Noise reduction optimization management method and system for compressed air system

Abstract

The invention discloses a noise reduction optimization management method and system for a compressed air system, and belongs to the technical field of noise reduction optimizing.The method comprises the steps of integration of noise reduction related data of the compressed air system, optimization of the noise reduction related data of the compressed air system, construction of a noise reduction optimization management model and noise reduction optimization management. According to the scheme, the dynamic credible coefficient of the data in the aspect of features is obtained according to the dissimilatory index of the data relative to the features, the weighted distance is calculated, the nearest neighbor set is constructed, the cluster deviation coefficient of the data is obtained, and abnormal data is identified and deleted; introducing a modified harmonic series function to obtain an Euler-harmonic information entropy of the set, selecting an optimal split feature and an optimal split feature value according to an Euler gain and a control factor to obtain a decision tree constructed based on a structure construction data subset, and redefining labels of leaf nodes in the decision tree based on a label definition data subset to obtain a decision tree set; noise reduction optimization management can be dynamically carried out on the compressed air system in real time, and the management efficiency and the noise reduction effect are improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of noise reduction optimization, and specifically refers to a noise reduction optimization management method and system for a compressed air system. Background Art

[0002] The noise reduction optimization management method is a method that uses artificial intelligence technology to collect and process multi-dimensional data of compressed air system noise reduction, mine potential laws in historical data, optimize the noise reduction control scheme of the compressed air system, effectively reduce the noise level during the operation of the compressed air system, improve the operation stability and working efficiency of the compressed air system, create a quiet and comfortable atmosphere for the working environment, reduce the impact of noise on the health and working status of surrounding personnel, and also help reduce the risk of equipment failure caused by noise and extend the service life of the equipment.

[0003] However, in the existing noise reduction optimization management methods, during the operation of the compressed air system, due to sensor errors, environmental interference, and working condition fluctuations, some of the data collected related to the noise reduction of the compressed air system will deviate from the normal range, resulting in low data quality, reducing the accuracy of noise reduction optimization management, and poor noise reduction optimization effect of the compressed air system; in the existing noise reduction optimization management methods, the data collected related to the noise reduction of the compressed air system has the characteristics of high complexity and diversity, and traditional methods are difficult to effectively adapt to the dynamic change characteristics of the data, resulting in poor noise reduction strategy effects and unable to effectively carry out noise reduction optimization management of the compressed air system. Summary of the Invention

[0004] In view of the above situation, to overcome the defects of the prior art, the present invention provides a noise reduction optimization management method and system for a compressed air system. In the existing noise reduction optimization management methods, during the operation of the compressed air system, due to sensor errors, environmental interference, and working condition fluctuations, some of the data collected related to the noise reduction of the compressed air system will deviate from the normal range, resulting in low data quality, reducing the accuracy of noise reduction optimization management, and poor noise reduction optimization effect of the compressed air system. According to the alienation index of the data relative to the feature, this solution obtains the dynamic credibility coefficient of the data on the feature, and can identify the abnormality of a single data point on a specific feature; calculates the weighted distance based on the dynamic credibility coefficient to make the distance calculation more accurate and reduce the false alarm rate; constructs the nearest neighbor set to obtain the cluster deviation coefficient of the data, identifies and deletes abnormal data, more accurately identifies the reliability of the data, effectively eliminates outliers, improves data quality, and makes the subsequent noise reduction optimization management analysis based on the data more accurate; in view of the problem that the data collected related to the noise reduction of the compressed air system in the existing noise reduction optimization management methods has the characteristics of high complexity and diversity, and traditional methods are difficult to effectively adapt to the dynamic change characteristics of the data, resulting in poor noise reduction strategy effect and inability to effectively perform noise reduction optimization management on the compressed air system, this solution introduces a modified harmonic series function to obtain the Euler-harmonic information entropy of the set, which can more comprehensively and accurately measure the information content and uncertainty of the data subset; selects the best splitting feature and the best splitting feature value according to the Euler gain and the control factor to obtain a decision tree constructed based on the structure to build data subsets, which helps to build a more efficient and accurate decision tree; then redefines the labels of the leaf nodes in the decision tree based on the label definition of the data subset to make the output of the decision tree more in line with the requirements of noise reduction optimization management; based on the ensemble learning technology, combines the decision trees to obtain a noise reduction optimization management model, better cope with the high complexity and dynamic change characteristics of the compressed air system data, formulate more targeted noise reduction strategies for different working conditions, and help to perform noise reduction optimization management on the compressed air system in real time and dynamically, improving the management efficiency and noise reduction effect.

[0005] The technical solution adopted by the present invention is as follows: A noise reduction optimization management method for a compressed air system provided by the present invention includes the following steps:

[0006] Step S1: Integration of data related to noise reduction of the compressed air system;

[0007] Step S2: Optimization of data related to noise reduction of the compressed air system;

[0008] Step S3: Construction of a noise reduction optimization management model;

[0009] Step S4: Noise reduction optimization management.

[0010] Further, in step S1, the integration of data related to noise reduction of the compressed air system is to collect and process historical data related to noise reduction of the compressed air system;

[0011] The collection of historical air compression system noise reduction related data is to collect air compression system operation data, air compression system equipment data, noise data, environmental data, and noise reduction adjustment data;

[0012] The processing of historical air compression system noise reduction related data is to perform data cleaning, data normalization, data encoding, and dataset construction on the historical air compression system noise reduction related data to obtain an air compression system noise reduction dataset.

[0013] Further, in step S2, the optimization of the air compression system noise reduction related data specifically includes the following steps:

[0014] Step S21: Calculate the dissimilarity index; calculate the Euclidean distance between every two data in the air compression system noise reduction dataset A on the same feature, and then perform normalization processing according to the maximum Euclidean distance and the minimum Euclidean distance to obtain the dissimilarity index of the data relative to the feature; the formula used is as follows:

[0015] ;

[0016] In the formula, and are respectively the maximum Euclidean distance and the minimum Euclidean distance between the data v in A and the remaining data on the feature m, is the dissimilarity index of the data v relative to the feature m, v and w are data indices in A, m is the feature index in A, and are respectively the projections of the data v and the data w on the feature m, is and the Euclidean distance between them, and H is the number of data in A;

[0017] Step S22: Calculate the dynamic credibility coefficient; calculate the average dissimilarity index of all data in A on the same feature, perform standardization processing on according to the average dissimilarity index to obtain the standardized dissimilarity index, and then use the inverse Sigmoid function to calculate the dynamic credibility coefficient of the feature;

[0018] Step S23: Calculate the nearest neighbor data; based on the dynamic credibility coefficient, calculate the weighted distance between the data v in A and each of the remaining data, sort the weighted distances in ascending order, and select the data corresponding to the first smaller weighted distances as the nearest neighbor data of the data v, and then obtain the nearest neighbor set of each data in A; where, is rounding down

[0019] Step S24: Calculate the cluster deviation coefficient; calculate the cluster deviation coefficient of the data according to the weighted distance and the nearest neighbor set;

[0020] Step S25: Optimization; preset a deviation threshold, delete the data with a cluster deviation coefficient greater than the deviation threshold from A as abnormal data to obtain an optimized management data set, select 70% of the data from the optimized management data set to construct a training data set, and the remaining 30% of the data as a test data set.

[0021] Further, in step S3, the construction of the noise reduction and optimization management model specifically includes the following steps:

[0022] Step S31: Training data set division; randomly divide the training data set into two non-overlapping subsets Z1 and Z2. Z1 and Z2 are the structure construction data subset and the label definition data subset respectively, record the ratio of the data quantities in Z1 and Z2, and set the values of the parameter group [β1, β2, p]; where, β1 and β2 are the first control factor and the second control factor respectively, and p is the threshold;

[0023] Step S32: Calculate the Euler-harmonic information entropy; design a modified harmonic series function based on the Euler constant and the harmonic series, and introduce the modified harmonic series function to obtain the Euler-harmonic information entropy of the set; the formula used is as follows:

[0024] ;

[0025] Where, is the Euler-harmonic information entropy of the structure construction data subset Z1, |Z1| is the number of data in Z1, Q is the number of data labels, q is the data label index, h q is the number of data in Z1 that belong to the data label q, is the modified harmonic series function, is constructed based on the harmonic series and corrected by adding the Euler constant;

[0026] Step S33: Construct a decision tree; construct a binary structure decision tree based on the structure construction data subset Z1, split each node of the decision tree based on the node splitting rule until the number of leaf nodes in the decision tree is equal to the threshold p to stop node splitting, and complete the construction of the decision tree; the node splitting rule includes the following steps:

[0027] Step S331: Calculate the Euler gain; subtract the weighted Euler-harmonic information entropy of the left and right child node data sets from the Euler-harmonic information entropy of the data set at the node respectively to obtain the Euler gain;

[0028] Step S332: Select the best splitting feature; construct a set of maximum Euler gains when splitting at the node using each feature, and normalize the set of maximum Euler gains to obtain a normalized set of maximum Euler gains. Calculate the selection probability of each feature in the normalized set of maximum Euler gains based on the first control factor β1 to obtain a set of feature selection probabilities. According to the set of feature selection probabilities, randomly select a feature as the best splitting feature;

[0029] Step S333: Select the best splitting feature value; construct a set of Euler gains when splitting at the node using all possible feature values of the best splitting feature, and normalize the set of Euler gains to obtain a normalized set of Euler gains. Calculate the selection probability of each feature value in the normalized set of Euler gains based on the second control factor β2 to obtain a set of feature value selection probabilities. Then, according to the set of feature value selection probabilities, randomly select a feature value from all possible feature values of the best splitting feature as the best splitting feature value;

[0030] Step S334: Split; use the selected best splitting feature value to split the node to obtain its left and right child nodes;

[0031] Step S34: Redefine the label; input the data in the label definition data subset Z2 into the constructed decision tree, and allocate the data to the corresponding leaf nodes according to the node splitting rule to obtain the data set corresponding to each leaf node on Z2. Calculate the prediction probability of each data label in the data set to obtain a set of prediction probabilities. Then, according to the set of prediction probabilities, select the data label with the maximum prediction probability from all data labels as the new data label of the leaf node;

[0032] Step S35: Decision tree combination; repeat steps S31 - S34 for N times to construct N decision trees in total. Based on the ensemble learning technique, combine the N decision trees to obtain a noise reduction and optimization management model, and output the predicted data label.

[0033] Furthermore, in step S4, the noise reduction and optimization management is to collect and process the data related to the noise reduction of the real-time air compressor system, input the processed data related to the noise reduction of the real-time air compressor system into the noise reduction and optimization management model for analysis, obtain the corresponding noise reduction adjustment data according to the output data label, generate a noise reduction adjustment strategy, and perform noise reduction and optimization management on the air compressor system according to the noise reduction adjustment strategy;

[0034] The collection of the data related to the noise reduction of the real-time air compressor system is to collect the operation data of the air compressor system, the equipment data of the air compressor system, the noise data, and the environmental data;

[0035] The processing of the noise reduction related data of the real-time compressed air system is to perform data cleaning, data normalization, and data encoding on the noise reduction related data of the real-time compressed air system.

[0036] A noise reduction optimization management system for a compressed air system provided by the present invention includes a module for integrating noise reduction related data of the compressed air system, a module for optimizing noise reduction related data of the compressed air system, a module for constructing a noise reduction optimization management model, and a noise reduction optimization management module;

[0037] The module for integrating noise reduction related data of the compressed air system collects and processes the historical noise reduction related data of the compressed air system, and sends the data to the module for optimizing noise reduction related data of the compressed air system;

[0038] The module for optimizing noise reduction related data of the compressed air system obtains the dynamic credibility coefficient of the data on the feature according to the dissimilation index of the data relative to the feature, calculates the weighted distance and constructs the nearest neighbor set, obtains the cluster deviation coefficient of the data, identifies and deletes the abnormal data, and sends the data to the module for constructing a noise reduction optimization management model;

[0039] The module for constructing a noise reduction optimization management model introduces a modified harmonic series function to obtain the Euler-harmonic information entropy of the set, selects the best splitting feature and the best splitting feature value according to the Euler gain and the control factor, obtains a decision tree constructed based on the data subset constructed by the structure, redefines the label of the leaf node in the decision tree based on the label definition data subset, and combines the decision trees based on the ensemble learning technology to obtain a noise reduction optimization management model, and sends the data to the noise reduction optimization management module;

[0040] The noise reduction optimization management module obtains the corresponding noise reduction adjustment data according to the data label output by the noise reduction optimization management model, generates a noise reduction adjustment strategy, and performs noise reduction optimization management on the compressed air system according to the noise reduction adjustment strategy.

[0041] The beneficial effects achieved by the present invention by adopting the above scheme are as follows:

[0042] (1)In view of the problem that in the existing noise reduction optimization management method, during the operation of the air compression system, due to sensor errors, environmental interference, and working condition fluctuations, some of the data collected related to the noise reduction of the air compression system will deviate from the normal range, resulting in low data quality, reducing the accuracy of noise reduction optimization management, and making the noise reduction optimization effect of the air compression system poor. This solution obtains the dynamic credibility coefficient of the data on the feature according to the dissimilation index of the data relative to the feature, and can identify the abnormality of a single data point on a specific feature; calculates the weighted distance based on the dynamic credibility coefficient to make the distance calculation more accurate and reduce the false alarm rate; constructs the nearest neighbor set to obtain the cluster deviation coefficient of the data, identifies and deletes abnormal data, more accurately identifies the reliability of the data, effectively eliminates outliers, improves data quality, makes the subsequent noise reduction optimization management analysis based on the data more accurate, and finally realizes more effective noise control.

[0043] (2)In view of the problem that in the existing noise reduction optimization management method, the data collected related to the noise reduction of the air compression system has the characteristics of high complexity and diversity, and traditional methods are difficult to effectively adapt to the dynamic change characteristics of the data, resulting in poor noise reduction strategy effects and being unable to effectively carry out noise reduction optimization management of the air compression system. This solution randomly divides the training data set into a structure construction data subset and a label definition data subset. Through this division, the information in the data can be mined more systematically; introduces the modified harmonic series function to obtain the Euler-harmonic information entropy of the set, which can more comprehensively and accurately measure the information content and uncertainty of the data subset, takes into account the dynamic changes of the data, and overcomes the problem of inaccurate information measurement of traditional methods when dealing with complex dynamic data; selects the best splitting feature and the best splitting feature value according to the Euler gain and control factor to obtain the decision tree constructed based on the structure construction data subset, which helps to construct a more efficient and accurate decision tree; then redefines the labels of the leaf nodes in the decision tree based on the label definition data subset to make the output of the decision tree more in line with the requirements of noise reduction optimization management, ensuring that the decision tree can accurately predict the appropriate noise reduction strategy according to the input data; based on the ensemble learning technology, combines the decision trees to obtain the noise reduction optimization management model, improves the generalization ability and stability, better responds to the high complexity and dynamic change characteristics of the air compression system data, formulates more targeted noise reduction strategies for different working conditions, and helps to carry out noise reduction optimization management of the air compression system in real time and dynamically, improving the management efficiency and noise reduction effect. Description of the Drawings

[0044] Figure 1 It is a schematic flow chart of a noise reduction optimization management method for an air compression system provided by the present invention;

[0045] Figure 2 It is a schematic diagram of a noise reduction optimization management system for an air compression system provided by the present invention;

[0046] Figure 3 It is a schematic flow diagram of step S2;

[0047] Figure 4 It is a schematic flow diagram of step S3.

[0048] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. They are used together with the embodiments of the present invention to explain the present invention and do not constitute a limitation to the present invention. Detailed Embodiments

[0049] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts belong to the scope of protection of the present invention.

[0050] In the description of the present invention, it should be understood that the terms "upper", "lower", "front", "rear", "left", "right", "top", "bottom", "inner", "outer", etc. indicating the orientation or positional relationship are based on the orientation or positional relationship shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and thus should not be construed as a limitation to the present invention.

[0051] Embodiment 1, refer to Figure 1 , a noise reduction optimization management method for a compressed air system provided by the present invention, the method includes the following steps:

[0052] Step S1: Integrate the data related to the noise reduction of the compressed air system; collect and process the historical data related to the noise reduction of the compressed air system;

[0053] Step S2: Optimize the data related to the noise reduction of the compressed air system; obtain the dynamic credibility coefficient of the data in terms of features according to the dissimilation index of the data relative to the features, calculate the weighted distance and construct the nearest neighbor set, obtain the cluster deviation coefficient of the data, and identify and delete the abnormal data;

[0054] Step S3: Construct a noise reduction optimization management model; introduce a modified harmonic series function to obtain the Euler-harmonic information entropy of the set, select the best splitting feature and the best splitting feature value according to the Euler gain and the control factor, obtain a decision tree constructed based on the structure to build a data subset, and then re-define the label of the leaf node in the decision tree based on the label definition of the data subset, and combine the decision trees to obtain a noise reduction optimization management model;

[0055] Step S4: Noise reduction optimization management; according to the data labels output by the noise reduction optimization management model, obtain the corresponding noise reduction adjustment data, generate a noise reduction adjustment strategy, and perform noise reduction optimization management on the compressed air system according to the noise reduction adjustment strategy.

[0056] Example 2, refer to Figure 1 , based on the above example, in step S1, the integration of compressed air system noise reduction related data is to collect and process historical compressed air system noise reduction related data;

[0057] The collection of historical compressed air system noise reduction related data is to collect compressed air system operation data, compressed air system equipment data, noise data, environmental data, and noise reduction adjustment data;

[0058] The compressed air system operation data includes fan speed, air pressure, air volume, fan vibration intensity, and operation time;

[0059] The compressed air system equipment data includes fan model, number of blades, maintenance record data, and failure rate;

[0060] The noise data includes noise intensity and noise frequency;

[0061] The environmental data includes temperature, humidity, atmospheric pressure, wind speed, and wind direction;

[0062] The processing of historical compressed air system noise reduction related data is to perform data cleaning, data normalization, data encoding, and dataset construction on historical compressed air system noise reduction related data to obtain a compressed air system noise reduction dataset;

[0063] Data cleaning is to remove error values, missing values, and outliers from the data;

[0064] Data normalization is to use the maximum-minimum scaling method to unify numerical data into the same range;

[0065] Data encoding is to use One-Hot encoding to convert categorical data into numerical data;

[0066] The construction of the dataset is to use the noise reduction adjustment data as data labels, and then construct a compressed air system noise reduction dataset based on the historical compressed air system noise reduction related data processed by data cleaning, data normalization, and data encoding.

[0067] Example 3, refer to Figure 1 and Figure 3 , based on the above example, in step S2, the optimization of compressed air system noise reduction related data specifically includes the following:

[0068] Step S21: Calculate the alienation index; calculate the Euclidean distance between every two data in the same feature of the noise reduction data set A of the compressed air system, and then perform normalization processing according to the maximum Euclidean distance and the minimum Euclidean distance to obtain the alienation index of the data relative to the feature; by calculating the Euclidean distance between data on the same feature, identify outliers that deviate significantly from other data on a specific feature, and eliminate the dimensional differences of different features through normalization processing to make subsequent analysis more balanced; the formula used is as follows:

[0069] ;

[0070] ;

[0071] ;

[0072] In the formula, and are respectively the maximum Euclidean distance and the minimum Euclidean distance between the data v in A and the remaining data on the feature m, is the alienation index of the data v relative to the feature m, v and w are data indexes in A, m is the feature index in A, and are respectively the projections of the data v and the data w on the feature m, is and the Euclidean distance between them, and H is the number of data in A;

[0073] Step S22: Calculate the dynamic credibility coefficient; calculate the average alienation index of all data in A on the same feature, perform standardization processing on according to the average alienation index to obtain the standardized alienation index, and then use the inverse Sigmoid function to calculate the dynamic credibility coefficient of the feature; through the inverse Sigmoid function, assign a low credibility coefficient to the feature with a high alienation index, and assign a high credibility coefficient to the feature with a low alienation index; the formula used is as follows:

[0074] ;

[0075] ;

[0076] ;

[0077] Among them, is the alienation index of the data v relative to the feature m, is the average alienation index of all data in A on the feature m, is the standardized alienation index of the data v relative to the feature m, is the standard deviation of the dissimilation index of all data in A on feature m, v is the data index in A, and m is the feature index in A. is the dynamic credibility coefficient of data i on feature m;

[0078] Step S23: Calculate the nearest neighbor data; Based on the dynamic credibility coefficient, calculate the weighted distance between data v in A and each of the remaining data, and sort the weighted distances in ascending order. Select the data corresponding to the smaller weighted distances as the nearest neighbor data of data v, and then obtain the nearest neighbor set of each data in A; Find the neighbor data most similar to the target data through the weighted distance, providing a local reference benchmark for subsequent anomaly detection; The formula used is as follows:

[0079] ;

[0080] In the formula, is the weighted distance between data v and data w in A, M is the number of features in A, is rounding down;

[0081] Step S24: Calculate the cluster deviation coefficient; Calculate the cluster deviation coefficient of the data according to the weighted distance and the nearest neighbor set; Identify isolated anomalies or group anomalies by comparing the sum of the weighted distances between the data and its neighbors. Introduce μ o to eliminate the scale effect and make the cluster deviation coefficients comparable under different conditions; The formula used is as follows:

[0082] ;

[0083] In the formula, D v is the cluster deviation coefficient of data v, C v is the sum of the weighted distances between data i and all data in its own nearest neighbor set, C o is the sum of the weighted distances between the nearest neighbor data o of data i and all data in its own nearest neighbor set, is the weighted distance between data v and the nearest neighbor data o in A, μ o is the mean of the weighted distances between the nearest neighbor data o and all data in its own nearest neighbor set, and o is the nearest neighbor data index;

[0084] Step S25: Optimization; Preset a deviation threshold, and delete the data with a cluster deviation coefficient greater than the deviation threshold from A as abnormal data to obtain an optimized management data set. Select 70% of the data from the optimized management data set to construct a training data set, and the remaining 30% of the data as a test data set; Eliminate abnormal data through the deviation threshold to improve the quality of the data set.

[0085] By performing the above operations, in view of the problem that in the existing noise reduction optimization management method, during the operation of the air compression system, due to sensor errors, environmental interference, and working condition fluctuations, some of the data collected related to the noise reduction of the air compression system will deviate from the normal range, resulting in low data quality, reducing the accuracy of noise reduction optimization management, and poor noise reduction optimization effect of the air compression system. According to the dissimilation index of the data relative to the feature, this solution obtains the dynamic credibility coefficient of the data on the feature, and can identify the abnormality of a single data point on a specific feature; calculates the weighted distance based on the dynamic credibility coefficient to make the distance calculation more accurate and reduce the false alarm rate; constructs the nearest neighbor set, obtains the cluster deviation coefficient of the data, identifies and deletes abnormal data, more accurately identifies the reliability of the data, effectively eliminates outliers, improves data quality, makes the subsequent noise reduction optimization management analysis based on the data more accurate, and finally realizes more effective noise control.

[0086] Example 4, refer to Figure 1 and Figure 4 , based on the above example, in step S3, constructing the noise reduction optimization management model specifically includes the following steps:

[0087] Step S31: Training dataset division; randomly divide the training dataset into two non-overlapping subsets Z1 and Z2. Z1 and Z2 are the structure construction data subset and the label definition data subset respectively, and record the ratio of the number of data in Z1 to Z2 , and set the values of the parameter group [β1, β2, p]; where β1 and β2 are the first control factor and the second control factor respectively, p is the threshold, |Z1| and |Z2| are the number of data in Z1 and Z2 respectively. Z1 is used to construct the structure of the decision tree, and Z2 is used to redefine the labels of the leaf nodes in the decision tree; Z2 is independent of Z1, can correct the label deviation of the decision tree, make the model more adaptable to the real situation, and control the model structure through the data volume ratio and the parameter group to avoid a single data distribution dominating the construction of the decision tree;

[0088] Step S32: Calculate the Euler-harmonic information entropy; design a modified harmonic series function based on the Euler constant and the harmonic series, and introduce the modified harmonic series function to obtain the Euler-harmonic information entropy of the set; make it more adaptable to the statistical characteristics of the data related to the noise reduction of the air compression system; the formula used is as follows:

[0089] ;

[0090] Among them, is the Euler-harmonic information entropy of the structure construction data subset Z1, |Z1| is the number of data in Z1, Q is the number of data labels, q is the data label index, h q is the number of data in Z1 that belong to the data label q, is the modified harmonic series function, Constructed based on the harmonic series and corrected by adding the Euler constant, , where γ is the Euler constant;

[0091] Step S33: Construct a decision tree; construct a binary decision tree based on the data subset Z1 of the structure, and split each node of the decision tree based on the node splitting rule until the number of leaf nodes in the decision tree is equal to the threshold p, then stop node splitting to complete the construction of the decision tree; the node splitting rule includes the following steps:

[0092] Step S331: Calculate the Euler gain; subtract the weighted Euler-harmonic information entropy of the left and right child node data sets from the Euler-harmonic information entropy of the data set at the node to obtain the Euler gain; the formula used is as follows:

[0093] ;

[0094] In the formula, is the Euler gain corresponding to splitting at node a using the feature value j of feature i, and Z1 a is the data set at node a of the decision tree constructed based on Z1, is the feature value j of feature i, and a, i, and j are the node index of the decision tree, the feature index in Z1, and the feature value index in Z1 respectively, and are the data sets of the left and right child nodes after splitting at node a using respectively, , and are the Euler-harmonic information entropies of the sets Z1 a , the set and the set respectively, , and are the data quantities in the sets Z1 a , the set and the set respectively, and are the weighted Euler-harmonic information entropies of the left and right child node data sets after splitting at node a using b ij respectively;

[0095] Step S332: Select the best splitting feature; construct a set of maximum Euler gains when splitting at the node using each feature, and normalize the set of maximum Euler gains to obtain a normalized set of maximum Euler gains. Calculate the selection probability of each feature in the normalized set of maximum Euler gains based on the first control factor β1 to obtain a set of feature selection probabilities. Randomly select a feature from the set of feature selection probabilities as the best splitting feature; improve the generalization ability so that the model can adapt to the data related to the noise reduction of the compressed air system under different working conditions; the formula used is as follows:

[0096] ;

[0097] ;

[0098] ;

[0099] ;

[0100] where L a and are the set of maximum Euler gains and the normalized set of maximum Euler gains at node a respectively, I is the total number of features, , and are the maximum values of the Euler gains generated when splitting using all possible feature values of the 1st, the ith, and the Ith features at node a respectively, G a is the set of feature selection probabilities at node a, is the softmax function; is to calculate the corresponding Euler gain for all possible feature values j of feature i at node a, and then take the maximum value; is to take the minimum value of the maximum Euler gains of all features i at node a; is to take the maximum value of the maximum Euler gains of all features i at node a;

[0101] Step S333: Select the best splitting feature value; construct a set of Euler gains when splitting using all possible feature values of the best splitting feature at the node, and normalize the set of Euler gains to obtain a normalized set of Euler gains. Calculate the selection probability of each feature value in the normalized set of Euler gains based on the second control factor β2 to obtain a set of feature value selection probabilities, and then randomly select a feature value from all possible feature values of the best splitting feature according to the set of feature value selection probabilities as the best splitting feature value; automatically adapt to the noise characteristics of different compressed air systems; the formula used is as follows:

[0102] ;

[0103] ;

[0104] ;

[0105] In the formula, and are the Euler gain set and the normalized Euler gain set respectively when splitting at node a using the best splitting feature ibest. ibest is the best splitting feature index, and J is the number of all feature values of the best splitting feature. , and are the 1st, j-th, and J-th feature values of the best splitting feature ibest respectively. , and are the corresponding Euler gains when splitting at node a using the 1st, j-th, and J-th feature values of the best splitting feature ibest respectively. is the feature value selection probability set when splitting at node a using the best splitting feature ibest. is, at node a, for all possible feature values j of the selected best splitting feature ibest, calculating their corresponding Euler gains and then taking the minimum value among them. is, at node a, for all possible feature values j of the selected best splitting feature ibest, calculating their corresponding Euler gains and then taking the maximum value among them.

[0106] Step S334: Split; Use the selected best splitting feature value to split the node to obtain its left and right child nodes.

[0107] Step S34: Redefine the label; Input the data in the label definition data subset Z2 into the constructed decision tree, allocate the data to the corresponding leaf nodes according to the node splitting rule, obtain the data set corresponding to each leaf node on Z2, calculate the prediction probability of each data label in the data set to obtain the prediction probability set, and then select the data label with the maximum prediction probability from all data labels as the new data label of the leaf node according to the prediction probability set; The labels of the leaf nodes of the decision tree may be incorrect due to the data deviation of Z1. Correct the label deviation through Z2 to ensure that the model can still make correct predictions in rare situations, avoid label oscillation caused by training data noise, and improve the model stability; The formula used is as follows:

[0108] ;

[0109] ;

[0110] Among them, G c is the set of predicted probabilities of leaf node c, and Z2 c is the data set corresponding to leaf node c on Z2, and |Z2 c | is the number of data in Z2 c , c is the index of the leaf node of the decision tree, , and are the predicted probabilities of the first, q-th, and Q-th data labels of leaf node c in Z2 c respectively. Q is the number of data labels, q is the data label index, and x d is the d-th data in Z2 c , and d is the data index in Z2 c . is the predicted data label of the decision tree for data x d ; is the indicator function; when , is 1, otherwise is 0;

[0111] Step S35: Decision tree combination; Repeat steps S31 - S34 N times to construct N decision trees in total. Based on the ensemble learning technique, combine the N decision trees to obtain a noise reduction optimization management model and output the predicted data label; A single decision tree may be affected by randomness, resulting in unstable noise reduction strategies. By jointly making decisions with multiple decision trees, it is applicable to different data related to the noise reduction of the air compressor system; The formula used is as follows:

[0112] ;

[0113] Among them, Y is the predicted data label output by the noise reduction optimization management model, is the predicted data label of the n-th decision tree for data x; is the indicator function; when , is 1, otherwise is 0.

[0114] By performing the above operations, in view of the problems in the existing noise reduction optimization management method that the noise reduction-related data collected from the air compressor system is highly complex and diverse, and the traditional method is difficult to effectively adapt to the dynamic change characteristics of the data, resulting in poor effects of the noise reduction strategy and inability to effectively perform noise reduction optimization management on the air compressor system, this solution randomly divides the training data set into a structure construction data subset and a label definition data subset. Through this division, the information in the data can be mined more systematically; the modified harmonic series function is introduced to obtain the Euler-harmonic information entropy of the set, which can more comprehensively and accurately measure the information content and uncertainty of the data subset, takes into account the dynamic changes of the data, and overcomes the problem of inaccurate information measurement of the traditional method when dealing with complex dynamic data; the best splitting feature and the best splitting feature value are selected according to the Euler gain and the control factor to obtain a decision tree constructed based on the structure construction data subset, which helps to construct a more efficient and accurate decision tree; then, based on the label definition data subset, the labels of the leaf nodes in the decision tree are redefined to make the output of the decision tree more in line with the requirements of noise reduction optimization management, ensuring that the decision tree can accurately predict the appropriate noise reduction strategy according to the input data; based on the ensemble learning technology, the decision trees are combined to obtain a noise reduction optimization management model, which improves the generalization ability and stability, better copes with the high complexity and dynamic change characteristics of the air compressor system data, formulates more targeted noise reduction strategies for different working conditions, helps to perform noise reduction optimization management on the air compressor system in real time and dynamically, and improves the management efficiency and noise reduction effect.

[0115] Embodiment 5, refer to Figure 1 , based on the above embodiment, in step S4, the noise reduction optimization management is to collect and process the real-time noise reduction-related data of the air compressor system, input the processed real-time noise reduction-related data of the air compressor system into the noise reduction optimization management model for analysis, obtain the corresponding noise reduction adjustment data according to the output data label, generate a noise reduction adjustment strategy, and perform noise reduction optimization management on the air compressor system according to the noise reduction adjustment strategy;

[0116] The collection of the real-time noise reduction-related data of the air compressor system is to collect the operation data of the air compressor system, the equipment data of the air compressor system, the noise data and the environmental data;

[0117] The processing of the real-time noise reduction-related data of the air compressor system is to perform data cleaning, data normalization and data encoding on the real-time noise reduction-related data of the air compressor system.

[0118] Embodiment 6, refer to Figure 2 , based on the above embodiment, a noise reduction optimization management system for an air compressor system provided by the present invention includes a noise reduction-related data integration module of the air compressor system, a noise reduction-related data optimization module of the air compressor system, a module for constructing a noise reduction optimization management model, and a noise reduction optimization management module;

[0119] The noise reduction related data integration module of the compressed air system collects and processes the historical noise reduction related data of the compressed air system, and sends the data to the noise reduction related data optimization module of the compressed air system;

[0120] The noise reduction related data optimization module of the compressed air system obtains the dynamic credibility coefficient of the data in terms of features according to the dissimilation index of the data relative to the features, calculates the weighted distance and constructs the nearest neighbor set, obtains the cluster deviation coefficient of the data, identifies and deletes the abnormal data, and sends the data to the module for constructing the noise reduction optimization management model;

[0121] The module for constructing the noise reduction optimization management model introduces the modified harmonic series function to obtain the Euler-harmonic information entropy of the set, selects the best splitting feature and the best splitting feature value according to the Euler gain and the control factor, obtains the decision tree constructed based on the data subset constructed by the structure, and then redefines the label of the leaf node in the decision tree based on the label definition of the data subset. Based on the ensemble learning technology, the decision trees are combined to obtain the noise reduction optimization management model, and the data is sent to the noise reduction optimization management module;

[0122] The noise reduction optimization management module obtains the corresponding noise reduction adjustment data according to the data label output by the noise reduction optimization management model, generates a noise reduction adjustment strategy, and performs noise reduction optimization management on the compressed air system according to the noise reduction adjustment strategy.

[0123] It should be noted that in this article, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the term "comprising", "including" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, article or device comprising a series of elements includes not only those elements, but also other elements not expressly listed, or elements inherent to such process, method, article or device.

[0124] Although the embodiments of the present invention have been shown and described, those of ordinary skill in the art can understand that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principle and spirit of the present invention.

[0125] The above describes the present invention and its implementation manners. This description is not restrictive. What is shown in the drawings is only one of the implementation manners of the present invention, and the actual structure is not limited thereto. Generally speaking, if those of ordinary skill in the art are inspired by it and design similar structural manners and embodiments without creative efforts without departing from the purpose of the present invention, they shall fall within the protection scope of the present invention.

Claims

1. A noise reduction and optimization management method for a compressed air system, characterized in that: The method includes the following steps: Step S1: Integration of data related to noise reduction of the compressed air system; collecting and processing historical data related to noise reduction of the compressed air system; Step S2: Optimization of data related to noise reduction of the compressed air system; obtaining the dynamic credibility coefficient of the data in terms of features according to the alienation index of the data with respect to the features, calculating the weighted distance and constructing the nearest neighbor set, obtaining the cluster deviation coefficient of the data, and identifying and deleting abnormal data; Step S3: Constructing a noise reduction optimization management model; introducing a modified harmonic series function to obtain the Euler-harmonic information entropy of the set, selecting the best splitting feature and the best splitting feature value according to the Euler gain and the control factor, obtaining a decision tree constructed based on the data subset constructed by the structure, and then redefining the labels of the leaf nodes in the decision tree based on the data subset defined by the labels, and combining the decision trees to obtain the noise reduction optimization management model; Step S4: Noise reduction optimization management; obtaining the corresponding noise reduction adjustment data according to the data labels output by the noise reduction optimization management model, generating a noise reduction adjustment strategy, and performing noise reduction optimization management on the compressed air system according to the noise reduction adjustment strategy.

2. The noise reduction and optimization management method for a compressed air system according to claim 1, characterized in that: In step S2, the optimization of the data related to noise reduction of the compressed air system specifically includes the following steps: Step S21: Calculating the alienation index; Step S22: Calculate the dynamic credibility coefficient; calculate the average dissimilation index of all data in A on the same feature, and perform normalization processing on to obtain the normalized dissimilation index, and then use the inverse Sigmoid function to calculate the dynamic credibility coefficient of the feature; where A is the noise reduction data set of the air compressor system, is the dissimilation index of data v relative to feature m, v is the data index in A, and m is the feature index in A; Step S23: Calculate the nearest neighbor data; Based on the dynamic trust coefficient, calculate the weighted distance between the data v in A and each of the remaining data, sort the weighted distances in ascending order, and select the data corresponding to the first smaller weighted distances as the nearest neighbor data of the data v, thereby obtaining the nearest neighbor set of each data in A; where H is the number of data in A, is rounding down; Step S24: Calculating the cluster deviation coefficient; calculating the cluster deviation coefficient of the data according to the weighted distance and the nearest neighbor set; Step S25: Optimization; presetting a deviation threshold, deleting the data with a cluster deviation coefficient greater than the deviation threshold as abnormal data from A to obtain an optimized management data set, selecting 70% of the data from the optimized management data set to construct a training data set, and using the remaining 30% of the data as a test data set.

3. The noise reduction and optimization management method for a compressed air system according to claim 2, characterized in that: In step S21, the calculation of the alienation index is to calculate the Euclidean distance between every two data in the noise reduction data set A of the compressed air system on the same feature, and then perform normalization processing according to the maximum Euclidean distance and the minimum Euclidean distance to obtain the alienation index of the data with respect to the feature; the formula used is as follows: ； Wherein, and are respectively the maximum Euclidean distance and the minimum Euclidean distance between the data v in A and the remaining data on the feature m, is the dissimilation index of the data v with respect to the feature m, v and w are data indices in A, m is a feature index in A, and are respectively the projections of the data v and the data w on the feature m, is and the Euclidean distance between them, and H is the number of data in A.

4. A noise reduction and optimization management method for a compressed air system according to claim 1, characterized in that: In step S3, the construction of the noise reduction optimization management model specifically includes the following steps: Step S31: Division of the training data set; randomly dividing the training data set into two non-overlapping subsets Z1 and Z2, where Z1 and Z2 are the data subsets for structure construction and label definition respectively, recording the ratio of the number of data in Z1 to Z2, and setting the values of the parameter group [β1, β2, p]; where β1 and β2 are the first control factor and the second control factor respectively, and p is the threshold; Step S32: Calculating the Euler-harmonic information entropy; designing a modified harmonic series function based on the Euler constant and the harmonic series, and introducing the modified harmonic series function to obtain the Euler-harmonic information entropy of the set; Step S33: Constructing a decision tree; constructing a binary-structured decision tree based on the data subset Z1 for structure construction, splitting each node of the decision tree based on the node splitting rule until the number of leaf nodes in the decision tree is equal to the threshold p, and then stopping the node splitting to complete the construction of the decision tree; Step S34: Redefine labels; input the data in the label definition data subset Z2 into the constructed decision tree, allocate the data to the corresponding leaf nodes according to the node splitting rules, obtain the data sets corresponding to each leaf node on Z2, calculate the prediction probabilities of each data label in the data sets to obtain a prediction probability set, and then select the data label with the maximum prediction probability from all data labels as the new data label of the leaf node according to the prediction probability set; Step S35: Decision tree combination; repeat steps S31 - S34 for N times to construct N decision trees in total. Based on the ensemble learning technique, combine the N decision trees to obtain a noise reduction and optimization management model, and output the predicted data labels.

5. The noise reduction and optimization management method for a compressed air system according to claim 4, characterized in that: In step S32, the calculation of the Euler - harmonic information entropy is to design a modified harmonic series function based on the Euler constant and the harmonic series, and introduce the modified harmonic series function to obtain the Euler - harmonic information entropy of the set; the formula used is as follows: ； Among them, is the Euler-harmonic information entropy of the structural construction data subset Z1, |Z1| is the number of data in Z1, Q is the number of data labels, q is the data label index, and h q is the number of data in Z1 that belong to the data label q, is the modified harmonic series function, is constructed based on the harmonic series and corrected by adding the Euler constant.

6. A noise reduction and optimization management method for a compressed air system according to claim 4, characterized in that: In step S33, the node splitting rules specifically include the following steps: Step S331: Calculate the Euler gain; subtract the weighted Euler - harmonic information entropy of the left and right child node data sets from the Euler - harmonic information entropy of the data set at the node respectively to obtain the Euler gain; Step S332: Select the best splitting feature; construct a set of maximum Euler gains when splitting at the node using each feature, normalize the set of maximum Euler gains to obtain a normalized set of maximum Euler gains, calculate the selection probability of each feature in the normalized set of maximum Euler gains based on the first control factor β1 to obtain a feature selection probability set, and randomly select a feature as the best splitting feature according to the feature selection probability set; Step S333: Select the best splitting feature value; construct a set of Euler gains when splitting at the node using all possible feature values of the best splitting feature, normalize the set of Euler gains to obtain a normalized set of Euler gains, calculate the selection probability of each feature value in the normalized set of Euler gains based on the second control factor β2 to obtain a feature value selection probability set, and then randomly select a feature value from all possible feature values of the best splitting feature as the best splitting feature value according to the feature value selection probability set; Step S334: Split; use the selected best splitting feature value to split the node to obtain its left and right child nodes.

7. A noise reduction and optimization management method for a compressed air system according to claim 1, characterized in that: In step S1, the integration of the data related to the noise reduction of the compressed air system is to collect and process the historical data related to the noise reduction of the compressed air system; The collection of the historical data related to the noise reduction of the compressed air system is to collect the operation data of the compressed air system, the equipment data of the compressed air system, the noise data, the environmental data, and the noise reduction adjustment data; The processing of the historical data related to the noise reduction of the compressed air system is to perform data cleaning, data normalization, data encoding, and data set construction processing on the historical data related to the noise reduction of the compressed air system to obtain a data set for the noise reduction of the compressed air system.

8. A noise reduction and optimization management method for a compressed air system according to claim 1, characterized in that: In step S4, the noise reduction optimization management collects and processes the data related to the noise reduction of the real-time compressed air system, inputs the processed data related to the noise reduction of the real-time compressed air system into the noise reduction optimization management model for analysis, obtains the corresponding noise reduction adjustment data according to the output data label, generates a noise reduction adjustment strategy, and performs noise reduction optimization management on the compressed air system according to the noise reduction adjustment strategy; The collection of the data related to the noise reduction of the real-time compressed air system is to collect the operation data of the compressed air system, the equipment data of the compressed air system, the noise data and the environmental data; The processing of the data related to the noise reduction of the real-time compressed air system is to perform data cleaning, data normalization and data encoding on the data related to the noise reduction of the real-time compressed air system.

9. A noise reduction and optimization management system for a compressed air system, which is used to implement a noise reduction and optimization management method for a compressed air system as described in any one of claims 1-8, characterized in that: It includes a module for integrating data related to the noise reduction of the compressed air system, a module for optimizing data related to the noise reduction of the compressed air system, a module for constructing a noise reduction optimization management model, and a noise reduction optimization management module; The module for integrating data related to the noise reduction of the compressed air system collects and processes the historical data related to the noise reduction of the compressed air system, and sends the data to the module for optimizing data related to the noise reduction of the compressed air system; The module for optimizing data related to the noise reduction of the compressed air system obtains the dynamic credibility coefficient of the data in terms of features according to the alienation index of the data relative to the features, calculates the weighted distance and constructs the nearest neighbor set, obtains the cluster deviation coefficient of the data, identifies and deletes the abnormal data, and sends the data to the module for constructing a noise reduction optimization management model; The module for constructing a noise reduction optimization management model introduces a modified harmonic series function to obtain the Euler-harmonic information entropy of the set, selects the best splitting feature and the best splitting feature value according to the Euler gain and the control factor, obtains a decision tree constructed based on the data subset constructed by the structure, and then redefines the label of the leaf node in the decision tree based on the data subset defined by the label. Based on the ensemble learning technology, the decision trees are combined to obtain a noise reduction optimization management model, and the data is sent to the noise reduction optimization management module; The noise reduction optimization management module obtains the corresponding noise reduction adjustment data according to the data label output by the noise reduction optimization management model, generates a noise reduction adjustment strategy, and performs noise reduction optimization management on the compressed air system according to the noise reduction adjustment strategy.

Images