equipment working pressure, temperature, gas flow, toe size, equipment vibration frequency, operating time and maintenance records. After data cleaning, normalization, outlier detection and missing value filling in the data preprocessing module, a preprocessed data set is formed. The feature extraction module extracts statistical features, time domain features and frequency domain features from the preprocessed data set according to the mechanism of tire toe bubble generation and the operating characteristics of the equipment using the principal component analysis (PCA) algorithm to form a feature data set. For example, in the collected equipment operation data, it is found through analysis that the temperature fluctuation, pressure change rate and other features of the equipment have a great influence on the bubble generation. After being processed by the feature extraction algorithm, these key features are integrated into the feature data set. The input layer receives the feature data set, and its feature vector dimension is [N], where N is the number of features. Assuming that 10 key features are determined after feature extraction, then the feature vector dimension N=10 at this time.

[0047] The bubble occurrence prediction model adopts a multi-layer deep neural network structure, and the hidden layer adopts a multi-layer fully connected layer, and each layer is followed by a ReLU activation function. Taking a model with three hidden layers as an example, the first hidden layer receives the feature data transmitted from the input layer, and the neurons in the layer perform weighted summation with the input data through the weight matrix, add the bias term, and then perform nonlinear transformation through the ReLU activation function. The formula of the ReLU activation function is f(x)=max(0,x), which can effectively solve the gradient disappearance problem and enable the model to learn more complex features. For example, the weighted summation result received by a neuron is -2, and after being processed by the ReLU activation function, the output is 0; if the weighted summation result is 3, the output is 3. After such processing, the first hidden layer passes the result to the second hidden layer, and the second hidden layer also performs weighted summation, adds the bias term, and processes the ReLU activation function. And so on, through the layer-by-layer transformation of multi-layer neurons, the deep features of the data are continuously extracted.

[0048] After multiple nonlinear transformations in the hidden layer, the data is passed to the output layer. The output layer uses the Sigmoid function to map the output value to the [0,1] interval to give the probability of bubble occurrence. The formula of the Sigmoid function is , the output value of this function represents the probability of bubble occurrence. For example, if the output layer is calculated and the value before Sigmoid function processing is 2, after Sigmoid function calculation, , the model predicts that the probability of bubble occurrence at this time is 0.88, that is, there is an 88% probability of bubble problem.

[0049] Through the collaborative work of the above input layer, hidden layer and output layer, the bubble occurrence prediction model can effectively utilize the feature data set and accurately predict the probability of tire toe bubble occurrence.

[0050] Embodiment 2: The equipment status monitoring module builds an equipment status monitoring model based on the random forest (RF) algorithm. The training steps are implemented as follows: E1: Prepare training set and labels Use the preprocessed feature data set as the training set. These feature data are extracted from the equipment operation data and are closely related to the equipment status. Select the equipment status as the label. The equipment status can be normal, abnormal, or a specific fault type, which is defined and classified according to actual needs.

[0051] E2: Determine the parameters of the random forest Determine key parameters of the random forest model, including the number of trees (i.e., the total number of decision trees in the forest), the maximum depth of each tree (i.e., the maximum number of layers a decision tree can grow), and the minimum number of sample splits (i.e., the minimum number of samples required for a decision tree node to split).

[0052] The choice of these parameters can be determined based on experience, experiments, or cross-validation. For example, the number of trees can be set to 100, 200, or more, the maximum depth can be set to 10, 20, or unlimited, and the minimum number of sample splits can be set to 2, 5, or more.

[0053] E3: Training a Random Forest Model The random forest model is trained using the training set. During the training process, each tree grows according to the feature dataset and labels, and splits by selecting the best feature until the maximum depth is reached or other stopping conditions are met.

[0054] The random forest model improves the accuracy and stability of the model by integrating the output results of multiple decision trees. The training process will continue until the model converges, that is, the model performance no longer improves significantly or reaches the preset training rounds.

[0055] E4: Verify model performance The trained random forest model is validated using the validation set. The validation set is a dataset independent of the training set and is used to evaluate the classification accuracy and generalization ability of the model.

[0056] Classification accuracy indicates the proportion of samples that the model correctly classifies. Generalization ability indicates the model's ability to predict unseen data. The performance of the model can be evaluated through indicators such as confusion matrix, accuracy, precision, and recall.

[0057] E5: Save model parameters When the model performance reaches the preset standard, such as the classification accuracy reaches or exceeds the threshold and the generalization ability is good, the model parameters are saved.

[0058] The saved model parameters include the structural information of each tree (such as node splitting characteristics, thresholds, leaf node outputs, etc.) and the overall parameters of the random forest (such as the number of trees, maximum depth, etc.). These parameters can be used for subsequent model loading and prediction to achieve real-time monitoring and evaluation of device status.

[0059] The following describes in detail the implementation process of the equipment status monitoring model based on the random forest (RF) algorithm to process high-dimensional data and accurately classify and identify abnormal equipment status, with specific examples.

[0060] Assume that in a tire manufacturing plant, a large amount of data will be generated when the tire toe bubble generating equipment is running, covering multiple dimensions such as equipment working pressure, temperature, gas flow, toe size, equipment vibration frequency, operating time and maintenance records. After these data are collected by the data collection module, they enter the data preprocessing module for cleaning, normalization, outlier detection and missing value filling to form a data set suitable for model analysis. After that, the feature extraction module extracts statistical features, time domain features and frequency domain features from the preprocessed data set to form a high-dimensional feature data set.

[0061] The input layer receives the feature data set after the above processing. For example, after feature extraction, the tire manufacturing plant obtains a data set containing 50 features. These features reflect the operating status of the equipment from different aspects. The input layer is responsible for receiving these 50-dimensional feature data.

[0062] The decision tree layer constructs multiple decision trees. When constructing each decision tree, some features and some samples are randomly selected from the feature data set for training. For example, for the first decision tree, 20 features are randomly selected from 50 features, and a certain proportion of samples are randomly selected from a large amount of sample data. The decision tree splits according to the input features. For example, taking the device temperature as an example, if the temperature is greater than a certain threshold (this threshold is determined by the algorithm during the training process), the sample is divided into one branch; if it is less than the threshold, it is divided into another branch, thereby forming a decision rule. Each decision tree divides the data step by step in this way until a certain stopping condition is met, such as reaching the maximum depth or the number of samples in the node is less than a certain set value.

[0063] The integration layer integrates the output results of all decision trees. Since each decision tree is trained based on different features and samples, their classification results may be different. The integration layer determines the final device status through a voting mechanism. For example, if there are 50 decision trees, 30 of which determine that the current status of the device is abnormal and 20 are normal, then according to the voting results, the device status is finally determined to be abnormal.

[0064] The output layer outputs the equipment status monitoring results, including normal operation, abnormal operation or fault warning. In the example of this tire manufacturing plant, if the integration layer determines that the equipment status is abnormal, the output layer will clearly give the result of "abnormal operation" and further determine whether a fault warning needs to be issued based on the preset rules.

[0065] The feature data set after preprocessing and feature extraction is used as the training set, and the device status is selected as the label. In actual scenarios, the device status can be determined based on historical experience, historical fault records, etc. For example, the status of the device running stably for a period of time without abnormalities related to bubble problems is marked as "normal", and the status of the device with abnormal pressure fluctuations that increase the risk of bubble generation is marked as "abnormal".

[0066] Determine the key parameters of the random forest model, such as the number of trees, maximum depth, and minimum number of sample splits. Through multiple experiments and analysis, in this tire manufacturing scenario, it was determined that the number of trees was 100, the maximum depth of each tree was 15 layers, and the minimum number of sample splits was 5. These parameters were selected to avoid overfitting of the model and improve the generalization ability of the model while ensuring the accuracy of the model. Use the training set to train the random forest model. During the training process, each decision tree grows and splits according to the given features and samples. As the training progresses, the model continuously adjusts the structure and parameters of the decision tree until the model converges, that is, the performance of the model is no longer significantly improved. For example, in the early stage of training, the model's recognition accuracy of abnormal equipment status may be only 60%. As the number of training rounds increases, after reaching a certain number of rounds, the accuracy stabilizes at around 90%, and there is no longer a significant improvement. At this time, the model is considered to have converged.

[0067] Use a validation set independent of the training set to validate the trained random forest model. The performance of the model is evaluated by calculating indicators such as confusion matrix, accuracy, precision, and recall. Assume that there are 100 samples in the validation set, of which 30 are actually abnormal samples and 70 are normal samples. The model correctly predicts 25 abnormal samples and 65 normal samples, then accuracy = (25 + 65) / 100 = 90%, precision = 25 / (25 + 5) = 83.3% (where 5 is the number of normal samples misjudged as abnormal), and recall = 25 / 30 = 83.3%. When these performance indicators reach the preset standards (such as accuracy reaching more than 90%), the model performance is considered good. When the model performance reaches the preset standards, save the model parameters, including the structural information of each tree (such as node splitting features, thresholds, leaf node outputs, etc.) and the overall parameters of the random forest (such as the number of trees, maximum depth, etc.). Through the above process of constructing the model structure and training, the equipment condition monitoring model based on the random forest (RF) algorithm can effectively process high-dimensional data and accurately classify and identify the abnormal status of the equipment causing tire toe bubble.

[0068] Embodiment 3: The parameter optimization model is built based on the genetic algorithm (GA) to find the optimal combination of equipment operating parameters to minimize the probability of bubble occurrence and maximize equipment operating efficiency. The training steps are specifically implemented as follows: F1: Define the optimization objective function Determine the optimization objective function, which comprehensively considers the two indicators of bubble occurrence probability and equipment operation efficiency. The objective function can be defined as: ; Where x represents the device operating parameter vector, represents the probability of bubble occurrence under parameter x, It represents the operating efficiency of the equipment under parameter x, α and β are the weight coefficients of bubble occurrence probability and operating efficiency, respectively, which are used to balance the relative importance of the two indicators.

[0069] The optimization goal is to find the parameter x that minimizes the objective function.

[0070] F2: Initialize the population Initialize a population, each individual in the population represents a set of device working parameters.

[0071] An individual can be represented as a vector, and each element of the vector corresponds to a device operating parameter, such as temperature, pressure, speed, etc. The size of the population can be determined according to the complexity of the problem and computing resources, and is usually set to dozens to hundreds of individuals.

[0072] F3: Evaluate the fitness of each individual in the population Each individual in the population is evaluated and its fitness value, that is, the value of the objective function, is calculated.

[0073] The smaller the fitness value, the better the performance of the individual, that is, the lower the probability of bubble occurrence and the higher the equipment operation efficiency.

[0074] The probability of bubble occurrence and equipment operating efficiency can be estimated using simulation experiments, historical data, or expert experience.

[0075] F4: Selection, crossover and mutation operations The individuals with the highest fitness are selected for crossover and mutation operations to generate a new generation of population.

[0076] The selection operation can adopt methods such as roulette selection and tournament selection to ensure that individuals with higher fitness have a greater chance of being selected.

[0077] The crossover operation can exchange some gene fragments of two individuals to generate a new individual.

[0078] The mutation operation can randomly change certain genes of individuals to increase the diversity of the population.

[0079] F5: Repeat the iterative process Repeat steps F3 to F4 until the preset number of iterations is reached or the optimal solution is found.

[0080] The number of iterations can be determined based on the complexity of the problem and computing resources, and is usually set to several hundred to several thousand times.

[0081] The optimal solution refers to the individual that minimizes the objective function value, that is, the optimized combination of equipment operating parameters.

[0082] F6: Output the best individual Output the optimal individual, that is, the optimized combination of equipment working parameters.

[0083] The optimal individuals can be used as recommended parameters for equipment operation to minimize the probability of bubble occurrence and maximize equipment operation efficiency.

[0084] Embodiment 4: The decision support module aims to integrate the outputs of the bubble occurrence prediction model, equipment status monitoring model and parameter optimization model to provide scientific equipment control strategies for equipment operators or control systems. The specific implementation steps are as follows: S1: Define model output The probability of the bubble occurrence prediction model output is defined as P_Bubble, which represents the probability of bubble occurrence under the current equipment working parameters.

[0085] The output status of the equipment status monitoring model is defined as S_Status, which indicates the current operating status of the equipment, such as normal, abnormal, or specific fault type.

[0086] The optimization parameter output by the parameter optimization model is defined as P_Optimized, which is represented by the combination of equipment operating parameters that minimizes the probability of bubble occurrence and maximizes the equipment operation efficiency.

[0087] S2: Determine the risk level of bubble occurrence The risk level of bubble occurrence is determined based on P_Bubble and the preset threshold.

[0088] For example, multiple thresholds can be set, such as low risk (P_Bubble < 0.3), medium risk (0.3 ≤ P_Bubble< 0.7), and high risk (P_Bubble ≥ 0.7).

[0089] According to the value of P_Bubble, the risk level of bubble occurrence is determined to provide a basis for subsequent decision-making.

[0090] S3: Evaluate the current operating status of the device Combined with S_Status, evaluate the current operating status of the device.

[0091] If S_Status indicates that the equipment is abnormal or faulty, an early warning signal is issued to remind the equipment operator or control system to pay attention to the equipment status.

[0092] Early warning signals can be sent to relevant personnel or systems through sound, light or email.

[0093] S4: Adjust equipment operating parameters According to P_Optimized, the operating parameters of the equipment are adjusted to reduce the risk of bubble occurrence.

[0094] Apply the optimized parameters in P_Optimized to the equipment, such as adjusting the temperature, pressure, speed and other parameters to reduce the probability of bubble occurrence.

[0095] During the adjustment process, it is necessary to ensure that the equipment operates within a safe range to avoid equipment damage or safety accidents caused by parameter adjustment.

[0096] S5: Generate device control strategy report Generate equipment control strategy report based on risk level, equipment status and optimization parameters.

[0097] The report content may include the risk level of bubble occurrence, the current status of the equipment, optimization parameter recommendations and corresponding control measures.

[0098] The report format can be text, chart or spreadsheet, etc., which is easy for equipment operators or control systems to understand and implement.

[0099] S6: Send control strategy report Send control strategy reports to equipment operators or control systems to implement appropriate control measures. Reports can be sent via email, internal network or control system interface.

[0100] The equipment operator or control system takes corresponding control measures based on the report content, such as adjusting equipment parameters, performing repairs or replacing parts, to ensure the safe and efficient operation of the equipment.

[0101] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device.

[0102] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A predictive analysis system for controlling a tire toe bubble generating device, characterized in that: The system comprises: Data collection module: collects the operating data of the tire toe bubble generating device; Data preprocessing module: preprocess the collected data, including data cleaning, normalization, outlier detection and missing value filling, to form a preprocessed data set; Feature extraction module: extracts features from the preprocessed data set to form a feature data set; Bubble Occurrence Prediction Module: Build a bubble occurrence prediction model based on the deep neural network (DNN) algorithm. The model is trained using feature data sets to predict the probability of bubble occurrence. Equipment status monitoring module: Build an equipment status monitoring model based on the random forest RF algorithm. This model uses the feature data in the preprocessed data set to identify abnormal status of equipment operation. Parameter optimization module: Build a parameter optimization model based on the genetic algorithm (GA) algorithm. This model uses historical operating data and equipment parameters to optimize the equipment's operating parameters to reduce the occurrence of bubbles. Model output integration module: input the characteristic data set into the bubble occurrence prediction model, equipment status monitoring model and parameter optimization model respectively to obtain the model output results; Decision support module: Generates equipment control strategy reports based on model output results, including bubble occurrence risk, equipment status assessment, and optimized operating parameters.

2. A predictive analysis system for controlling a tire toe bubble generating device according to claim 1, characterized in that: The operation data includes equipment working pressure, temperature, gas flow, toe size, equipment vibration frequency, operation time and maintenance records; the feature data set includes statistical features, time domain features and frequency domain features.

3. A predictive analysis system for controlling a tire toe bubble generating device according to claim 2, characterized in that: The bubble occurrence prediction model adopts a multi-layer deep neural network structure, extracts features through multiple fully connected layers and predicts the probability of bubble occurrence. Its network structure includes: Input layer: receives data from the feature dataset, and the feature vector dimension is [N], where N is the number of features; Hidden layer: multiple layers of fully connected layers are used, each layer is followed by a ReLU activation function to extract deep features; Output layer: Outputs the probability of bubble occurrence and uses the Sigmoid function to map the output value to the [0,1] interval.

4. A predictive analysis system for controlling a tire toe bubble generating device according to claim 3, characterized in that: The training steps of the bubble generation prediction model include: D1: Initialize the parameters of the DNN model, including weights, bias terms, and learning rates; D2: Divide the preprocessed feature data set into a training set and a validation set. The training set is used for model training, and the validation set is used for model validation. D3: Use the training set to iteratively train the DNN model. In each iteration, the model output is calculated by forward propagation, and the loss between the predicted value and the true value is calculated according to the loss function; the loss function adopts the binary cross entropy loss function; D4: Update the model parameters based on the loss value through the back propagation algorithm and gradient descent method; D5: Repeat steps D3 to D4 until the loss converges or the preset number of iterations is reached; D6: Use the validation set to validate the trained DNN model and evaluate the accuracy, precision, and recall of the model; D7: When the model performance reaches the preset standard, the model parameters are saved to obtain the trained bubble occurrence prediction model.

5. A prediction and analysis system for controlling a tire toe bubble generating device according to claim 2, characterized in that: The equipment status monitoring model realizes the monitoring of the equipment operation status by constructing multiple decision trees and integrating their output results. Its structure includes: Input layer: receives data from feature datasets; Decision tree layer: construct multiple decision trees, each of which is split according to the input features to form decision rules; Integration layer: Integrate the output results of all decision trees and determine the final device status through a voting mechanism; Output layer: Output equipment status monitoring results, including normal operation, abnormal operation or fault warning.

6. A predictive analysis system for controlling a tire toe bubble generating device according to claim 5, characterized in that: The steps to train the equipment condition monitoring model include: E1: Use the feature dataset as the training set and select the device status as the label; E2: Determine the parameters of the random forest, including the number of trees, maximum depth, and minimum number of sample splits; E3: Use the training set to train the random forest model until the model converges; E4: Use the validation set to validate the random forest model and evaluate the classification accuracy and generalization ability of the model; E5: When the model performance reaches the preset standard, the model parameters are saved to obtain the trained equipment status monitoring model.

7. A predictive analysis system for controlling a tire toe bubble generating device according to claim 2, characterized in that: The training steps of the parameter optimization model include: F1: Define the optimization objective function, including minimizing the probability of bubble occurrence and maximizing equipment operation efficiency; F2: Initialize the population, each individual represents a set of equipment working parameters; F3: Evaluates the fitness of each individual in the population, indicating the value of the objective function; F4: Select the individuals with the highest fitness to perform crossover and mutation operations to generate a new generation of population; F5: Repeat steps F3 to F4 until the preset number of iterations is reached or the optimal solution is found; F6: Output the best individual, indicating the optimized equipment working parameters.

8. The predictive analysis system for controlling the tire toe bubble generating device according to claim 1, characterized in that: The specific implementation steps of the decision support module include: S1: Define the probability output by the bubble occurrence prediction model as P_Bubble, the status output by the equipment status monitoring model as S_Status, and the optimization parameter output by the parameter optimization model as P_Optimized; S2: Determine the risk level of bubble occurrence based on P_Bubble and the preset threshold; S3: Combined with S_Status, evaluate the current operating status of the equipment and issue an early warning if there is an abnormality; S4: According to P_Optimized, adjust the operating parameters of the equipment to reduce the risk of bubble generation; S5: Generate equipment control strategy report based on risk level, equipment status and optimization parameters; S6: Send the control strategy report to the equipment operator or control system to implement corresponding control measures.

9. A predictive analysis system for controlling a tire toe bubble generating device according to claim 8, characterized in that: The risk level output by the decision support module is divided into four levels: low risk, medium risk, high risk, and extremely high risk.

10. A prediction and analysis system for controlling a tire toe bubble generating device according to claim 9, characterized in that: It also includes: an early warning release module, which automatically selects the corresponding early warning information release method according to the risk level output by the policy support module, including SMS notification, email notification, sound and light alarm and voice alarm.

Images