Equipment fault intelligent early warning system and method based on Internet of Things

Abstract

The invention provides an equipment fault intelligent early warning system and method based on the Internet of Things, and relates to the technical field of fault detection, the system comprises an equipment terminal, a multi-modal data acquisition device, an edge end, a cloud end and a user terminal; the equipment terminal is connected with the cloud end through the multi-modal data acquisition equipment and the edge end in sequence; the equipment terminal is connected with the cloud end through the multi-modal data acquisition equipment, and the multi-modal data acquisition equipment is used for acquiring real-time data of the equipment terminal and transmitting the real-time data to the cloud end; the edge end and the cloud end are both connected with the user terminal, and the user terminal is notified when an equipment fault occurs; the user terminal is connected with the equipment terminal so as to control start and stop of the equipment terminal through the user terminal. The whole system carries out fault early warning on the equipment in a cloud side cooperation mode, and the equipment fault prediction accuracy is ensured while the early warning real-time performance is maintained.

Description

Technical Field

[0001] The present invention relates to the technical field of fault detection, and in particular to an intelligent early warning system and method for equipment faults based on the Internet of Things. Background Art

[0002] IoT technology connects devices, sensors, control systems, etc. to the Internet to achieve real-time monitoring and data collection of device status. Multiple sensors (such as temperature, pressure, vibration, speed, etc.) are installed on the equipment, which can collect various data of equipment operation in real time. Intelligent warning of equipment failure based on IoT refers to real-time monitoring, data collection, anomaly detection and fault warning of equipment through IoT technology, so that when equipment failure data occurs, early warning information is issued in time to help users or enterprises take preventive measures to avoid major impacts of equipment failure on production and operation.

[0003] If equipment fails during operation, it may cause production shutdown, equipment damage, and even affect safety. Through the intelligent early warning system, potential equipment problems can be discovered in time, maintenance measures can be taken in advance, serious consequences can be avoided, maintenance costs and downtime losses can be reduced, and equipment operation efficiency and service life can be improved. This early warning mechanism is crucial to ensure production continuity and equipment stability.

[0004] However, the existing detection of equipment failures is often done by monitoring some equipment data tables and judging whether a failure has occurred based on subjective data conditions. This is difficult to achieve for complex equipment failure monitoring with high-dimensional equipment parameters. In addition, directly judging whether the equipment is abnormal based on single-dimensional data cannot capture the correlation between high-dimensional data, which is prone to misjudgment, increasing the number of equipment shutdowns and maintenance times, and increasing equipment operation and maintenance costs and operating efficiency. Summary of the invention

[0005] In order to solve the technical problems that the detection of equipment failure in the prior art is often still carried out by monitoring some equipment data tables and judging whether a failure occurs by human subjective judgment based on the data situation, which is difficult to achieve for equipment failure monitoring with complex equipment parameters and high-dimensional equipment. In addition, directly judging whether the equipment is abnormal based on single-dimensional data cannot capture the correlation between high-dimensional data, which is prone to misjudgment, increases the number of equipment shutdowns and maintenance, and increases equipment operation and maintenance costs and operating efficiency. The present invention provides an intelligent early warning system and method for equipment failure based on the Internet of Things.

[0006] The technical solution provided by the embodiment of the present invention is as follows:

[0007] First aspect

[0008] An embodiment of the present invention provides an intelligent early warning system for equipment failure based on the Internet of Things, the system comprising: an equipment terminal, a multimodal data acquisition device, an edge terminal, a cloud terminal and a user terminal;

[0009] The device terminal is connected to the cloud through the multimodal data acquisition device and the edge terminal in turn;

[0010] The device terminal is connected to the cloud through a multimodal data acquisition device, which is used to collect real-time data from the device terminal and transmit the real-time data to the cloud;

[0011] Both the edge and the cloud are connected to the user terminal, and the user terminal is notified when a device failure occurs;

[0012] The user terminal is connected to the device terminal so as to control the start and stop of the device terminal through the user terminal.

[0013] Second aspect

[0014] An embodiment of the present invention provides an intelligent early warning method for equipment failure based on the Internet of Things, which is applied to the fault testing system of the first aspect, and the method includes:

[0015] S1: Obtain historical operation data of the device terminal in normal operation state through the cloud, wherein the historical operation data includes multiple multi-dimensional sample points;

[0016] S2: Regularize the covariance matrix with the minimized covariance determinant and construct a regularized covariance matrix;

[0017] S3: constructing an abnormal data recognition model using historical operation data, wherein the abnormal data recognition model includes an input module, an abnormal data identifier based on a regularized covariance matrix, an autoencoder, and an output module connected in sequence;

[0018] S4: Send the abnormal data recognition model to the edge end;

[0019] S5: Collecting real-time operation data of the device terminal through a multimodal data collection device;

[0020] S6: Analyze the real-time operation data through the abnormal data recognition model to determine whether there is abnormal operation data. If so, output the abnormal component corresponding to the abnormal operation data and proceed to step S8. Otherwise, proceed to step S7.

[0021] S7: Upload the real-time operation data to the historical operation data in the cloud to update the historical operation data, and return to step S3;

[0022] S8: Send the abnormal component to the user terminal for abnormal reminder.

[0023] The beneficial effects brought about by the technical solution provided by the embodiment of the present invention include at least:

[0024] In the present invention, an intelligent early warning system for equipment failures is constructed, which includes equipment terminals, multimodal data acquisition equipment, edge terminals, cloud terminals and user terminals. The computing power difference between edge terminals and cloud terminals is fully utilized, and the powerful computing power of the cloud terminal is used to update the abnormal data recognition model. Then, the edge terminal uses limited computing power to support the updated abnormal data recognition model to perform abnormal detection on the equipment terminal, so as to achieve fast, accurate and real-time abnormal detection. In the fault warning process, the regularized covariance matrix is ​​first determined based on the historical operation data of the equipment, and the regularization parameter is introduced to effectively avoid the irreversible state of the matrix caused by the number of samples being less than the number of dimensions, thereby ensuring the reversibility of the covariance matrix, so that when performing abnormal detection, the degree of deviation of each sample point relative to the data as a whole can be accurately measured, and an abnormal data identifier is established based on the matrix to determine whether each sample data with high-dimensional information is abnormal in a coarse-grained manner. In this process, a comprehensive judgment is made from the perspective of all dimensions whether the equipment abnormality occurs at each moment, which can effectively discover the linear and nonlinear correlations between parameters and avoid the single-dimensional data judgment-induced In order to solve the problem of low accuracy of monitoring results, after determining the abnormal state, the autoencoder is used to determine the reconstruction error of the operating data of each component of the equipment, and then more computing resources are gathered to accurately locate the faulty components in a fine-grained manner. Without wasting computing resources, it can complete the real-time response to the fault state with higher accuracy. Moreover, the model relies on the powerful computing power of the cloud, and uses the collected operating data as historical data to continuously update the abnormal data recognition model, continuously improve the abnormal state recognition ability, and thus form a virtuous circle, continuously improve the equipment's abnormality detection accuracy, reduce equipment downtime and maintenance times, reduce equipment operation and maintenance costs, and improve equipment operation efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0026] Figure 1 A schematic diagram of the structure of an intelligent early warning system for equipment failure based on the Internet of Things provided by an embodiment of the present invention;

[0027] Figure 2 A schematic diagram of a flow chart of an intelligent early warning method for equipment failure based on the Internet of Things provided by an embodiment of the present invention;

[0028] Figure 3A schematic diagram of the structure of an abnormal data recognition model provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0029] The technical solution of the present invention is described below in conjunction with the accompanying drawings.

[0030] In the embodiments of the present invention, words such as "exemplarily" and "for example" are used to indicate examples, illustrations or explanations. Any embodiment or design described as "example" in the present invention should not be interpreted as being more preferred or more advantageous than other embodiments or designs. Specifically, the use of the word "example" is intended to present the concept in a specific way. In addition, in the embodiments of the present invention, the meaning expressed by "and / or" can be both, or it can be either of the two.

[0031] In the embodiments of the present invention, "image" and "picture" can sometimes be used interchangeably. It should be noted that when the difference between them is not emphasized, the meanings they intend to express are the same. "of", "corresponding, relevant" and "corresponding" can sometimes be used interchangeably. It should be noted that when the difference between them is not emphasized, the meanings they intend to express are the same.

[0032] In the embodiments of the present invention, sometimes a subscript such as W1 may be mistakenly written as a non-subscript such as W1. When the difference is not emphasized, the meanings to be expressed are consistent.

[0033] In order to make the technical problems, technical solutions and advantages to be solved by the present invention more clear, a detailed description will be given below with reference to the accompanying drawings and specific embodiments.

[0034] Reference Manual Attached Figure 1 , showing a structural schematic diagram of an intelligent early warning system for equipment failure based on the Internet of Things provided by an embodiment of the present invention.

[0035] The embodiment of the present invention provides an intelligent early warning system for equipment failure based on the Internet of Things, and the system includes: an equipment terminal, a multimodal data acquisition device, an edge terminal, a cloud terminal and a user terminal.

[0036] The device terminal is connected to the cloud through the multimodal data acquisition device and the edge terminal in turn;

[0037] The device terminal is connected to the cloud through a multimodal data acquisition device, which is used to collect real-time data from the device terminal and transmit the real-time data to the cloud;

[0038] Both the edge and the cloud are connected to the user terminal, and the user terminal is notified when a device failure occurs;

[0039] The user terminal is connected to the device terminal so as to control the start and stop of the device terminal through the user terminal.

[0040] Among them, device terminal: the actual running device, which collects various operating status data through the sensors installed on it. Multimodal data acquisition equipment: connected to the device terminal, collects multi-dimensional data of the device (such as temperature, pressure, vibration, etc.) in real time, and transmits the data to the cloud for further processing and analysis. Edge: a data processing unit located close to the device, which can pre-process and analyze the data locally, reduce data transmission latency, and identify some preliminary abnormalities locally. Cloud: stores and analyzes data collected from device terminals and edge terminals, uses more complex algorithms for in-depth data analysis, and transmits the analysis results back to the system. User terminal: connects the device terminal, edge terminal and cloud terminal. Users can receive fault notifications, view device status and control the start and stop operations of the equipment through the user terminal.

[0041] It should be noted that this system architecture realizes real-time monitoring and intelligent early warning of equipment through the Internet of Things technology, and uses edge computing to reduce the delay of data transmission and improve the efficiency of fault detection. At the same time, users can remotely control the equipment through the terminal, optimize equipment management, and reduce downtime and maintenance costs caused by faults.

[0042] In a possible implementation manner, all connection modes are connection modes based on a heartbeat mechanism.

[0043] Among them, the heartbeat mechanism is a communication mechanism used to maintain the stability of network connections and monitor the connection status. Specifically, two systems (such as device terminals and servers) periodically send heartbeat packets to confirm whether the network connection is normal and whether the system is online. If no heartbeat packet or response is received within a period of time, the system will consider the connection to be interrupted or the device to be offline, and take corresponding processing measures. Using a connection method based on the heartbeat mechanism can enhance the robustness of the system, promptly detect and handle potential faults in devices and networks, and ensure the high availability and stability of the system.

[0044] In actual application, the system realizes real-time monitoring and intelligent early warning of equipment through Internet of Things technology, and uses edge computing to reduce data transmission delay and improve fault detection efficiency. At the same time, users can remotely manage equipment through terminals, prevent faults in time, reduce downtime and maintenance costs, and ensure efficient and stable operation of the system.

[0045] The beneficial effects brought about by the technical solution provided by the embodiment of the present invention include at least:

[0046] In the present invention, an intelligent early warning system for equipment failures is constructed, which includes equipment terminals, multimodal data acquisition equipment, edge terminals, cloud terminals and user terminals. The computing power difference between edge terminals and cloud terminals is fully utilized, and the powerful computing power of the cloud terminal is used to update the abnormal data recognition model. Then, the edge terminal uses limited computing power to support the updated abnormal data recognition model to perform abnormal detection on the equipment terminal, so as to achieve fast, accurate and real-time abnormal detection. In the fault warning process, the regularized covariance matrix is ​​first determined based on the historical operation data of the equipment, and the regularization parameter is introduced to effectively avoid the irreversible state of the matrix caused by the number of samples being less than the number of dimensions, thereby ensuring the reversibility of the covariance matrix, so that when performing abnormal detection, the degree of deviation of each sample point relative to the data as a whole can be accurately measured, and an abnormal data identifier is established based on the matrix to determine whether each sample data with high-dimensional information is abnormal in a coarse-grained manner. In this process, a comprehensive judgment is made from the perspective of all dimensions whether the equipment abnormality occurs at each moment, which can effectively discover the linear and nonlinear correlations between parameters and avoid the single-dimensional data judgment-induced In order to solve the problem of low accuracy of monitoring results, after determining the abnormal state, the autoencoder is used to determine the reconstruction error of the operating data of each component of the equipment, and then more computing resources are gathered to accurately locate the faulty components in a fine-grained manner. Without wasting computing resources, it can complete the real-time response to the fault state with higher accuracy. Moreover, the model relies on the powerful computing power of the cloud, and uses the collected operating data as historical data to continuously update the abnormal data recognition model, continuously improve the abnormal state recognition ability, and thus form a virtuous circle, continuously improve the equipment's abnormality detection accuracy, reduce equipment downtime and maintenance times, reduce equipment operation and maintenance costs, and improve equipment operation efficiency.

[0047] Reference Manual Attached Figure 2 , which shows a flow chart of an intelligent early warning method for equipment failure based on the Internet of Things provided by an embodiment of the present invention.

[0048] The present invention also provides an intelligent early warning method for equipment failure based on the Internet of Things, which is applied to the above-mentioned intelligent early warning system for equipment failure based on the Internet of Things, and the method comprises:

[0049] S1: Obtain historical operation data of the device terminal under normal operating conditions through the cloud.

[0050] Among them, the historical operation data includes multiple multi-dimensional sample points.

[0051] Each multi-dimensional sample point is the operating data of all components of the equipment at a moment.

[0052] It should be noted that the historical operating data of the device terminal under normal conditions is obtained through the cloud. This data contains the operating status information of all components of the device at different time points, which in turn provides reliable benchmark data for subsequent anomaly detection, helps identify the normal operating mode of the equipment, more accurately identifies equipment anomalies, and improves the accuracy and timeliness of fault warnings.

[0053] S2: Regularize the covariance matrix with the minimized covariance determinant and construct a regularized covariance matrix.

[0054] Among them, the covariance determinant is the determinant of the covariance matrix, which refers to the calculation of the determinant of the covariance matrix. Regularization is a technique to prevent model overfitting, especially when dealing with high-dimensional data. It limits the complexity of the model by introducing some constraints or penalty terms, thereby improving the generalization ability of the model. The covariance matrix is ​​a matrix used to describe the correlation between different dimensions in a multidimensional data set. A stable regularized covariance matrix is ​​constructed by regularizing the covariance matrix that minimizes the covariance determinant. The advantage of this is that it prevents the covariance matrix from being singular (irreversible) in high-dimensional small sample scenarios, ensuring the robustness and accuracy of anomaly detection, and is particularly suitable for complex data sets with dimensions greater than the number of samples.

[0055] In actual use, due to the different acquisition frequencies and the different complexities of different devices, fat data, also known as high-dimensional small samples, will inevitably appear. That is, the number of samples collected is very close to the data dimension or less than the data dimension. At this time, the covariance matrix of the minimum regularized covariance determinant will be in a singular state, that is, an irreversible state. At this time, the number of dimensions of the data is greater than the number of samples, and the samples are not sufficient to fully represent the relationship between all dimensions, resulting in linear dependence between certain dimensions, making the determinant of the covariance matrix zero, and thus unable to generate a stable covariance matrix. The reversibility of the covariance matrix ensures that when performing anomaly detection, the degree of deviation of each sample point relative to the data as a whole can be accurately measured. The regularized covariance matrix introduces a regularization parameter to control the weights of the unit matrix I and the sample covariance matrix. In high-dimensional scenarios, it prevents the covariance matrix from being singular (irreversible) and ensures the stability of the calculation.

[0056] In a possible implementation, the regularized covariance matrix is ​​specifically:

[0057] K(H)=QΛ 1 / 2 [ρI+(1-ρ)c α S W (H)]Λ 1 / 2 Q T

[0058] S W (H) = Λ -1 / 2 QT S U (H)QΛ -1 / 2

[0059] S U (H) = (h-1) -1 (U H -m U (H) T (U H -m U (H)

[0060]

[0061] Where K(H) represents the regularized covariance matrix with respect to the optimal subset H, Q represents the eigenvector matrix of the regularized covariance matrix, Λ represents the eigenvalue matrix of the regularized covariance matrix, ρ represents the regularization parameter, I represents the identity matrix, S W (H) represents the intermediate variable, S U (H) represents the covariance matrix calculated from H selected from the sample set U, c α Indicates adjustment S W (H) correction factor, the subscript T indicates transposition, m U (H) represents the mean vector of H in U, h represents the number of samples selected from H, 1 h represents a vector of length h that is all 1, U H Represents a sample matrix of H of size h×p, where p represents the total dimension of the data, and the optimal subset is the subset that minimizes the regularized covariance matrix.

[0062] Specifically, K(H) represents the regularized covariance matrix, which is used to evaluate the covariance structure of the sample set after regularization. Q is an orthogonal matrix whose column vectors are the eigenvectors of K(H). These eigenvectors correspond to different directions of the data, indicating the direction of data dispersion in the direction of each eigenvector. Λ is a diagonal matrix containing the eigenvalues ​​of the covariance matrix K(H), which represents the variance (dispersion) of the data in the direction of the eigenvector. The larger the eigenvalue, the greater the dispersion of the data in the corresponding eigenvector direction. Λ 1 / 2It is the square root of the eigenvalue matrix, that is, the square root of each eigenvalue in the diagonal matrix is ​​taken, which is used for the normalization of the covariance matrix, so that the matrix can adjust its influence through the square root of the eigenvalue. The regularization parameter controls the weight of the unit matrix I and the sample covariance matrix. In high-dimensional scenarios, it prevents the covariance matrix from being singular (irreversible) and ensures the stability of the calculation. The unit matrix represents the basic structure of regularization. The introduction of the unit matrix can prevent the eigenvalue from being too small (close to zero), thereby ensuring the reversibility of the covariance matrix. The correction factor is used to adjust the estimate of the sample covariance matrix, especially when there are finite samples, it can make the estimation result more accurate. S W (H) The sample covariance matrix based on the optimal subset H represents the covariance matrix of the sample points calculated from the subset. This matrix reflects the data distribution on the currently selected subset. The transposed matrix of Q ensures that the structure of the final matrix K(H) is consistent with the dimension and direction of the original data.

[0063] Among them, the regularized covariance matrix is ​​a decomposed matrix, which converts the problem of solving the regularized covariance matrix into the problem of finding eigenvalues ​​and eigenvectors, making these complex matrix operations more efficient and stable, and simplifying the problem that when the data dimension is large, directly calculating the determinant, inverse matrix and other operations of the covariance matrix will be very complicated.

[0064] Refer to the instruction manual Figure 3 , shows a structural schematic diagram of an abnormal data identification model provided by an embodiment of the present invention.

[0065] S3: Use historical operation data to build an abnormal data identification model.

[0066] Among them, the abnormal data recognition model includes an input module, an abnormal data identifier based on a regularized covariance matrix, an autoencoder and an output module connected in sequence.

[0067] Among them, the autoencoder is an unsupervised neural network model, which is usually used for data dimension reduction, feature learning and anomaly detection. It consists of an encoder and a decoder: Encoder: compresses high-dimensional data into low-dimensional implicit representation (feature extraction). Decoder: reconstructs high-dimensional original data based on low-dimensional implicit representation. In the anomaly detection task, the autoencoder uses the normal data pattern it has learned for reconstruction. If the model can reconstruct the input data well, it means that the data conforms to the normal pattern. If the reconstruction error is large, it means that the data may be abnormal. The abnormal data identifier based on the regularized covariance matrix is ​​mainly used to globally detect the deviation of data points from the normal distribution, which is suitable for coarse-grained detection, while the autoencoder performs fine-grained detection dimension by dimension through reconstruction error to accurately locate anomalies. Combining the two, the system can perform accurate anomaly detection on the basis of high efficiency and low computational consumption, adapt to high-dimensional data scenarios, and provide timely fault warnings.

[0068] Specifically, the multi-dimensional sample points are the combined vectors of all dimensions of the equipment that need to be monitored. The sample points are first detected for anomalies through the regularized covariance matrix with the minimum regularized covariance determinant, and the abnormal sample points are found as a whole. Then, the autoencoder is used to calculate the data of each dimension for separate data reconstruction, and the abnormal equipment is accurately locked according to the reconstruction error. This method of performing coarse-grained detection first and then fine-grained detection can greatly reduce the consumption of computing resources to achieve real-time equipment anomaly monitoring. This two-step equipment anomaly detection method combines the regularized covariance matrix and the autoencoder, and can effectively realize real-time anomaly monitoring of multi-dimensional data of the equipment. The core idea of ​​this method is to achieve efficient and accurate anomaly identification while reducing computing resource consumption through a combination of coarse-grained and fine-grained detection.

[0069] In a possible implementation, the construction process of the abnormal data identifier is specifically as follows:

[0070] Standardize historical operating data.

[0071] With the goal of minimizing the determinant of the regularized covariance matrix, a subset of a preset proportion is selected from the standardized historical operating data as the optimal subset.

[0072] The process of selecting the optimal subset is as follows: from the sample data, i.e., the historical running data after the table conversion, select data points Q that account for a preset proportion of the total number of samples as the initial subset to ensure robustness. For this initial subset, calculate its mean and covariance matrix, and then perform iterative optimization. The iterative process is as follows: calculate the Mahalanobis distance from all data points to the current covariance matrix, select Q data points with the smallest distance, update the subset, and repeat this process until the determinant of the covariance matrix no longer decreases significantly. The initial subset at this time is taken as the optimal subset.

[0073] It should be noted that those skilled in the art can set the size of the preset ratio according to actual needs, and the present invention is not limited here. Specifically, the preset ratio can be set to 0.75.

[0074] Determine the regularized covariance matrix under the optimal subset.

[0075] The regularization parameter of the regularized covariance matrix is ​​determined based on the regularized covariance matrix condition number.

[0076] In a possible implementation, the regularization parameter is determined as follows:

[0077]

[0078] CN target =10·log(p)

[0079] Among them, λ max and λ min denote the maximum and minimum eigenvalues ​​of the regularized covariance matrix, respectively, and CN target represents the regularized covariance matrix condition number, p represents the total dimension of historical operation data, and log represents the logarithmic function.

[0080] Among them, the regularization parameter is selected based on the condition number of the data. The purpose is to ensure the reversibility of the covariance matrix, especially in high-dimensional data, to prevent the matrix from becoming singular (i.e., irreversible). The condition number is an important indicator for measuring the stability of the matrix. It represents the ratio of the maximum and minimum eigenvalues ​​in the matrix. A larger condition number indicates that the matrix is ​​close to being singular and numerically unstable. The goal of selecting the regularization parameter is to control the condition number of the regularized matrix within an acceptable range.

[0081] The calculation formula of the regularized covariance matrix condition number is a target condition number that grows with the dimension, so that the target condition number can effectively control the reversibility of the matrix in high-dimensional data. If the data dimension is high, this method will automatically give a higher target condition number, improve the application scenario of anomaly recognition, and complete adaptive and accurate anomaly data recognition.

[0082] Substitute the regularization parameters into the regularization covariance matrix to obtain the optimal regularization covariance matrix.

[0083] Calculate the mean vector of the standardized historical running data.

[0084] The optimal regularized covariance matrix is ​​combined with the mean vector to obtain an abnormal data identifier.

[0085] For example, the process of the anomaly identifier to identify abnormal data is as follows: First, each monitoring parameter is standardized. 450 samples are selected from 600 samples as a subset, and their covariance matrix is ​​calculated and their determinant is minimized to obtain a robust mean and covariance matrix. Then, based on the robust mean and covariance matrix, the Mahalanobis distance of the newly acquired equipment operation sample points is calculated. Then a threshold is set, and samples exceeding this distance are marked as abnormal points. Finally, possible outliers are marked and further analyzed to see whether these abnormal points represent equipment failures or other problems.

[0086] It should be noted that the abnormal data identifier established in this way ensures its reversibility and stability in high-dimensional data by minimizing the determinant of the regularized covariance matrix, and combines the robust mean and Mahalanobis distance to improve the accuracy and robustness of anomaly detection, and can more accurately identify abnormal points of the equipment.

[0087] In a possible implementation, the autoencoder includes an encoder and a decoder, and the construction process of the autoencoder is specifically as follows:

[0088] Cluster the historical operation data according to the components to which they belong, and obtain multi-dimensional operation data of different equipment components:

[0089] X={x1,x2,...,x n}

[0090] x i =[x i1 ,x i2 ,...,x im ],i=1,2,…,n

[0091] Among them, x i represents the i-th sample of the multidimensional running data, and n represents the number of samples x of the multidimensional running data X. ij Represents the data of the jth dimension in the i-th sample in the multidimensional data, j=1, 2, ..., m, where m represents the total dimension of the multidimensional running data.

[0092] Each sample is input into the encoder to generate the sample implicit representation:

[0093] z i =f encoder (x i )

[0094] Among them, f encoder () represents the encoder, z i Represents x i The sample implicit representation of .

[0095] Input the sample implicit representation into the decoder to obtain the reconstructed original data:

[0096]

[0097] in, Represents the reconstruction of the original data, f decoder () indicates a decoder.

[0098] Calculate the mean square reconstruction error between the original data and the reconstructed original data:

[0099]

[0100] Where L represents the mean square reconstruction error.

[0101] The autoencoder is trained with the goal of minimizing the mean square reconstruction error.

[0102] Output the trained autoencoder.

[0103] It should be noted that the autoencoder established in this way can learn the normal operation mode of equipment components by compressing and reconstructing historical operation data, and accurately detect anomalies by minimizing reconstruction errors. Its advantages are that it can automatically extract features, process complex multi-dimensional data, accurately identify equipment failures, and improve the accuracy and efficiency of fault detection.

[0104] S4: Send the abnormal data recognition model to the edge.

[0105] It should be noted that sending the abnormal data identification model to the edge can utilize the real-time and localized processing capabilities of edge computing, reduce data transmission delays and bandwidth occupancy, improve the response speed and system efficiency of anomaly detection, and ensure rapid warning and timely handling of equipment failures.

[0106] S5: Collect real-time operation data of the device terminal through a multimodal data acquisition device.

[0107] S6: Analyze the real-time operation data through the abnormal data recognition model to determine whether there is abnormal operation data. If so, output the abnormal component corresponding to the abnormal operation data and go to step S8. Otherwise, go to step S7.

[0108] In a possible implementation, the real-time operation data is analyzed by using an abnormal data recognition model to determine whether there is abnormal operation data, specifically including:

[0109] Normalize real-time operational data.

[0110] The standardized real-time operation data is substituted into the abnormal data identifier, and the Mahalanobis distance between the data point corresponding to the standardized real-time operation data and the historical operation data represented by the abnormal data identifier is calculated.

[0111] In a possible implementation, the Mahalanobis distance is calculated as follows:

[0112]

[0113] Among them, D M (x new ) represents the real-time running data x calculated by the abnormal data identifier new The Mahalanobis distance of , μ represents the mean vector in the abnormal data identifier.

[0114] Among them, Mahalanobis distance is a method to measure the distance between multidimensional data points. It not only takes into account the differences between dimensions, but also the correlation between dimensions. Mahalanobis distance can measure the degree to which sample points deviate from the center of the overall distribution based on the covariance structure of the data. It is scaled by the covariance matrix and is particularly suitable for detecting multidimensional abnormal data. The Mahalanobis distance can accurately detect abnormal points in multidimensional data, especially when there is correlation between features. Its advantages are more obvious and it is suitable for use in complex multidimensional data scenarios.

[0115] When the Mahalanobis distance is greater than the preset Mahalanobis distance, it is determined that the real-time operation data is abnormal.

[0116] The real-time operation data is substituted into the autoencoder according to the component category for reconstruction to obtain the reconstructed real-time operation data.

[0117] A reconstruction error between the reconstructed real-time operation data and the real-time operation data of the corresponding component is calculated.

[0118] In a possible implementation manner, the reconstruction error is calculated as follows:

[0119]

[0120] in, and denote the real-time operation data of the vth component and the reconstructed real-time operation data of the vth component, respectively. e represents the reconstruction error of the vth component.

[0121] Among them, reconstruction error refers to the difference between the original data and the data reconstructed by the model when using models such as autoencoders to reconstruct data. The advantage of detecting anomalies through reconstruction error is that the autoencoder can automatically learn the normal operating mode of the equipment, use the reconstruction error to measure the degree of deviation between the real-time data and the normal mode, and can accurately identify the anomalies of specific components, which is particularly suitable for processing complex and multi-dimensional data scenarios.

[0122] When the reconstruction error is greater than the reconstruction error of the autoencoder, it is determined that the component corresponding to the reconstruction error is abnormal.

[0123] The reconstruction error of the autoencoder refers to the reconstruction error of the autoencoder obtained during the training process with the goal of minimizing the mean square reconstruction error, that is, the minimized mean square reconstruction error.

[0124] When the Mahalanobis distance is less than the preset Mahalanobis distance, it is determined that the real-time operation data is normal.

[0125] It should be noted that those skilled in the art can set the size of the preset Mahalanobis distance according to actual needs, and the present invention is not limited here. Specifically, the size of the preset Mahalanobis distance can be determined using the chi-square distribution. The specific process is to first determine the dimension of the real-time running data, refer to the chi-square distribution table according to the dimension to determine the significance level, and then calculate the size relationship between the square of the Mahalanobis distance and the quantile of the chi-square distribution corresponding to the significance level. If the square of the Mahalanobis distance is greater than the quantile, it means that the real-time running data is abnormal, otherwise it is normal.

[0126] It should be noted that the advantage of judging whether there is abnormal data in this way is that, by combining the Mahalanobis distance and the reconstruction error of the autoencoder, it can not only globally screen out abnormal data points, but also determine the specific abnormal components through fine-grained analysis, thereby improving the accuracy and reliability of anomaly detection and adapting to complex multi-dimensional data scenarios.

[0127] S7: Upload the real-time operation data to the historical operation data in the cloud to update the historical operation data, and return to step S3.

[0128] In this way, the model has continuous learning capabilities, which can continuously improve the abnormal data recognition capabilities of the abnormal data recognition model. The powerful learning capabilities of the cloud can be used to quickly build a more accurate abnormal data recognition model, which is then sent to the edge with certain data processing capabilities. The trained abnormal data model is used to perform anomaly detection on real-time operating data, thus forming a virtuous circle, continuously expanding the normal data boundaries, and more accurately identifying abnormal data of device operation.

[0129] S8: Send the abnormal component to the user terminal for abnormal reminder.

[0130] In actual application, the solution combines the regularized covariance matrix and the autoencoder, screens global abnormal data through the Mahalanobis distance, and uses the reconstruction error to accurately locate specific abnormal components to ensure the efficiency and accuracy of abnormality detection. This method is suitable for high-dimensional complex data scenarios, uses cloud and edge computing to achieve real-time detection, and has continuous learning capabilities, which can dynamically update the model and improve the accuracy and reliability of detection.

[0131] The beneficial effects brought about by the technical solution provided by the embodiment of the present invention include at least:

[0132] In the present invention, an intelligent early warning system for equipment failures is constructed, which includes equipment terminals, multimodal data acquisition equipment, edge terminals, cloud terminals and user terminals. The computing power difference between edge terminals and cloud terminals is fully utilized, and the powerful computing power of the cloud terminal is used to update the abnormal data recognition model. Then, the edge terminal uses limited computing power to support the updated abnormal data recognition model to perform abnormal detection on the equipment terminal, so as to achieve fast, accurate and real-time abnormal detection. In the fault warning process, the regularized covariance matrix is ​​first determined based on the historical operation data of the equipment, and the regularization parameter is introduced to effectively avoid the irreversible state of the matrix caused by the number of samples being less than the number of dimensions, thereby ensuring the reversibility of the covariance matrix, so that when performing abnormal detection, the degree of deviation of each sample point relative to the data as a whole can be accurately measured, and an abnormal data identifier is established based on the matrix to determine whether each sample data with high-dimensional information is abnormal in a coarse-grained manner. In this process, a comprehensive judgment is made from the perspective of all dimensions whether the equipment abnormality occurs at each moment, which can effectively discover the linear and nonlinear correlations between parameters and avoid the single-dimensional data judgment-induced In order to solve the problem of low accuracy of monitoring results, after determining the abnormal state, the autoencoder is used to determine the reconstruction error of the operating data of each component of the equipment, and then more computing resources are gathered to accurately locate the faulty components in a fine-grained manner. Without wasting computing resources, it can complete the real-time response to the fault state with higher accuracy. Moreover, the model relies on the powerful computing power of the cloud, and uses the collected operating data as historical data to continuously update the abnormal data recognition model, continuously improve the abnormal state recognition ability, and thus form a virtuous circle, continuously improve the equipment's abnormality detection accuracy, reduce equipment downtime and maintenance times, reduce equipment operation and maintenance costs, and improve equipment operation efficiency.

[0133] It should be understood that the processor in the embodiment of the present invention may be a central processing unit (CPU), and the processor may also be other general-purpose processors, digital signal processors (DSP), application specific integrated circuits (ASIC), field programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

[0134] It should also be understood that the memory in the embodiments of the present invention may be a volatile memory or a non-volatile memory, or may include both volatile and non-volatile memories. Among them, the non-volatile memory may be a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), or a flash memory. The volatile memory may be a random access memory (RAM), which is used as an external cache. By way of example and not limitation, many forms of random access memory (RAM) are available, such as static RAM (SRAM), dynamic random access memory (DRAM), synchronous DRAM (SDRAM), double data rate synchronous dynamic random access memory (DDR SDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchronous link DRAM (SLDRAM), and direct RAM bus random access memory (DR RAM).

[0135] The above embodiments can be implemented in whole or in part by software, hardware (such as circuits), firmware or any other combination. When implemented by software, the above embodiments can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the process or function described in the embodiment of the present invention is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions can be transmitted from one website, computer, server or data center to another website, computer, server or data center by wired (such as infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage system such as a server, data center, etc. that contains one or more available media sets. The available medium can be a magnetic medium (for example, a floppy disk, a hard disk, a tape), an optical medium (for example, a DVD), or a semiconductor medium. The semiconductor medium can be a solid-state hard disk.

[0136] It should be understood that the term "and / or" in this article is only a description of the association relationship of associated objects, indicating that there can be three relationships. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. A and B can be singular or plural. In addition, the character " / " in this article generally indicates that the associated objects before and after are in an "or" relationship, but it may also indicate an "and / or" relationship. Please refer to the context for specific understanding.

[0137] In the present invention, "at least one" means one or more, and "more than one" means two or more. "At least one of the following" or similar expressions refers to any combination of these items, including any combination of single or plural items. For example, at least one of a, b, or c can be represented by: a, b, c, ab, ac, bc, or abc, where a, b, c can be single or multiple.

[0138] It should be understood that in various embodiments of the present invention, the size of the serial numbers of the above-mentioned processes does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0139] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0140] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0141] In the several embodiments provided by the present invention, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0142] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0143] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0144] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for enabling a computer system (which can be a personal computer, server, or network system, etc.) to perform all or part of the steps of the methods described in various embodiments of the present invention. The aforementioned storage media include: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk, and other media that can store program codes.

[0145] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

[0146] There are a few points to note:

[0147] (1) The drawings of the embodiments of the present invention only relate to the structures related to the embodiments of the present invention, and other structures may refer to the general design.

[0148] (2) For the sake of clarity, in the drawings used to describe the embodiments of the present invention, the thickness of the layers or regions is exaggerated or reduced, that is, these drawings are not drawn according to the actual scale. It is understood that when an element such as a layer, film, region or substrate is referred to as being "on" or "under" another element, the element may be "directly" "on" or "under" the other element or there may be intermediate elements.

[0149] (3) In the absence of conflict, the embodiments of the present invention and the features therein may be combined with each other to obtain new embodiments.

[0150] The above are only specific embodiments of the present invention, but the protection scope of the present invention is not limited thereto. The protection scope of the present invention shall be based on the protection scope of the claims.

Claims

1. An intelligent early warning system for equipment failure based on the Internet of Things, characterized in that: The system includes device terminals, multimodal data acquisition equipment, edge terminals, cloud terminals, and user terminals; The device terminal is connected to the cloud via the multimodal data acquisition device and the edge terminal in sequence; The device terminal is connected to the cloud through the multimodal data acquisition device, and the multimodal data acquisition device is used to collect real-time data of the device terminal and transmit the real-time data to the cloud; The edge terminal and the cloud are both connected to the user terminal, and the user terminal is notified when a device failure occurs; The user terminal is connected to the device terminal so as to control the start and stop of the device terminal through the user terminal.

2. The intelligent early warning system for equipment failure based on the Internet of Things according to claim 1 is characterized in that: The connection methods are all based on the heartbeat mechanism.

3. An intelligent early warning method for equipment failure based on the Internet of Things, characterized in that: The intelligent early warning system for equipment failure applied to any one of claims 1 to 2 comprises: S1: Obtaining historical operation data of the device terminal in a normal operating state through the cloud, wherein the historical operation data includes a plurality of multi-dimensional sample points; S2: Regularize the covariance matrix with the minimized covariance determinant and construct a regularized covariance matrix; S3: constructing an abnormal data recognition model using the historical operation data, wherein the abnormal data recognition model includes an input module, an abnormal data identifier based on the regularized covariance matrix, an autoencoder, and an output module connected in sequence; S4: Sending the abnormal data recognition model to the edge end; S5: collecting real-time operation data of the device terminal through the multimodal data collection device; S6: parsing the real-time operation data through the abnormal data recognition model to determine whether there is abnormal operation data. If so, outputting the abnormal component corresponding to the abnormal operation data and proceeding to step S8; otherwise, proceeding to step S7; S7: Upload the real-time operation data to the historical operation data in the cloud to update the historical operation data, and return to step S3; S8: Send the abnormal component to the user terminal for abnormal reminder.

4. According to claim 3, the intelligent early warning method for equipment failure based on the Internet of Things is characterized in that: The regularized covariance matrix is ​​specifically: K(H)=QΛ 1 / 2 [ρI+(1-ρ)c α S W (H)]Λ 1 / 2 Q T S W (H)=Λ -1 / 2 Q T S U (H)QL -1 / 2 S U (H)=(h-1) -1 (U H -m U (H)) T (U H -m U (H)) Where K(H) represents the regularized covariance matrix with respect to the optimal subset H, Q represents the eigenvector matrix of the regularized covariance matrix, Λ represents the eigenvalue matrix of the regularized covariance matrix, ρ represents the regularization parameter, I represents the identity matrix, S W (H) represents the intermediate variable, S U (H) represents the covariance matrix calculated from H selected from the sample set U, c α Indicates adjustment S W (H) correction factor, the subscript T indicates transposition, m U (H) represents the mean vector of H in U, h represents the number of samples selected from H, 1 h represents a vector of length h that is all 1, U H Represents a sample matrix of H of size h×p, where p represents the total dimension of the data, and the optimal subset is the subset that minimizes the regularized covariance matrix.

5. The intelligent early warning method for equipment failure based on the Internet of Things according to claim 4 is characterized in that: The construction process of the abnormal data identifier is specifically as follows: Standardizing the historical operation data; With the goal of minimizing the determinant of the regularized covariance matrix, a subset of a preset proportion is selected from the standardized historical operation data as the optimal subset; Determining a regularized covariance matrix under the optimal subset; Determining a regularization parameter of the regularized covariance matrix according to the regularized covariance matrix condition number; Substituting the regularization parameter into the regularization covariance matrix to obtain an optimal regularization covariance matrix; Calculate the mean vector of the standardized historical running data; The optimal regularized covariance matrix is ​​combined with the mean vector to obtain the abnormal data identifier.

6. According to claim 5, the intelligent early warning method for equipment failure based on the Internet of Things is characterized in that: The regularization parameter is determined in the following way: CN target =10 log(p) Among them, λ max and λ min denote the maximum and minimum eigenvalues ​​of the regularized covariance matrix, respectively, and CN target represents the regularized covariance matrix condition number, p represents the total dimension of historical operation data, and log represents the logarithmic function.

7. According to claim 3, the intelligent early warning method for equipment failure based on the Internet of Things is characterized in that: The autoencoder includes an encoder and a decoder, and the construction process of the autoencoder is specifically as follows: The historical operation data are clustered according to the components to which they belong, to obtain multi-dimensional operation data of different equipment components: X={x1,x2,...,x n } x i =[x i1 ,x i2 ,...,x im ],i=1,2,…,n Among them, x i represents the i-th sample of the multidimensional running data, and n represents the number of samples x of the multidimensional running data X. ij represents the data of the jth dimension in the i-th sample in the multi-dimensional data, j = 1, 2, ..., m, where m represents the total dimension of the multi-dimensional running data; Each sample is input into the encoder to generate the sample implicit representation: z i =f encoder (x i ) Among them, f encoder () represents the encoder, z i Represents x i The sample implicit representation of ; The sample implicit representation is input to the decoder to obtain the reconstructed original data: in, Represents the reconstruction of the original data, f decoder () indicates a decoder; Calculate the mean square reconstruction error between the original data and the reconstructed original data: Where L represents the mean square reconstruction error; Training the autoencoder with the goal of minimizing the mean square reconstruction error; Output the trained autoencoder.

8. The intelligent early warning method for equipment failure based on the Internet of Things according to claim 3 is characterized in that: The performing data analysis on the real-time operation data by using the abnormal data recognition model to determine whether there is abnormal operation data specifically includes: Standardizing the real-time operation data; Substituting the normalized real-time operation data into the abnormal data identifier, and calculating the Mahalanobis distance between the data point corresponding to the normalized real-time operation data and the historical operation data represented by the abnormal data identifier; When the Mahalanobis distance is greater than a preset Mahalanobis distance, determining that the real-time operation data is abnormal; Substituting the real-time operation data into the autoencoder according to the component category for reconstruction to obtain reconstructed real-time operation data; Calculating a reconstruction error between the reconstructed real-time operation data and the real-time operation data of the corresponding component; When the reconstruction error is greater than the reconstruction error of the autoencoder, determining that a component corresponding to the reconstruction error is abnormal; When the Mahalanobis distance is less than a preset Mahalanobis distance, it is determined that the real-time operation data is normal.

9. The intelligent early warning method for equipment failure based on the Internet of Things according to claim 8 is characterized in that: The Mahalanobis distance is calculated as follows: Among them, D M (x new ) represents the real-time running data x calculated by the abnormal data identifier new The Mahalanobis distance of , μ represents the mean vector in the abnormal data identifier.

10. The intelligent early warning method for equipment failure based on the Internet of Things according to claim 8 is characterized in that: The calculation method of the reconstruction error is specifically as follows: in, and denote the real-time operation data of the vth component and the reconstructed real-time operation data of the vth component, respectively. e represents the reconstruction error of the vth component.

Images