Airbag pump operation abnormity detection method, equipment and medium

Abstract

The embodiment of the invention discloses a method and equipment for detecting abnormal operation of an airbag pump and a medium, belongs to the technical field of airbag pumps, and solves the problem that the working efficiency of the airbag pump is reduced due to equipment shutdown caused by gradual deterioration of faults due to neglect of the initial fault abnormity of the airbag pump. Feature extraction is carried out on the operation data of the airbag pump through a Transform model, and a feature space of the airbag pump is constructed; performing data detection on the characteristic space of the airbag pump to determine actual operation abnormal data of the airbag pump; on the basis of a time sequence prediction model of Transform and the feature data in the feature space of the airbag pump, the operation state of the airbag pump in a future preset time period is predicted, and predicted operation abnormal data of the airbag pump are obtained; and combining the actual abnormal operation data of the air bag pump with the predicted abnormal operation data of the air bag pump to obtain an abnormal data set of the air bag pump, determining the abnormal operation duration of the air bag pump and the abnormal degree value of the air bag pump, and performing abnormal alarm of the air bag pump.

Description

Technical Field

[0001] This application relates to the technical field of air bladder pumps, and particularly to a method, device, and medium for detecting abnormal operation of an air bladder pump. Background Art

[0002] In the current industrial production and fluid transportation fields, as an important conveying device, the air bladder pump is widely used in the conveying tasks of various fluid media due to its unique structural design and high working performance. However, although the air bladder pump performs well in most application scenarios, its fault monitoring and diagnosis technologies still face many challenges, which is particularly prominent in current industrial practices.

[0003] Traditional air bladder pump fault monitoring methods often rely on manual inspections and regular maintenance. This method is not only time-consuming and laborious but also difficult to detect potential faults in the equipment in a timely manner. Since the working environment of the air bladder pump is usually complex, it is difficult to accurately capture various parameter changes during its operation with the naked eye or simple detection tools. Therefore, in the initial stage of a fault, even if the equipment has shown some subtle abnormal manifestations, they are often ignored due to the lack of effective monitoring means, resulting in the gradual deterioration of the fault, ultimately leading to equipment shutdown or more serious consequences, thus reducing the working efficiency of the air bladder pump. Summary of the Invention

[0004] Embodiments of this application provide a method, device, and medium for detecting abnormal operation of an air bladder pump to solve the following technical problems: In the initial stage of an air bladder pump fault, even if the equipment has shown some subtle abnormal manifestations, they are often ignored due to the lack of effective monitoring means, resulting in the gradual deterioration of the fault, ultimately leading to equipment shutdown or more serious consequences, thus reducing the working efficiency of the air bladder pump.

[0005] Embodiments of this application adopt the following technical solutions:

[0006] An embodiment of the present application provides a method for detecting abnormal operation of an air bag pump, including: obtaining the operation data of the air bag pump; extracting features from the operation data of the air bag pump through the self-attention mechanism of the Transformer model, and constructing an air bag pump feature space based on the extracted feature data; detecting the data in the air bag pump feature space through a feature point detection algorithm to determine the actual abnormal operation data of the air bag pump; predicting the operation state of the air bag pump within a preset future time period based on the time series prediction model of the Transformer and the feature data in the air bag pump feature space to obtain the predicted abnormal operation data of the air bag pump; combining the actual abnormal operation data of the air bag pump with the predicted abnormal operation data of the air bag pump to obtain an air bag pump abnormal data group, and determining the abnormal operation duration of the air bag pump based on the air bag pump abnormal data group; comparing the air bag pump abnormal data group with a preset abnormal database to determine the air bag pump abnormal degree value; and performing an abnormal alarm for the air bag pump based on the air bag pump abnormal operation duration and the air bag pump abnormal degree value.

[0007] In the embodiment of the present application, the key features in the operation data of the air bag pump are extracted through the self-attention mechanism of the Transformer model, and a feature space is constructed to achieve comprehensive monitoring of the operation state of the air bag pump, and the operation data of the air bag pumps at different structural positions can be associated to more comprehensively and accurately identify the abnormal conditions of the equipment. The feature point detection algorithm can accurately identify potential faults of the equipment in the feature space, avoiding further deterioration of the faults resulting in shutdown or more serious consequences. Through the time series prediction ability of the Transformer model, the operation state of the air bag pump within a preset future time period is predicted, so as to discover potential abnormal trends in advance and provide sufficient time for maintenance personnel to carry out fault prevention and preparation work. The actual abnormal operation data and the predicted abnormal operation data are combined to form a complete abnormal data group, so that the maintenance personnel can quickly obtain the fault type and severity of the equipment, and then take appropriate countermeasures to timely maintain the air bag pump fault.

[0008] In an implementation manner of the present application, the self-attention mechanism of the Transformer model is used to extract features from the operation data of the airbag pump, and a feature space is constructed based on the extracted feature data, specifically including: converting the operation data of the airbag pump into sequence data, obtaining sequence features based on the acquisition position order of each data in the operation data of the airbag pump, performing positional encoding on the sequence features to generate embedding vectors corresponding to the sequence features; through the global self-attention mechanism of the Transformer model, determining a first derivative vector group corresponding to each embedding vector respectively, and performing multiplication and summation processing on the first derivative vector group to obtain the global most relevant information corresponding to each operation feature of the airbag pump; wherein, the global most relevant information is the comprehensive correlation information between each operation feature of the airbag pump and all other operation features of the airbag pump; through the local self-attention mechanism of the Transformer model, determining a second derivative vector group corresponding to each embedding vector respectively, and performing multiplication and summation processing on the second derivative vector group to obtain the adjacent most relevant information corresponding to each operation feature of the airbag pump; wherein, the adjacent most relevant information is the correlation information adjacent to the position of the current operation feature of the airbag pump; based on the global most relevant information and the adjacent most relevant information, obtaining the correlation relationship between the operation features of the airbag pump, and constructing the airbag pump feature space based on the correlation relationship between the operation features of the airbag pump.

[0009] In an implementation manner of the present application, constructing the airbag pump feature space based on the correlation relationship between the operation features of the airbag pump specifically includes: obtaining a global correlation matrix based on the global most relevant information, and obtaining a local correlation matrix based on the adjacent most relevant information; determining a corresponding weight combination based on the operation feature type of the airbag pump, and performing weighted processing on the global correlation matrix and the local correlation matrix based on the weight combination to obtain an integrated correlation matrix; comparing the integrated correlation matrix with a preset matrix threshold to screen out the correlations that do not meet the correlation conditions; generating a node graph based on the remaining correlations to construct the airbag pump feature space.

[0010] In an implementation manner of the present application, generating a node graph based on the remaining correlations to construct the airbag pump feature space specifically includes: determining the position of each node in the airbag pump based on the node graph; and determining the feature type corresponding to each node respectively; determining the correlation coefficient between each node based on the position and the feature type; adjusting the eigenvalue corresponding to each node based on the correlation coefficient to construct the airbag pump feature space based on the adjusted eigenvalue.

[0011] In one implementation of the present application, data detection is performed on the feature space of the air bladder pump through a feature point detection algorithm to determine the actual abnormal operation data of the air bladder pump, which specifically includes: sorting different types of data in the feature space of the air bladder pump to generate multiple feature trajectories; determining the low-dimensional representations corresponding to each data point in the multiple feature trajectories through a pre-set autoencoder, and determining the distances between each data point and other data points based on the low-dimensional representations to construct a distance matrix; determining the K nearest neighbors corresponding to each data point based on the distance matrix; and performing data detection on each data point in the feature trajectory based on the pre-set outlier threshold and the K nearest neighbors to determine the actual abnormal operation data of the air bladder pump.

[0012] In one implementation of the present application, based on the time series prediction model of Transformer and the feature data in the feature space of the air bladder pump, the operation state of the air bladder pump in a future preset time period is predicted to obtain the predicted abnormal operation data of the air bladder pump, which specifically includes: inputting the feature data in the feature space of the air bladder pump into the time series prediction model of Transformer to obtain the predicted operation data of the air bladder pump in a future preset time period; constructing a line graph of the operation data based on the predicted operation data of the air bladder pump, determining the differences between adjacent data in the line graph of the operation data, and determining the operation change section and the operation stable section of the air bladder pump based on the differences; comparing the differences between adjacent data corresponding to the operation stable section of the air bladder pump with a first preset difference threshold to determine the first abnormal data; comparing the differences between adjacent data corresponding to the operation change section of the air bladder pump with a second preset difference threshold to determine the second abnormal data; and obtaining the predicted abnormal operation data of the air bladder pump based on the first abnormal data and the second abnormal data.

[0013] In one implementation of the present application, the actual abnormal operation data of the air bladder pump and the predicted abnormal operation data of the air bladder pump are combined to obtain an abnormal data group of the air bladder pump, and the abnormal operation duration of the air bladder pump is determined based on the abnormal data group of the air bladder pump, which specifically includes: arranging the actual abnormal operation data of the air bladder pump and the predicted abnormal operation data of the air bladder pump in chronological order and marking the corresponding time points; determining the time differences between adjacent time points, and determining the abnormal change rate of the air bladder pump based on the ratio between the latter time difference and the former time difference; and inputting the abnormal change rate of the air bladder pump into the predicted operation duration prediction model of the air bladder pump to output the abnormal operation duration of the air bladder pump through the predicted operation duration prediction model of the air bladder pump.

[0014] In an implementation manner of the present application, the abnormal data group of the air bladder pump is compared with a preset abnormal database to determine the abnormal degree value of the air bladder pump, which specifically includes: comparing the abnormal data group of the air bladder pump with the preset abnormal database to obtain a reference abnormal type; wherein, the preset abnormal database includes multiple abnormal data groups of the air bladder pump, and also includes the corresponding air bladder pump abnormal types of multiple abnormal data groups of the air bladder pump; determining the abnormal data difference between the abnormal data group of the air bladder pump and the data group corresponding to the reference abnormal type, and determining a first weight based on the abnormal data difference; comparing the reference abnormal type with a preset abnormal type level table to determine a second weight based on the comparison result; and determining the abnormal degree value of the abnormal data group of the air bladder pump based on the abnormal data group of the air bladder pump, the first weight, and the second weight.

[0015] An embodiment of the present application provides an air bladder pump operation abnormality detection device, including: at least one processor; and a memory communicatively connected to the at least one processor; wherein, the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to: obtain the operation data of the air bladder pump; extract features from the operation data of the air bladder pump through the self-attention mechanism of the Transformer model, and construct an air bladder pump feature space based on the extracted feature data; detect the data in the air bladder pump feature space through a feature point detection algorithm to determine the actual operation abnormal data of the air bladder pump; predict the operation state of the air bladder pump within a preset future time period based on the time series prediction model of the Transformer and the feature data in the air bladder pump feature space to obtain the predicted operation abnormal data of the air bladder pump; combine the actual operation abnormal data of the air bladder pump with the predicted operation abnormal data of the air bladder pump to obtain an abnormal data group of the air bladder pump, and determine the abnormal operation duration of the air bladder pump based on the abnormal data group of the air bladder pump; compare the abnormal data group of the air bladder pump with a preset abnormal database to determine the abnormal degree value of the air bladder pump; and perform an abnormal alarm for the air bladder pump based on the abnormal operation duration of the air bladder pump and the abnormal degree value of the air bladder pump.

[0016] A non - volatile computer storage medium provided by an embodiment of the present application stores computer - executable instructions, and the computer - executable instructions are set as follows: obtain the operation data of the air - bag pump; perform feature extraction on the operation data of the air - bag pump through the self - attention mechanism of the Transformer model, and construct an air - bag pump feature space based on the extracted feature data; perform data detection on the air - bag pump feature space through a feature - point detection algorithm to determine the actually abnormal operation data of the air - bag pump; predict the operation state of the air - bag pump within a preset future time period based on the time - series prediction model of the Transformer and the feature data in the air - bag pump feature space to obtain the predicted abnormal operation data of the air - bag pump; combine the actually abnormal operation data of the air - bag pump with the predicted abnormal operation data of the air - bag pump to obtain an air - bag pump abnormal data group, and determine the abnormal operation duration of the air - bag pump based on the air - bag pump abnormal data group; compare the air - bag pump abnormal data group with a preset abnormal database to determine the air - bag pump abnormal degree value; perform air - bag pump abnormal alarm based on the air - bag pump abnormal operation duration and the air - bag pump abnormal degree value.

[0017] The above - mentioned at least one technical solution adopted in the embodiment of the present application can achieve the following beneficial effects: The embodiment of the present application extracts the key features in the operation data of the air - bag pump through the self - attention mechanism of the Transformer model and constructs a feature space, realizing the comprehensive monitoring of the operation state of the air - bag pump, and can correlate the operation data of air - bag pumps in different structural positions to more comprehensively and accurately identify the abnormal conditions of the equipment. The feature - point detection algorithm can accurately identify potential faults of the equipment in the feature space, avoiding further deterioration of the faults resulting in shutdown or more serious consequences. Through the time - series prediction ability of the Transformer model, predict the operation state of the air - bag pump within a preset future time period, so as to discover potential abnormal trends in advance and provide sufficient time for maintenance personnel to carry out fault prevention and preparation work. Combine the actually abnormal operation data with the predicted abnormal operation data to form a complete abnormal data group, so that the maintenance personnel can quickly obtain the fault type and severity of the equipment, and then take appropriate countermeasures to timely maintain the air - bag pump fault. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the drawings described below are only some embodiments recorded in the present application. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings. In the drawings:

[0019] Figure 1 It is a flowchart of a method for detecting abnormal operation of an air - bag pump provided by an embodiment of the present application;

[0020] Figure 2 This is a schematic structural diagram of an abnormal operation detection device for an air bladder pump provided by an embodiment of the present application.

[0021] Reference numerals:

[0022] 200: Abnormal operation detection device for an air bladder pump, 201: Processor, 202: Memory. Specific embodiments

[0023] The embodiments of the present application provide an abnormal operation detection method, device and medium for an air bladder pump.

[0024] In order to enable those skilled in the art to better understand the technical solutions in the present application, the following will clearly and completely describe the technical solutions in the embodiments of the present application with reference to the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments of this specification, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present application.

[0025] The following will detail the technical solutions proposed in the embodiments of the present invention through the accompanying drawings.

[0026] Figure 1 This is a flowchart of an abnormal operation detection method for an air bladder pump provided by an embodiment of the present application. As Figure 1 shown, the abnormal operation detection method for an air bladder pump includes the following steps:

[0027] Step 101, obtain the operation data of the air bladder pump.

[0028] In an embodiment of the present application, a plurality of operation data of the air bladder pump are obtained. For example, the air supply pressure data, flow rate data, discharge pressure data, transmission efficiency, vibration data at different positions of the air bladder pump, temperature data at different positions of the air bladder pump, etc. can be obtained. The data can be obtained in real time through sensors arranged at different positions of the air bladder pump. For example, through a pressure sensor installed on the air supply pipeline, the air supply pressure when the air bladder pump is working is monitored in real time; a pressure sensor is installed on the discharge pipeline to obtain the pressure value of the gas discharged by the air bladder pump; vibration sensors are installed at different positions of the air bladder pump (such as bearings, housings, etc.) to monitor the vibration condition of the air bladder pump in real time; a temperature sensor is used to monitor the temperature change at different positions of the air bladder pump.

[0029] Step 102, perform feature extraction on the operation data of the air bladder pump through the self-attention mechanism of the Transformer model, and construct an air bladder pump feature space based on the extracted feature data.

[0030] In one embodiment of the present application, the operation data of the air bladder pump is converted into sequence data, sequence features are obtained based on the acquisition position order of each data in the operation data of the air bladder pump, position encoding is performed on the sequence features, and embedding vectors corresponding to the sequence features are generated. Through the global self-attention mechanism of the Transformer model, the first derivative vector groups respectively corresponding to the embedding vectors are determined, and the first derivative vector groups are multiplied and summed to obtain the global most relevant information respectively corresponding to each operation feature of the air bladder pump; wherein, the global most relevant information is the comprehensive association information between each operation feature of the air bladder pump and all other operation features of the air bladder pump. Through the local self-attention mechanism of the Transformer model, the second derivative vector groups respectively corresponding to the embedding vectors are determined, the global association matrix is obtained based on the global most relevant information, and the local association matrix is obtained based on the adjacent most relevant information. Based on the operation feature type of the air bladder pump, the corresponding weight combination is determined, and the global association matrix and the local association matrix are weighted based on the weight combination to obtain the integrated association matrix. The integrated association matrix is compared with the preset matrix threshold to screen out the associations that do not meet the association conditions. Based on the remaining associations, a node graph is generated to construct the air bladder pump feature space. The groups are multiplied and summed to obtain the adjacent most relevant information respectively corresponding to each operation feature of the air bladder pump; wherein, the adjacent most relevant information is the association information adjacent to the position of the current operation feature of the air bladder pump. Based on the global most relevant information and the adjacent most relevant information, the association relationship between the operation features of the air bladder pump is obtained, and the air bladder pump feature space is constructed based on the association relationship between the operation features of the air bladder pump.

[0031] Specifically, according to the installation position order of the sensors on the air bladder pump, these data are organized into sequence data, and position encoding is performed on each data point in the sequence to distinguish their positions in the sequence, and each data point is converted into an embedding vector. Through the global self-attention mechanism of the Transformer model, the attention weights between each embedding vector and all other embedding vectors are calculated, wherein these weights reflect the association degree between each embedding vector and other embedding vectors. According to the attention weights, the first derivative vector group corresponding to each embedding vector is generated, wherein the first derivative vector group contains the comprehensive association information between each embedding vector and other embedding vectors. The first derivative vector groups are multiplied and summed to obtain the global most relevant information corresponding to each operation feature of the air bladder pump.

[0032] Further, using the local self-attention mechanism of the Transformer model, calculate the attention weights between each embedding vector and its adjacent embedding vectors, where these weights reflect the degree of association between each embedding vector and its adjacent embedding vectors. According to the attention weights, generate a second derivative vector group corresponding to each embedding vector. The second derivative vector group contains the association information between each embedding vector and its adjacent embedding vectors. Perform multiplication and summation processing on the second derivative vector group to obtain the adjacent most relevant information corresponding to each air bladder pump operation characteristic.

[0033] Further, based on the global most relevant information and the adjacent most relevant information, extract the association relationships between the air bladder pump operation characteristics. These association relationships reflect the mutual influence and dependence between different operation characteristics. Using the extracted association relationships, construct the feature space of the air bladder pump, where each dimension in the feature space of the embodiment of the present application corresponds to an air bladder pump operation characteristic, and the relative positions and distances between the features reflect the degree of association between them.

[0034] Suppose the air bladder pump to be detected, and its operation data includes supply pressure, flow rate, discharge pressure, data from two vibration sensors, and data from two temperature sensors. Organize these data into a sequence according to the sensor installation order and convert them into embedding vectors. Then, use the global and local self-attention mechanisms of the Transformer model to calculate the degree of association between each embedding vector and other embedding vectors. For example, for the supply pressure embedding vector, the global self-attention mechanism can output that it has a strong association with the flow rate embedding vector and the discharge pressure embedding vector, and a weak association with the vibration and temperature embedding vectors. And the local self-attention mechanism can output that it has the strongest association with the adjacent flow rate embedding vector. Based on this association information, construct the feature space of the air bladder pump, where the connection lines between the supply pressure feature and the flow rate and discharge pressure features are thicker to indicate a stronger association relationship, and the connection lines with the vibration and temperature features are thinner to indicate a weaker association relationship. Through the constructed feature space, the operation state of the air bladder pump can be better obtained, and potential faults or abnormal conditions can be discovered.

[0035] In an embodiment of the present application, obtain a global association matrix based on the global most relevant information, and obtain a local association matrix based on the adjacent most relevant information. Determine the corresponding weight combination based on the air bladder pump operation characteristic type, and perform weighted processing on the global association matrix and the local association matrix based on the weight combination to obtain an integrated association matrix. Compare the integrated association matrix with a preset matrix threshold to screen out the associations that do not meet the association conditions. Generate a node graph based on the remaining associations to construct the air bladder pump feature space.

[0036] Specifically, organize the globally most relevant information into a matrix, namely the global correlation matrix. Each element in the matrix represents the degree of correlation between two features. The larger the element value, the stronger the correlation. And organize the adjacent most relevant information into a matrix, namely the local correlation matrix. Each element in the matrix represents the degree of correlation between two adjacent features. According to the type and importance of the operation characteristics of the air bladder pump, different weights are assigned to the global correlation matrix and the local correlation matrix, where the assigned weights can be obtained based on expert experience. Using the determined weight combination, perform weighted processing on the global correlation matrix and the local correlation matrix. The purpose of the weighted processing is to combine the global and local correlation information to obtain a more comprehensive and accurate integrated correlation matrix.

[0037] Furthermore, compare the integrated correlation matrix with a preset matrix threshold to retain strongly correlated feature pairs and remove weakly correlated or irrelevant feature pairs, thereby simplifying the structure of the feature space. Based on the remaining correlations, that is, the elements in the integrated correlation matrix after screening, generate a node graph. Each node in the node graph represents an operation feature of the air bladder pump, and the connection lines between the nodes represent the correlations between the features.

[0038] In an embodiment of the present application, based on the node graph, determine the positions of each node in the air bladder pump. And determine the corresponding feature types of each node. Based on the positions and feature types, determine the correlation coefficients between each node. Adjust the feature values corresponding to each node based on the correlation coefficients to construct the air bladder pump feature space based on the adjusted feature values.

[0039] Specifically, in the node graph, each node represents an operation feature in the air bladder pump. The position of the node can be determined according to the actual installation position of the sensor on the air bladder pump. Each node corresponds to a specific operation feature of the air bladder pump, such as supply pressure, flow rate, discharge pressure, vibration, temperature, etc. In the node graph, the connection lines between the nodes represent the correlations between the features. The thickness or color of the connection lines can represent the strength of the correlation. By calculating the correlation between the nodes, the correlation coefficients between the nodes are obtained. The magnitude of the correlation coefficient reflects the strength and direction of the interaction between the features. After determining the correlation coefficients, adjust the feature value of each node by weighted summation. The adjusted feature value more accurately reflects the actual operation state of the air bladder pump and takes into account the mutual influence between the features.

[0040] Step 103: Detect the data in the air bladder pump feature space through a feature point detection algorithm to determine the abnormal data of the actual operation of the air bladder pump.

[0041] In one embodiment of the present application, different types of data in the air bladder pump feature space are sorted to generate multiple feature trajectories. By presetting an autoencoder, the low-dimensional representations corresponding to each data point in the multiple feature trajectories are determined, and based on the low-dimensional representations, the distances between each data point and other data points are determined to construct a distance matrix. Based on the distance matrix, the K nearest neighbors corresponding to each data point are determined. Based on the preset outlier threshold and the K nearest neighbors, data detection is performed on each data point in the feature trajectory to determine the abnormal data of the actual operation of the air bladder pump.

[0042] Specifically, in the air bladder pump feature space, different types of data, such as supply pressure, flow rate, discharge pressure, vibration, temperature, etc., are sorted to ensure the continuity between data points, forming multiple feature trajectories. Each feature trajectory represents a specific type of air bladder pump operation data.

[0043] Furthermore, the high-dimensional data points in the feature trajectory are converted into low-dimensional representations through a preset autoencoder, and the low-dimensional representations can more efficiently calculate the distances between data points. Using the low-dimensional representations, the Euclidean distances between each data point and other data points are calculated, and these distances are organized into a matrix, namely the distance matrix. Each element in the matrix represents the distance between two data points.

[0044] Furthermore, for each data point in the distance matrix, its K nearest neighbors are found, that is, the K data points with the closest distances, and a preset outlier threshold is set. For each data point in the feature trajectory, the sum of its distances to the K nearest neighbors is calculated and compared with the outlier threshold. If the sum of the distances of the data point exceeds the threshold, it is marked as an outlier. These outliers indicate abnormal conditions or faults in the operation of the air bladder pump.

[0045] Step 104: Based on the time series prediction model based on Transformer and the feature data in the air bladder pump feature space, predict the operation state of the air bladder pump within a preset future time period to obtain the predicted abnormal operation data of the air bladder pump.

[0046] In an embodiment of the present application, the feature data in the air bladder pump feature space is input into the time series prediction model of the Transformer to obtain the predicted operation data of the air bladder pump within a preset future time period. Based on the predicted operation data of the air bladder pump, a line graph of the operation data is constructed, the difference between adjacent data in the line graph of the operation data is determined, and the operation change section and the operation stable section of the air bladder pump are determined based on the difference. The difference between adjacent data corresponding to the operation stable section of the air bladder pump is compared with a first preset difference threshold to determine the first abnormal data. The difference between adjacent data corresponding to the operation change section of the air bladder pump is compared with a second preset difference threshold to determine the second abnormal data. Based on the first abnormal data and the second abnormal data, the predicted operation abnormal data of the air bladder pump is obtained.

[0047] Specifically, the feature data in the air bladder pump feature space, such as the supply air pressure, flow rate, temperature, etc., is used as the input of the Transformer time series prediction model. Among them, these feature data are arranged in a time series, that is, each feature has a corresponding timestamp. The Transformer model learns the temporal dependence relationships between the feature data through the self-attention mechanism. Based on these relationships, the model can predict the operation data of the air bladder pump within a preset future time period.

[0048] Furthermore, the predicted future operation data of the Transformer model is plotted as a line graph, and each point in the line graph represents the predicted value at a time point. Calculate the difference between adjacent data points in the line graph. These differences reflect the change trend of the operation data of the air bladder pump. Based on the magnitude of the difference, the line graph is divided into the operation change section and the operation stable section of the air bladder pump. Among them, the change section refers to the part with a larger difference, indicating that the operation data of the air bladder pump has changed significantly in a short period of time, and the stable section refers to the part with a smaller difference, indicating that the operation data of the air bladder pump remains relatively stable.

[0049] Furthermore, the difference between adjacent data corresponding to the stable section is compared with the first preset difference threshold. If a certain difference exceeds the first preset difference threshold, then this data point is considered the first abnormal data. Among them, the first preset difference threshold is used to determine the stability of the stable section, that is, the difference between adjacent data remains within a stable range.

[0050] Furthermore, in the change section, although the operation data of the air bladder pump will change significantly, this change is usually gradual. The difference between adjacent data corresponding to the change section is compared with the second preset difference threshold. If a certain difference exceeds the second preset difference threshold, for example, the change is too drastic or does not conform to the expected change trend, then this data point is considered the second abnormal data. The first abnormal data and the second abnormal data are summarized to obtain a complete set of the predicted operation abnormal data of the air bladder pump.

[0051] Step 105: Combine the actual abnormal operation data of the air bladder pump with the predicted abnormal operation data of the air bladder pump to obtain an air bladder pump abnormal data group, and determine the abnormal operation duration of the air bladder pump based on the air bladder pump abnormal data group.

[0052] In an embodiment of the present application, arrange the actual abnormal operation data of the air bladder pump and the predicted abnormal operation data of the air bladder pump in chronological order, and mark the corresponding time points. Determine the time difference between adjacent time points, and determine the abnormal change rate of the air bladder pump based on the ratio between the latter time difference and the former time difference. Input the abnormal change rate of the air bladder pump into the predicted operable duration model of the air bladder pump, so as to output the abnormal operation duration of the air bladder pump through the predicted operable duration model of the air bladder pump.

[0053] Specifically, arrange these two types of abnormal data in chronological order respectively. Each data point corresponds to a specific time point, and this time point represents the moment when the abnormal event occurs. In the arranged data sequence, mark each abnormal data point with the corresponding time point. These time points can be specific dates and times, or offsets relative to a certain starting point. Calculate the time difference between adjacent abnormal data points. The time difference represents the time length between two abnormal events. Based on the ratio between the latter time difference and the former time difference, calculate the abnormal change rate of the air bladder pump. If the ratio is greater than 1, it means that the frequency of abnormal events is accelerating; if the ratio is less than 1, it means that the frequency of abnormal events is slowing down; if the ratio is close to 1, it means that the frequency of abnormal events is relatively stable.

[0054] Furthermore, use the calculated abnormal change rate as the input of the predicted operable duration model of the air bladder pump. Among them, the training process of the predicted operable duration model of the air bladder pump is to use the change rate of historical abnormal events as sample inputs, and use the operable duration of the air bladder pump corresponding to the input samples as outputs to train a preset neural network model to obtain the predicted operable duration model of the air bladder pump, which is used to predict the operable duration of the air bladder pump under a given abnormal change rate. Through the predicted operable duration model of the air bladder pump, obtain the predicted operable duration of the air bladder pump under the current abnormal change rate. This duration can evaluate the remaining service life of the air bladder pump and whether maintenance or replacement needs to be carried out in advance.

[0055] Step 106: Compare the air bladder pump abnormal data group with the preset abnormal database to determine the air bladder pump abnormal degree value.

[0056] In an embodiment of the present application, the abnormal data group of the air bladder pump is compared with a preset abnormal database to obtain a reference abnormal type. The preset abnormal database includes various abnormal data groups of the air bladder pump, and also includes the corresponding air bladder pump abnormal types for various abnormal data groups of the air bladder pump. Determine the abnormal data difference between the abnormal data group of the air bladder pump and the data group corresponding to the reference abnormal type, and determine the first weight based on the abnormal data difference. Compare the reference abnormal type with the preset abnormal type grading table to determine the second weight based on the comparison result. Based on the abnormal data group of the air bladder pump, the first weight, and the second weight, determine the abnormal degree value of the abnormal data group of the air bladder pump.

[0057] Specifically, the identified abnormal data group of the air bladder pump is compared with the preset abnormal database. The preset abnormal database in the embodiment of the present application includes various known abnormal data groups of the air bladder pump and their corresponding abnormal types. For example, assume that a set of operation data of an air bladder pump is monitored in real time, and it is found that both the pressure value and the flow value are lower than the normal value, forming an abnormal data group. Compare this abnormal data group with the preset abnormal database and find that it is most similar to the abnormal data group of the "blockage" type in the database.

[0058] Furthermore, determine the data difference between the abnormally monitored data group and the data group corresponding to the reference abnormal type. The difference reflects the deviation degree between the actual abnormality and the known abnormal type. Based on this abnormal data difference, determine a first weight. The size of the weight is usually inversely proportional to the abnormal data difference, that is, the smaller the difference, the larger the weight, indicating that the abnormality is closer to a certain known abnormal type. Compare the determined reference abnormal type with the preset abnormal type grading table. This grading table usually classifies abnormal types into different levels according to the severity or impact degree of the abnormality. Based on the comparison result, determine a second weight, which reflects the severity of the reference abnormal type.

[0059] Furthermore, based on the abnormally monitored data group of the air bladder pump, the first weight, and the second weight, determine the abnormal degree value of the abnormal data group by weighted summation.

[0060] Step 107: Perform an abnormal alarm for the air bladder pump based on the abnormal operation duration of the air bladder pump and the abnormal degree value of the air bladder pump.

[0061] In one embodiment of the present application, when an abnormality of the air bag pump is detected, the time when the abnormality starts is recorded, and based on the start time of the abnormality and the current time, the duration of the abnormal operation of the air bag pump is calculated. Combining the air bag pump abnormality degree value and the abnormal operation duration, an alarm trigger condition is set. For example, when the abnormality degree exceeds a certain level and the abnormal operation duration exceeds the set time, an alarm is triggered. The alarm information should include key information such as the abnormal type, abnormal degree, and abnormal operation duration of the air bag pump, so that relevant personnel can quickly understand the fault situation and make a response.

[0062] An embodiment of the present application provides an air bag pump abnormal operation detection device 200, including: at least one processor 201; and a memory 202 communicatively connected to the at least one processor 201; wherein, the memory 202 stores instructions executable by the at least one processor 201, and the instructions are executed by the at least one processor 201 to enable the at least one processor 201 to: obtain the air bag pump operation data; perform feature extraction on the air bag pump operation data through the self-attention mechanism of the Transformer model, and construct an air bag pump feature space based on the extracted feature data; perform data detection on the air bag pump feature space through a feature point detection algorithm to determine the actually abnormal operation data of the air bag pump; predict the operation state of the air bag pump within a preset time period in the future based on the time series prediction model of the Transformer and the feature data in the air bag pump feature space to obtain the predicted abnormal operation data of the air bag pump; combine the actually abnormal operation data of the air bag pump with the predicted abnormal operation data of the air bag pump to obtain an air bag pump abnormal data group, and determine the abnormal operation duration of the air bag pump based on the air bag pump abnormal data group; compare the air bag pump abnormal data group with a preset abnormal database to determine the air bag pump abnormality degree value; perform an air bag pump abnormal alarm based on the air bag pump abnormal operation duration and the air bag pump abnormality degree value.

[0063] A non-volatile computer storage medium provided by an embodiment of the present application stores computer-executable instructions, and the computer-executable instructions are set to: obtain the operation data of the air bag pump; perform feature extraction on the operation data of the air bag pump through the self-attention mechanism of the Transformer model, and construct an air bag pump feature space based on the extracted feature data; perform data detection on the air bag pump feature space through a feature point detection algorithm to determine the actual operation abnormal data of the air bag pump; based on the time series prediction model of the Transformer and the feature data in the air bag pump feature space, predict the operation state of the air bag pump within a preset future time period to obtain the predicted operation abnormal data of the air bag pump; combine the actual operation abnormal data of the air bag pump with the predicted operation abnormal data of the air bag pump to obtain an air bag pump abnormal data group, and determine the abnormal operation duration of the air bag pump based on the air bag pump abnormal data group; compare the air bag pump abnormal data group with a preset abnormal database to determine the air bag pump abnormal degree value; perform air bag pump abnormal alarm based on the air bag pump abnormal operation duration and the air bag pump abnormal degree value.

[0064] The various embodiments in the present application are described in a progressive manner. For the same or similar parts among the various embodiments, reference can be made to each other. Each embodiment focuses on the differences from other embodiments. In particular, for the embodiments of the device, equipment, and non-volatile computer storage medium, since they are basically similar to the method embodiments, the description is relatively simple, and the relevant parts can refer to the partial description of the method embodiments.

[0065] The above are only the embodiments of the present application and are not used to limit the present application. For those skilled in the art, various changes and modifications can be made to the embodiments of the present application. These modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the various embodiments of the present application.

Claims

1. A method for detecting abnormal operation of a bellows pump, characterized in that: The method comprises: Obtain bellows pump operation data; Extracting features from the wind bladder pump operation data through the self-attention mechanism of the Transformer model, and constructing a wind bladder pump feature space based on the extracted feature data; Performing data detection on the characteristic space of the air bag pump by using a characteristic point detection algorithm to determine the actual abnormal operation data of the air bag pump; Based on the Transformer time series prediction model and the feature data in the feature space of the wind bladder pump, the operation state of the wind bladder pump in a preset time period in the future is predicted to obtain the predicted abnormal operation data of the wind bladder pump; The actual abnormal operation data of the wind bladder pump is combined with the predicted abnormal operation data of the wind bladder pump to obtain an abnormal data group of the wind bladder pump, and the abnormal operation time of the wind bladder pump is determined based on the abnormal data group of the wind bladder pump; Comparing the wind bag pump abnormal data group with a preset abnormal database to determine the abnormality degree value of the wind bag pump; Based on the abnormal operation time of the bellows pump and the abnormal degree value of the bellows pump, an abnormal alarm of the bellows pump is performed.

2. A method for detecting abnormal operation of a bellows pump according to claim 1, characterized in that: The feature extraction of the wind bladder pump operation data by the self-attention mechanism of the Transformer model and the construction of the wind bladder pump feature space based on the extracted feature data specifically include: Convert the bellows pump operation data into sequence data, obtain sequence features based on the acquisition position order of each data in the bellows pump operation data, perform position encoding on the sequence features, and generate an embedding vector corresponding to the sequence features; Through the global self-attention mechanism of the Transformer model, the first derivative vector groups corresponding to each of the embedded vectors are determined, and the first derivative vector groups are multiplied and summed to obtain the global most relevant information corresponding to each wind bladder pump operation feature; wherein the global most relevant information is the comprehensive correlation information between each wind bladder pump operation feature and all other wind bladder pump operation features; Through the local self-attention mechanism of the Transformer model, the second derivative vector groups corresponding to each of the embedded vectors are determined, and the second derivative vector groups are multiplied and summed to obtain the adjacent most relevant information corresponding to each of the wind bladder pump operation characteristics; wherein the adjacent most relevant information is the associated information adjacent to the current wind bladder pump operation feature position; Based on the global most relevant information and the adjacent most relevant information, the correlation relationship between the wind bladder pump operation characteristics is obtained, and based on the correlation relationship between the wind bladder pump operation characteristics, a wind bladder pump feature space is constructed.

3. A method for detecting abnormal operation of a bellows pump according to claim 2, characterized in that: The step of constructing a wind bladder pump feature space based on the correlation between the wind bladder pump operation features specifically includes: Obtain a global correlation matrix based on the global most relevant information, and obtain a local correlation matrix based on the adjacent most relevant information; Based on the operation characteristic type of the bellows pump, a corresponding weight combination is determined, and based on the weight combination, the global correlation matrix and the local correlation matrix are weighted to obtain an integrated correlation matrix; Comparing the integrated association matrix with a preset matrix threshold to filter out associations that do not meet the association condition; A node graph is generated based on the remaining associations to construct the wind bladder pump feature space.

4. A method for detecting abnormal operation of a bellows pump according to claim 3, characterized in that: The node graph is generated based on the remaining associations to construct the wind bladder pump feature space, specifically including: Based on the node graph, determining the position of each node in the air bag pump; And, determining the feature type corresponding to each of the nodes; Based on the positions and the feature types, determining a correlation coefficient between the nodes; The characteristic value corresponding to each of the nodes is adjusted based on the correlation coefficient, so as to construct the bellows pump characteristic space based on the adjusted characteristic value.

5. A method for detecting abnormal operation of a bellows pump according to claim 1, characterized in that: The method of performing data detection on the characteristic space of the air bag pump by using a characteristic point detection algorithm to determine the actual abnormal operation data of the air bag pump specifically includes: Sorting different types of data in the bellows pump feature space to generate multiple feature trajectories; By presetting an autoencoder, a low-dimensional representation corresponding to each data point in the plurality of feature trajectories is determined, and a distance between each data point and other data points is determined based on the low-dimensional representation to construct a distance matrix; Based on the distance matrix, determine the K nearest neighbors corresponding to each data point; Based on the preset abnormal point threshold and the K nearest neighbors, data detection is performed on each data point in the characteristic trajectory to determine the actual abnormal operation data of the bellows pump.

6. A method for detecting abnormal operation of a bellows pump according to claim 1, characterized in that: The Transformer-based time series prediction model and the feature data in the feature space of the wind bladder pump predict the operating state of the wind bladder pump in a future preset time period to obtain the predicted abnormal operation data of the wind bladder pump, specifically including: Inputting the feature data in the feature space of the wind bladder pump into the time series prediction model of the Transformer to obtain the predicted operation data of the wind bladder pump within a preset time period in the future; Constructing an operation data line graph based on the predicted operation data of the bellows pump, determining the difference between adjacent data in the operation data line graph, and determining an operation change section and a stable operation section of the bellows pump based on the difference; Comparing the difference between adjacent data corresponding to the stable operation section of the bellows pump with a first preset difference threshold to determine the first abnormal data; Comparing the difference between adjacent data corresponding to the change section of the bellows pump operation with a second preset difference threshold to determine the second abnormal data; Based on the first abnormal data and the second abnormal data, the predicted operation abnormal data of the bellows pump is obtained.

7. A method for detecting abnormal operation of a bellows pump according to claim 1, characterized in that: The step of combining the actual abnormal operation data of the wind bladder pump with the predicted abnormal operation data of the wind bladder pump to obtain an abnormal data group of the wind bladder pump, and determining the abnormal operation time of the wind bladder pump based on the abnormal data group of the wind bladder pump, specifically includes: Arrange the actual abnormal operation data of the wind bag pump and the predicted abnormal operation data of the wind bag pump in chronological order, and mark the corresponding time points; Determine the time difference between adjacent time points, and determine the abnormal change rate of the air bag pump based on the ratio between the latter time difference and the previous time difference; The abnormal change rate of the bellows pump is input into a bellows pump operable duration prediction model, so that the abnormal operation duration of the bellows pump is output through the bellows pump operable duration prediction model.

8. A method for detecting abnormal operation of a bellows pump according to claim 7, characterized in that: The step of comparing the abnormal data set of the wind bladder pump with a preset abnormal database to determine the abnormality level of the wind bladder pump specifically includes: Comparing the wind sac pump abnormal data group with a preset abnormal database to obtain a reference abnormal type; wherein the preset abnormal database includes a plurality of wind sac pump abnormal data groups, and also includes wind sac pump abnormal types corresponding to the plurality of wind sac pump abnormal data groups; Determine an abnormal data difference between the abnormal data group of the bellows pump and the data group corresponding to the reference abnormal type, and determine a first weight based on the abnormal data difference; Comparing the reference abnormality type with a preset abnormality type level table to determine a second weight based on the comparison result; Based on the wind bladder pump abnormal data group, the first weight, and the second weight, an abnormality degree value of the wind bladder pump abnormal data group is determined.

9. A device for detecting abnormal operation of a bellows pump, characterized in that: The device comprises a memory for storing computer program instructions and a processor for executing the program instructions, wherein when the computer program instructions are executed by the processor, the device is triggered to execute the method according to any one of claims 1 to 8.

10. A non-volatile computer storage medium storing computer executable instructions, characterized in that: The computer executable instructions can execute the method according to any one of claims 1 to 8.

Images