Packer predictive maintenance method and system based on Internet of Things

Abstract

The invention discloses a packer predictive maintenance method and system based on the Internet of Things, and relates to the technical field of equipment maintenance, and the method comprises the following steps: obtaining original machine state data of a packer and original emotion data of maintenance personnel, and carrying out the preprocessing to obtain preprocessed machine data and preprocessed emotion data; uploading the preprocessed machine data and the preprocessed emotion data to a cloud end by using an Internet of Things technology, and extracting key machine features and key emotion features in the preprocessed machine data and the preprocessed emotion data; effective information in the key machine features and the key emotion features is screened, and a comprehensive feature vector matrix is constructed; and training a predictive maintenance model based on the comprehensive feature vector matrix, and formulating a maintenance strategy of the packer. According to the method, the health state of the packer can be predicted more accurately, and the relationship between the equipment state and the emotion of the maintainer can be deeply excavated through multi-dimensional correlation analysis.

Description

Technical Field

[0001] The present invention relates to the technical field of equipment maintenance, and in particular to a predictive maintenance method and system for a baler based on the Internet of Things. Background Art

[0002] The baler is an automated device used in the packaging industry to bundle products or items for transportation, storage or sale. Balers are widely used in food, medicine, textiles, postal services, logistics and other industries. By monitoring the operating status and performance parameters of the baler equipment, potential failures can be predicted so that maintenance can be performed before the failure occurs. Through predictive maintenance, maintenance activities can be planned in advance, the allocation of maintenance resources can be optimized, unexpected downtime and production losses can be reduced, and the reliability and life of the equipment can be improved.

[0003] IoT technology can be used to remotely monitor and collect the operating data of the baler in real time, such as temperature, vibration, power consumption, operating time and other key performance indicators. These data are transmitted to the cloud or data center through IoT devices for analysis and processing. Based on these data, data analysis and machine learning algorithms can be used to build a predictive model to monitor the equipment status in real time, predict possible future failures, and formulate corresponding maintenance plans based on the prediction results.

[0004] For example, Chinese patent 202010278878.7 discloses a predictive maintenance method and maintenance system for mechanical equipment, which uses an artificial intelligence engine to analyze and process vibration acceleration amplitude data to obtain labeled fault features, and performs maintenance on the equipment under test based on the labeled fault features or original temperature data. However, the above method still has the following shortcomings: it does not take into account the emotional data of maintenance personnel. As an important auxiliary variable, emotional data, if not taken into account, is not conducive to improving the accuracy of fault prediction, and thus it is impossible to scientifically formulate maintenance strategies.

[0005] Currently, no effective solution has been proposed for the problems in the related technologies. Summary of the invention

[0006] In view of the problems in the related technology, the present invention proposes a predictive maintenance method and system for a baler based on the Internet of Things to overcome the above-mentioned technical problems existing in the existing related technology.

[0007] To this end, the specific technical solution adopted by the present invention is as follows: According to one aspect of the present invention, a predictive maintenance method for a baler based on the Internet of Things is provided, and the predictive maintenance method for a baler based on the Internet of Things comprises the following steps: S1. Obtain original machine status data of the baler and original emotion data of the maintenance personnel, and preprocess them to obtain preprocessed machine data and preprocessed emotion data.

[0008] S2. Use the Internet of Things technology to upload the preprocessed machine data and the preprocessed emotion data to the cloud, and extract key machine features and key emotion features from the preprocessed machine data and the preprocessed emotion data.

[0009] S3. Filter the effective information in key machine features and key emotional features, and construct a comprehensive feature vector matrix.

[0010] S4. Train the predictive maintenance model based on the comprehensive feature vector matrix and formulate the maintenance strategy for the baler.

[0011] Among them, screening effective information from key machine features and key emotional features and constructing a comprehensive feature vector matrix includes the following steps: S31. Combine key machine features and key emotional features to construct an initial feature matrix and determine the target variable; S32, using the initial feature matrix and target variables to train a random forest model as a baseline model; S33. Use cross-validation to evaluate the performance of the random forest model; S34, permuting the features in each initial feature matrix and selecting features according to the permutation test results; S35. Calculate the Pearson correlation coefficient between the selected features, and eliminate redundant features based on the correlation matrix to obtain a comprehensive feature vector matrix.

[0012] Further, uploading the preprocessed machine data and the preprocessed emotion data to the cloud using the Internet of Things technology, and extracting key machine features and key emotion features from the preprocessed machine data and the preprocessed emotion data includes the following steps: S21. Verify the quality of the pre-processed machine data and sentiment data, and convert them into a format for IoT transmission; S22, uploading the formatted data to the cloud server through the IoT device; S23. Extract key machine features and key emotional features.

[0013] Further, permutation testing each feature in the initial feature matrix and selecting features according to the permutation test results include the following steps: S341. Use each decision tree in the random forest model to evaluate the features in the initial feature matrix and use the average reduction impurity to measure the importance of the features. S342. For each feature, aggregate the average importance score of the feature in all decision trees; S343. Select a feature and randomly shuffle the value of the feature on the validation set, while keeping other features unchanged; S344. Retrain the random forest model using the randomly shuffled data set; S345, obtaining the change in performance of the retrained random forest model compared to the baseline model; S346, repeat the permutation test for all features to evaluate the impact of each feature on the performance of the random forest model; S347. Analyze the degree of degradation of the performance of the random forest model after each feature replacement. According to the degree of performance degradation, rank the features by importance and select several features with the highest ranking.

[0014] Further, calculating the Pearson correlation coefficient between the selected features and eliminating redundant features based on the correlation matrix to obtain a comprehensive feature vector matrix includes the following steps: S351, calculating the Pearson correlation coefficient between the selected features and constructing a correlation matrix; S352, eliminating redundant features according to the correlation matrix; S353. Construct a comprehensive feature vector matrix based on the results after redundant features are eliminated.

[0015] Furthermore, training a predictive maintenance model based on the comprehensive feature vector matrix and formulating a maintenance strategy for the baler includes the following steps: S41, obtaining labels corresponding to each feature in the comprehensive feature vector matrix, including machine health status prediction labels and emotion impact assessment labels, and combining the comprehensive feature vector matrix, machine health status prediction labels and emotion impact assessment labels to obtain input variables and target variables; S42, based on the machine learning algorithm, and using the input variables and the target variables to train the corresponding machine learning model to obtain a predictive maintenance model; S43, obtaining new input variables and inputting them into the predictive maintenance model to obtain the health status prediction result of the baler at each time point; S44, time-aligning the prediction results with the emotional data of the maintenance personnel to form a comprehensive data set including the health status prediction of the baler and the emotional status of the maintenance personnel; S45, multi-dimensional correlation analysis integrated the data set to obtain potential emotional threat factors in the maintenance process; S46. Formulate a maintenance strategy for the baler based on the output of the predictive maintenance model and the results of the multi-dimensional correlation analysis.

[0016] Furthermore, based on the machine learning algorithm, and using the input variables and the target variables to train the corresponding machine learning model, obtaining the predictive maintenance model includes the following steps: S421, divide the input variables and the corresponding target variables to obtain a training set and a test set; S422, selecting a support vector machine algorithm, and using the training set data to train the support vector machine algorithm model, and using a hybrid optimization algorithm to optimize the support vector machine algorithm model parameters to obtain a predictive maintenance model; S423, verifying the predictive maintenance model using the test set, and optimizing the parameters of the predictive maintenance model according to the evaluation results; Among them, the decision function of the predictive maintenance model is: ; In the formula, f ( x ) represents the prediction of new input variables; α i Indicates i The Lagrange multiplier of training samples, y i Indicates i The labels of the support vectors, K represents the kernel function; b represents the bias term; N represents the number of support vectors.

[0017] Further, optimizing the support vector machine algorithm model parameters using the hybrid optimization algorithm includes the following steps: S4221. Randomly generate a particle swarm of a particle swarm optimization algorithm and a chromosome population of a genetic algorithm, and each individual represents a parameter combination of a support vector machine algorithm model; S4222, determining a fitness function, and encoding and decoding chromosomes; S4223, running particle swarm optimization algorithm and genetic algorithm; S4224, combining the population processed by the genetic algorithm with the position update rule of particles in the particle swarm optimization algorithm; S4225. Repeat steps S4223 to S4224 until the stop condition is met.

[0018] Furthermore, the multi-dimensional correlation analysis integrated the data set to obtain the potential emotional threat factors in the maintenance process, including the following steps: S451, calculating the correlation coefficient between the predicted health state of the baler and the emotional state of the maintenance personnel using the Pearson correlation coefficient; S452. Determine the emotional features that are significantly associated with the health status of the baler equipment through a significance test; S453, clustering the comprehensive data set using a clustering algorithm; S454. Through clustering results, the association pattern between high-risk emotional state and baler failure is identified.

[0019] Furthermore, formulating a maintenance strategy for the baler based on the output of the predictive maintenance model and the multi-dimensional correlation analysis results includes the following steps: S461. Monitor the emotional characteristics that are significantly associated with the health status of the baler equipment. If the significant emotional characteristic values ​​of the maintenance personnel are beyond the normal range, take preventive intervention measures; S462. Dynamically adjust the allocation of maintenance tasks according to the emotional state of the maintenance personnel.

[0020] According to another aspect of the present invention, a predictive maintenance system for a baler based on the Internet of Things is also provided. The predictive maintenance system for a baler based on the Internet of Things includes a raw data acquisition module, a key feature extraction module, a comprehensive feature construction module and a maintenance strategy formulation module; wherein the raw data acquisition module, the key feature extraction module, the comprehensive feature construction module and the maintenance strategy formulation module are connected in sequence.

[0021] The original data acquisition module is used to acquire the original machine status data of the baler and the original emotional data of the maintenance personnel, and preprocess them to obtain the preprocessed machine data and preprocessed emotional data.

[0022] The key feature extraction module is used to upload the preprocessed machine data and the preprocessed emotion data to the cloud by using the Internet of Things technology, and to extract the key machine features and key emotion features from the preprocessed machine data and the preprocessed emotion data.

[0023] The comprehensive feature construction module is used to filter the effective information in the key machine features and key emotional features, and construct a comprehensive feature vector matrix.

[0024] The maintenance strategy formulation module is used to train the predictive maintenance model based on the comprehensive feature vector matrix and formulate the maintenance strategy of the baler.

[0025] The beneficial effects of the present invention are: (1) The present invention provides a predictive maintenance method and system for balers based on the Internet of Things, which ensures the integrity of information by collecting machine status data and maintenance personnel's emotional data. The data is collected in real time through Internet of Things devices to ensure the timeliness of the data, and cloud storage technology is used to provide an efficient and secure storage solution.

[0026] (2) The present invention provides a more comprehensive data perspective by combining the machine status and maintenance personnel's emotional characteristics, which helps to capture more potential influencing factors. The random forest algorithm can provide a preliminary assessment of feature importance and provide guidance for feature selection. The true importance of features is evaluated by observing the impact of random permutation of features on model performance. By calculating the correlation between features and eliminating highly correlated features, the redundancy between features can be reduced, the generalization ability of the model can be improved, and the retained features are more representative and discriminative, which helps to explain the model's prediction results and makes the model decision more transparent and credible.

[0027] (3) The predictive maintenance model based on comprehensive feature vector matrix training can more accurately predict the health status of the baler. Through multi-dimensional correlation analysis, it can deeply explore the relationship between the equipment status and the emotions of the maintenance personnel and identify potential threat factors. Combining the output of the predictive model and the results of the correlation analysis, a scientific maintenance strategy can be formulated to improve the accuracy and rationality of decision-making. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0029] Figure 1 is a flow chart of a predictive maintenance method for a baler based on the Internet of Things according to an embodiment of the present invention; Figure 2 It is a principle block diagram of a baler predictive maintenance system based on the Internet of Things according to an embodiment of the present invention.

[0030] In the figure: 1. Original data acquisition module; 2. Key feature extraction module; 3. Comprehensive feature construction module; 4. Maintenance strategy formulation module. DETAILED DESCRIPTION

[0031] To further illustrate each embodiment, the present invention provides drawings, which are part of the disclosure of the present invention and are mainly used to illustrate the embodiments and can be used in conjunction with the relevant descriptions in the specification to explain the operating principles of the embodiments. With reference to these contents, ordinary technicians in the field should be able to understand other possible implementations and advantages of the present invention. The components in the figures are not drawn to scale, and similar component symbols are generally used to represent similar components.

[0032] According to an embodiment of the present invention, a method and system for predictive maintenance of a baler based on the Internet of Things are provided.

[0033] The present invention is further described with reference to the accompanying drawings and specific embodiments. Figure 1 As shown, according to one embodiment of the present invention, a predictive maintenance method for a baler based on the Internet of Things is provided, and the predictive maintenance method for a baler based on the Internet of Things comprises the following steps: S1. Obtain original machine status data of the baler and original emotion data of the maintenance personnel, and preprocess them to obtain preprocessed machine data and preprocessed emotion data.

[0034] Specifically, various sensors (such as temperature sensors, vibration sensors, etc.) installed on the baler collect machine status data in real time: Use emotion recognition tools (such as questionnaires, voice emotion analysis, etc.) to collect emotional data of maintenance personnel.

[0035] Filter out noise and invalid data to ensure data accuracy and completeness. Convert data into a unified format and unit for subsequent analysis.

[0036] S2. Use the Internet of Things technology to upload the preprocessed machine data and the preprocessed emotion data to the cloud, and extract key machine features and key emotion features from the preprocessed machine data and the preprocessed emotion data.

[0037] In a further embodiment, uploading the preprocessed machine data and the preprocessed emotion data to the cloud using the Internet of Things technology, and extracting key machine features and key emotion features from the preprocessed machine data and the preprocessed emotion data includes the following steps: S21. Verify the quality of pre-processed machine data and sentiment data, and convert them into formats for IoT transmission, such as JSON, XML, etc., to ensure that the uploaded data is of high quality and in a unified format.

[0038] S22. Upload the formatted data to the cloud server through an IoT device (such as a gateway, a router, etc.).

[0039] S23. Extract key machine features and key emotional features. The extraction of key features can reduce the amount of data and improve the efficiency of data analysis.

[0040] S3. Filter the effective information in key machine features and key emotional features, and construct a comprehensive feature vector matrix.

[0041] In a further embodiment, screening effective information in key machine features and key emotion features and constructing a comprehensive feature vector matrix includes the following steps: S31. Combine key machine features and key emotion features to construct an initial feature matrix and determine the target variable. The target variable is the object of model prediction, such as predicting the remaining service life of the baler or predicting the time of the next maintenance or failure.

[0042] S32. Use the initial feature matrix and target variables to train a random forest model as a baseline model. Random forest is an ensemble learning method that improves the generalization ability of the model by building multiple decision trees and voting. As a baseline model, the random forest model can provide a preliminary performance evaluation benchmark.

[0043] S33. Use cross-validation to evaluate the performance of the random forest model. Cross-validation can divide the data set into multiple subsets, cyclically use one of the subsets as the validation set, and the remaining subsets as the training set, and finally take the average performance as the evaluation result. Cross-validation effectively reduces overfitting and provides more stable and reliable model performance evaluation results.

[0044] S34. Permutation test each feature in the initial feature matrix, and select features that have a significant impact on model performance based on the permutation test results; permutation test is a feature selection method that evaluates the importance of each feature to the model by randomly shuffling the value of each feature and observing the changes in model performance. Select features that have a significant impact on model performance.

[0045] S35. Calculate the Pearson correlation coefficient between the selected features, and remove redundant features based on the correlation matrix to obtain a comprehensive feature vector matrix. Removing redundant features reduces the multicollinearity problem between features, simplifies the model, and improves the stability and generalization ability of the model.

[0046] For example, extract key machine features from the baler dataset, such as temperature, vibration frequency, pressure, etc. Extract key emotional features from the maintenance personnel dataset, such as stress level, fatigue index, satisfaction, etc. Construct an initial feature matrix where each row represents all feature values ​​at a time point. The target variable is the remaining useful life of the baler.

[0047] Train a random forest model using the initial feature matrix and the target variable, perform permutation tests on each feature to evaluate its impact on model performance, calculate the Pearson correlation coefficient between the selected features, construct a correlation matrix, and remove highly correlated redundant features.

[0048] In a further embodiment, permutation testing the features in each initial feature matrix and selecting features based on the permutation test results comprises the following steps: S341. Use each decision tree in the random forest model to evaluate the features in the initial feature matrix, and use the average reduction impurity to measure the importance of the features, and calculate the reduction of each feature to the impurity of the classification node (such as Gini impurity or information gain). By calculating the average of these reductions, the importance of each feature can be measured.

[0049] S342. For each feature, aggregate the average importance score of the feature in all decision trees to evaluate its overall contribution to the model.

[0050] S343. Select a feature and randomly shuffle (permute) the value of the feature on the validation set, while keeping other features unchanged.

[0051] S344. Retrain the random forest model using the randomly shuffled dataset and evaluate the model performance, such as accuracy, F1 score, etc. By retraining the model and evaluating the performance, we can understand the impact of the feature replacement on the model performance.

[0052] S345. Obtain the change in performance of the retrained random forest model compared to the baseline model.

[0053] S346. Repeat the permutation test for all features to evaluate the impact of each feature on the performance of the random forest model.

[0054] S347. Analyze the degree of decline in the performance of the random forest model after each feature replacement to determine the importance of the feature. According to the degree of performance decline, rank the features by importance and select the top-ranked features. Through sorting and selection, the features that contribute most to the model performance can be screened out, and irrelevant or minor features can be removed to improve the efficiency and accuracy of the model.

[0055] In a further embodiment, calculating the Pearson correlation coefficient between the selected features and eliminating redundant features based on the correlation matrix to obtain a comprehensive feature vector matrix includes the following steps: S351, calculating the Pearson correlation coefficient between the selected features and constructing a correlation matrix; It should be noted that the Pearson correlation coefficient is a statistical indicator that measures the linear relationship between two variables, with a value range of -1 to 1. Calculate the Pearson correlation coefficients between the selected features and organize these coefficients into a correlation matrix. The correlation matrix is ​​a symmetric matrix in which each element represents the correlation coefficient between two features. Constructing a correlation matrix can help understand the linear relationship between features and identify which features are highly correlated, thereby providing a basis for subsequent redundant feature elimination.

[0056] S352. Eliminate highly correlated redundant features according to the correlation matrix, and identify highly correlated feature pairs (usually the absolute value of the correlation coefficient is greater than a certain threshold, such as 0.8 or 0.9) according to the correlation coefficient value in the correlation matrix. For each pair of highly correlated features, retain one feature and eliminate the other redundant feature.

[0057] S353. Construct a comprehensive feature vector matrix based on the results after redundant features are eliminated.

[0058] S4. Train the predictive maintenance model based on the comprehensive feature vector matrix and formulate the maintenance strategy for the baler.

[0059] In a further embodiment, training a predictive maintenance model based on the comprehensive feature vector matrix and formulating a maintenance strategy for a baler includes the following steps: S41. Obtain the label corresponding to each feature in the comprehensive feature vector matrix, including the machine health status prediction label (health status of the baler) and the emotion impact assessment label (such as stress, satisfaction), and combine the comprehensive feature vector matrix, the machine health status prediction label and the emotion impact assessment label to obtain the input variable (feature) and the target variable (label).

[0060] S42. Based on the machine learning algorithm, the corresponding machine learning model is trained using the input variables and the target variables to obtain a predictive maintenance model.

[0061] S43. Obtain new input variables and input them into the predictive maintenance model to obtain the prediction results of the health status of the baler at each time point; that is, the machine data and emotion data collected in real time or regularly are input into the trained predictive maintenance model to obtain the prediction results of the health status of the baler at each time point.

[0062] S44, time-aligning the prediction results with the emotional data of the maintenance personnel to form a comprehensive data set including the health status prediction of the baler and the emotional status of the maintenance personnel.

[0063] S45. Multidimensional correlation analysis integrates the data set to obtain potential emotional threat factors in the maintenance process.

[0064] S46. Formulate a maintenance strategy for the baler based on the output of the predictive maintenance model and the results of the multi-dimensional correlation analysis.

[0065] In a further embodiment, based on a machine learning algorithm, and using input variables and target variables to train a corresponding machine learning model, obtaining a predictive maintenance model includes the following steps: S421. Divide the input variables and the corresponding target variables to obtain a training set and a test set.

[0066] S422, selecting a support vector machine algorithm, and using the training set data to train the support vector machine algorithm model, and using a hybrid optimization algorithm to optimize the support vector machine algorithm model parameters, namely, the penalty coefficient and the kernel parameter, to obtain a predictive maintenance model.

[0067] S423. Use the test set to verify the predictive maintenance model, and optimize the parameters of the predictive maintenance model based on the evaluation results.

[0068] Among them, the decision function of the predictive maintenance model is: ; In the formula, f ( x ) represents the prediction of new input variables; α i Indicates i The Lagrange multiplier of training samples is a weight coefficient used to balance classification error and model complexity. During the training process of SVM, the optimal solution is found by solving the Lagrange multiplier. y i Indicates i The labels of the support vectors, K represents the kernel function; b represents the bias term; N represents the number of support vectors.

[0069] In a further embodiment, optimizing the support vector machine algorithm model parameters using a hybrid optimization algorithm comprises the following steps: S4221. Randomly generate a particle swarm of the particle swarm optimization algorithm and a chromosome population of the genetic algorithm, and each individual represents a parameter combination of the support vector machine algorithm model.

[0070] S4222. Determine the fitness function, usually using the performance indicators of the support vector machine algorithm model on the validation set, such as mean square error (MSE), accuracy or other related indicators, and encode and decode chromosomes.

[0071] S4223. Run the particle swarm optimization algorithm and the genetic algorithm.

[0072] The fitness value of each particle is calculated, and the individual optimum and global optimum are updated. The position and speed of the particle are adjusted according to the update rule of the particle swarm optimization algorithm.

[0073] Perform genetic algorithm selection, crossover, and mutation operations to increase the diversity of the population and explore the solution space.

[0074] S4224. Combine the population processed by the genetic algorithm with the position update rule of particles in the particle swarm optimization algorithm, use the global search capability of the genetic algorithm to avoid the particle swarm optimization algorithm from falling into the local optimum, and utilize the fast convergence characteristics of the particle swarm optimization algorithm.

[0075] S4225. Repeat steps S4223 to S4224 until the stop condition is met.

[0076] In a further embodiment, the multi-dimensional correlation analysis of the integrated data set to obtain potential emotional threat factors in the maintenance process includes the following steps: S451. Calculate the correlation coefficient between the predicted health status of the baler and the emotional state of the maintenance personnel using the Pearson correlation coefficient; for example, calculate the Pearson correlation coefficient between the predicted health status of the baler (such as healthy, sub-healthy, faulty) and the emotional state of the maintenance personnel (such as stress, fatigue, satisfaction, etc.).

[0077] S452. Determine the emotional features that are significantly associated with the health status of the baler through a significance test; wherein the significance test can screen out the emotional features that have a significant impact on the health status of the baler.

[0078] S453. Use K-means and other clustering algorithms to complete clustering of the comprehensive data set and identify the health status patterns of the packer under different emotional states.

[0079] S454. Through the clustering results, the association pattern between high-risk emotional states and baler failures is identified. For example, the equipment failure rate is higher under certain specific emotional combinations (such as high stress and high fatigue).

[0080] In a further embodiment, formulating a maintenance strategy for a baler according to the output of the predictive maintenance model and the multi-dimensional correlation analysis results includes the following steps: S461. Monitor the emotional characteristics that are significantly associated with the health status of the baler equipment. If the significant emotional characteristic values ​​of the maintenance personnel are beyond the normal range, take preventive intervention measures, such as temporary rest, psychological counseling, etc.

[0081] S462. Dynamically adjust the distribution of maintenance tasks according to the emotional state of maintenance personnel to ensure that key tasks are carried out in the best emotional state, establish an emotional state and task matching model, and automatically adjust task distribution according to the real-time emotional state.

[0082] For example, first determine the dataset: Machine characteristics, temperature, vibration frequency, pressure.

[0083] Emotional characteristics, stress level, fatigue index, and satisfaction.

[0084] Target variable, remaining useful life (RUL) of the baler.

[0085] The prediction results and the comprehensive data set of the emotional state of the maintenance personnel are obtained, as shown in Table 1: Table 1 Prediction results and time alignment of maintenance personnel’s emotional states The Pearson correlation coefficient between the health status prediction and each emotional feature (stress level, fatigue index, satisfaction) was calculated, and clustering algorithms such as K-means were used to complete the clustering of the comprehensive data set to identify the health status patterns of the packer under different emotional states, and the following Table 2 was obtained: Table 2 Cluster analysis results based on emotional characteristics and health status prediction From the clustering results, it was found that the group with clustering result 3 (T4, T6, T8) had a low predicted health status, and the emotional characteristics were high stress and high fatigue, and low satisfaction.

[0086] Maintenance strategy of baler: Monitor the emotional characteristics of maintenance personnel in real time. If the stress level exceeds 4, the fatigue index exceeds 6, or the satisfaction level is lower than 2, take preventive intervention measures such as temporary rest and psychological counseling.

[0087] Adjust task assignments based on real-time emotional status to ensure that key tasks are performed in a good emotional state (stress level ≤ 3, fatigue index ≤ 5, satisfaction ≥ 4).

[0088] like Figure 2 As shown, according to another embodiment of the present invention, a baler predictive maintenance system based on the Internet of Things is also provided, which is used to implement a baler predictive maintenance method based on the Internet of Things as described in any one of claims 1-9, and is characterized in that the baler predictive maintenance system based on the Internet of Things includes an original data acquisition module 1, a key feature extraction module 2, a comprehensive feature construction module 3 and a maintenance strategy formulation module 4; wherein the original data acquisition module 1, the key feature extraction module 2, the comprehensive feature construction module 3 and the maintenance strategy formulation module 4 are connected in sequence.

[0089] The original data acquisition module 1 is used to acquire the original machine status data of the baler and the original emotional data of the maintenance personnel, and pre-process them to obtain the pre-processed machine data and the pre-processed emotional data.

[0090] The key feature extraction module 2 is used to upload the preprocessed machine data and the preprocessed emotion data to the cloud using the Internet of Things technology, and at the same time extract the key machine features and key emotion features from the preprocessed machine data and the preprocessed emotion data.

[0091] The comprehensive feature construction module 3 is used to filter the effective information in the key machine features and key emotional features, and construct a comprehensive feature vector matrix.

[0092] The maintenance strategy formulation module 4 is used to train the predictive maintenance model based on the comprehensive feature vector matrix and formulate the maintenance strategy of the baler.

[0093] In summary, the present invention provides a predictive maintenance method and system for a baler based on the Internet of Things, which ensures the integrity of information by collecting machine status data and maintenance personnel's emotional data. Data is collected in real time through Internet of Things devices to ensure the timeliness of data, and cloud storage technology is used to provide an efficient and secure storage solution. The present invention provides a more comprehensive data perspective by combining machine status and maintenance personnel's emotional characteristics, which helps to capture more potential influencing factors. The random forest algorithm can provide a preliminary assessment of feature importance and provide guidance for feature selection. The true importance of features is evaluated by observing the impact of random permutation of features on model performance. By calculating the correlation between features and eliminating highly correlated features, the redundancy between features can be reduced, the generalization ability of the model can be improved, and the retained features are more representative and discriminative, which helps to explain the prediction results of the model, making the model decision more transparent and credible. The predictive maintenance model based on comprehensive feature vector matrix training can more accurately predict the health status of the baler. Through multi-dimensional correlation analysis, the relationship between equipment status and maintenance personnel's emotions can be deeply explored to identify potential threat factors. Combined with the output of the prediction model and the results of correlation analysis, a scientific maintenance strategy is formulated to improve the accuracy and rationality of decision-making.

[0094] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A predictive maintenance method for a baler based on the Internet of Things, characterized in that: The IoT-based predictive maintenance method for balers includes the following steps: S1. Obtaining the original machine status data of the baler and the original emotional data of the maintenance personnel, and preprocessing to obtain the preprocessed machine data and preprocessed emotional data; S2. Upload the preprocessed machine data and the preprocessed emotion data to the cloud using the Internet of Things technology, and extract key machine features and key emotion features from the preprocessed machine data and the preprocessed emotion data; S3, filter out effective information from key machine features and key emotional features, and construct a comprehensive feature vector matrix; S4, train the predictive maintenance model based on the comprehensive feature vector matrix and formulate the maintenance strategy of the baler; The filtering of effective information in key machine features and key emotional features and constructing a comprehensive feature vector matrix includes the following steps: S31. Combine key machine features and key emotional features to construct an initial feature matrix and determine the target variable; S32, using the initial feature matrix and target variables to train a random forest model as a baseline model; S33. Use cross-validation to evaluate the performance of the random forest model; S34, permuting the features in each initial feature matrix and selecting features according to the permutation test results; S35. Calculate the Pearson correlation coefficient between the selected features, and eliminate redundant features based on the correlation matrix to obtain a comprehensive feature vector matrix.

2. The method for predictive maintenance of a baler based on the Internet of Things according to claim 1, characterized in that: The method of uploading the preprocessed machine data and the preprocessed emotion data to the cloud by using the Internet of Things technology and extracting the key machine features and the key emotion features from the preprocessed machine data and the preprocessed emotion data comprises the following steps: S21. Verify the quality of the pre-processed machine data and sentiment data, and convert them into a format for IoT transmission; S22, uploading the formatted data to the cloud server through the IoT device; S23. Extract key machine features and key emotional features.

3. The predictive maintenance method for a baler based on the Internet of Things according to claim 2 is characterized in that: The permutation test of the features in each initial feature matrix and the selection of features according to the permutation test results include the following steps: S341. Use each decision tree in the random forest model to evaluate the features in the initial feature matrix and use the average reduction impurity to measure the importance of the features. S342. For each feature, aggregate the average importance score of the feature in all decision trees; S343. Select a feature and randomly shuffle the value of the feature on the validation set, while keeping other features unchanged; S344. Retrain the random forest model using the randomly shuffled data set; S345, obtaining the change in performance of the retrained random forest model compared to the baseline model; S346, repeat the permutation test for all features to evaluate the impact of each feature on the performance of the random forest model; S347. Analyze the degree of degradation of the performance of the random forest model after each feature replacement. According to the degree of performance degradation, rank the features by importance and select several features with the highest ranking.

4. The predictive maintenance method for a baler based on the Internet of Things according to claim 3 is characterized in that: The step of calculating the Pearson correlation coefficient between the selected features and eliminating redundant features based on the correlation matrix to obtain a comprehensive feature vector matrix includes the following steps: S351, calculating the Pearson correlation coefficient between the selected features and constructing a correlation matrix; S352, eliminating redundant features according to the correlation matrix; S353. Construct a comprehensive feature vector matrix based on the results after redundant features are eliminated.

5. The method for predictive maintenance of a baler based on the Internet of Things according to claim 4, characterized in that: The method of training a predictive maintenance model based on a comprehensive feature vector matrix and formulating a maintenance strategy for a baler includes the following steps: S41, obtaining labels corresponding to each feature in the comprehensive feature vector matrix, including machine health status prediction labels and emotion impact assessment labels, and combining the comprehensive feature vector matrix, machine health status prediction labels and emotion impact assessment labels to obtain input variables and target variables; S42, based on the machine learning algorithm, and using the input variables and the target variables to train the corresponding machine learning model to obtain a predictive maintenance model; S43, obtaining new input variables and inputting them into the predictive maintenance model to obtain the health status prediction result of the baler at each time point; S44, time-aligning the prediction results with the emotional data of the maintenance personnel to form a comprehensive data set including the health status prediction of the baler and the emotional status of the maintenance personnel; S45, multi-dimensional correlation analysis integrated the data set to obtain potential emotional threat factors in the maintenance process; S46. Formulate a maintenance strategy for the baler based on the output of the predictive maintenance model and the results of the multi-dimensional correlation analysis.

6. The predictive maintenance method for a baler based on the Internet of Things according to claim 5 is characterized in that: The method of obtaining a predictive maintenance model based on a machine learning algorithm and using input variables and target variables to train a corresponding machine learning model includes the following steps: S421, divide the input variables and the corresponding target variables to obtain a training set and a test set; S422, selecting a support vector machine algorithm, and using the training set data to train the support vector machine algorithm model, and using a hybrid optimization algorithm to optimize the support vector machine algorithm model parameters to obtain a predictive maintenance model; S423, verifying the predictive maintenance model using the test set, and optimizing the parameters of the predictive maintenance model according to the evaluation results; Among them, the decision function of the predictive maintenance model is: ； In the formula, f ( x ) represents the prediction of new input variables; α i Indicates i The Lagrange multiplier of training samples, y i Indicates i The labels of the support vectors, K represents the kernel function; b represents the bias term; N Represents the number of support vectors.

7. The predictive maintenance method for a baler based on the Internet of Things according to claim 6 is characterized in that: Optimizing the support vector machine algorithm model parameters using a hybrid optimization algorithm includes the following steps: S4221. Randomly generate a particle swarm of a particle swarm optimization algorithm and a chromosome population of a genetic algorithm, and each individual represents a parameter combination of a support vector machine algorithm model; S4222, determining a fitness function, and encoding and decoding chromosomes; S4223, running particle swarm optimization algorithm and genetic algorithm; S4224, combining the population processed by the genetic algorithm with the position update rule of particles in the particle swarm optimization algorithm; S4225. Repeat steps S4223 to S4224 until the stop condition is met.

8. The method for predictive maintenance of a baler based on the Internet of Things according to claim 7, characterized in that: The multi-dimensional correlation analysis integrates the data set to obtain potential emotional threat factors in the maintenance process, including the following steps: S451, calculating the correlation coefficient between the predicted health state of the baler and the emotional state of the maintenance personnel using the Pearson correlation coefficient; S452. Determine the emotional features that are significantly associated with the health status of the baler equipment through a significance test; S453, clustering the comprehensive data set using a clustering algorithm; S454. Through clustering results, the association pattern between high-risk emotional state and baler failure is identified.

9. The method for predictive maintenance of a baler based on the Internet of Things according to claim 8, characterized in that: The method of formulating a maintenance strategy for a baler according to the output of the predictive maintenance model and the multi-dimensional correlation analysis results includes the following steps: S461. Monitor the emotional characteristics that are significantly associated with the health status of the baler equipment. If the significant emotional characteristic values ​​of the maintenance personnel are beyond the normal range, take preventive intervention measures; S462. Dynamically adjust the allocation of maintenance tasks according to the emotional state of the maintenance personnel.

10. A predictive maintenance system for a baler based on the Internet of Things, used to implement a predictive maintenance method for a baler based on the Internet of Things as described in any one of claims 1 to 9, characterized in that: The predictive maintenance system for balers based on the Internet of Things includes a raw data acquisition module, a key feature extraction module, a comprehensive feature construction module and a maintenance strategy formulation module; Wherein, the original data acquisition module, the key feature extraction module, the comprehensive feature construction module and the maintenance strategy formulation module are connected in sequence; The original data acquisition module is used to acquire the original machine status data of the baler and the original emotional data of the maintenance personnel, and preprocess to obtain the preprocessed machine data and preprocessed emotional data; The key feature extraction module is used to upload the preprocessed machine data and the preprocessed emotion data to the cloud using the Internet of Things technology, and extract key machine features and key emotion features from the preprocessed machine data and the preprocessed emotion data; The comprehensive feature construction module is used to filter effective information in key machine features and key emotional features, and construct a comprehensive feature vector matrix; The maintenance strategy formulation module is used to train a predictive maintenance model based on a comprehensive feature vector matrix and formulate a maintenance strategy for the baler.

Images