A method and system for analyzing electromechanical equipment health monitoring data
By using the timing data abnormality detection model and prediction model trained by the sample balance database in the railway station electromechanical equipment monitoring system, the abnormal status of the equipment is identified and predicted, and the problem that the existing system cannot be analyzed in depth and early warning is solved, and the electromechanical equipment is realized health monitoring and future status prediction.
Patent Information
- Application Number
- CN202411180388.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-27
- Publication Date
- 2024-11-22
- Estimated Expiration
- 2044-08-27
AI Technical Summary
The existing railway station electromechanical equipment monitoring system cannot deeply analyze the operating status data of each subsystem, lacks the early warning function for equipment health monitoring, and it is difficult to achieve effective health monitoring of electromechanical equipment.
A sample equalization database is used to train multiple abnormal detection models based on time sequence data, and is configured in each device subsystem. By building a data transmission channel, an abnormal detection result is input to the prediction model, an early warning output is generated, and an abnormal scoring function and local adaptive threshold are used. Identify the exception sequence and predict the future status of the device through prediction models with convolutional layer, LSTM layer and other structures.
It improves the accuracy of abnormal data detection, can identify abnormal data in the monitoring system and predict future health trends of the equipment, provides data support for the maintenance of electromechanical equipment, and reduces data transmission pressure.
Smart Images

Figure CN118709118B_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the technical field of data analysis, and in particular, relates to a method and system for analyzing electromechanical equipment health monitoring data. Background Art
[0002] With the rapid development of railway transportation system and the continuous improvement of people's living standards, both staff and passengers have higher and higher requirements for the safety and comfort of transportation, travel, work, rest and other environments. In order to improve the safety level of railway enterprises and the level of modern management, and achieve the goal of energy saving and consumption reduction, the operating status monitoring of various electromechanical equipment in important buildings such as railway stations is also particularly important.
[0003] The station house electromechanical equipment monitoring system aggregates the key equipment of subsystems such as the BA system, intelligent lighting system, energy management system, power supply system, network communication system, and the operating data of each subsystem in the monitoring system for display, making it convenient for management personnel to understand the operating status of the system.
[0004] However, although this solution can summarize the data of each subsystem, it lacks in-depth analysis and processing of the data of each subsystem, and cannot combine the operating status data of each subsystem to conduct health monitoring and early warning of various electromechanical equipment in the railway system station. Summary of the invention
[0005] The purpose of the embodiments of the present invention is to provide a method and system for analyzing electromechanical equipment health monitoring data, aiming to solve the technical problems raised in the above-mentioned background technology.
[0006] To achieve the above object, the present invention provides the following technical solutions.
[0007] In one embodiment of the present invention, a method for analyzing electromechanical equipment health monitoring data is provided, the data analysis method comprising the following steps:
[0008] A sample-balanced database is used to train multiple time series data anomaly detection models, and the multiple time series data anomaly detection models are respectively configured in corresponding device subsystems according to their device identification information;
[0009] Construct multiple data transmission channels between the monitoring main system and multiple equipment subsystems;
[0010] During the maintenance cycle of electromechanical equipment, in response to the abnormal detection results sent to the monitoring main system in the data transmission channel, the abnormal data corresponding to the abnormal detection results are used as the input of the prediction model, and the output of the prediction model is output as an early warning in the monitoring main system together with the abnormal detection results.
[0011] As a further limitation of the solution of the present invention, the step of training multiple time series data anomaly detection models using a sample-balanced database includes:
[0012] Obtain a plurality of sample-balanced databases containing device identification information;
[0013] A sample-balanced database is used to train a time series data anomaly detection model containing device identification information.
[0014] As a further solution of the present invention, the anomaly detection result sent to the monitoring main system in the data transmission channel is outputted through a trained time series data based anomaly detection model, and the time series data based anomaly detection model includes a data processing module and an anomaly detection module;
[0015] In the data processing module, the newly acquired device time series data is preprocessed, and the frequency components and the corresponding number of channels, number of sequences and length of the detection time window are obtained by decomposition to form a time series to be detected;
[0016] In the anomaly detection module, an anomaly scoring function is constructed to represent the difference between the time series to be detected and the normal sequence model, and the anomaly score is calculated using the anomaly scoring function; the abnormal sequence is found using a local adaptive threshold method, wherein the anomaly threshold for each sliding window is defined as 4 standard deviations different from the average value of the anomaly score in the sliding window; the sequence with an anomaly score greater than the anomaly threshold is identified as an anomaly, and a continuous anomaly sequence is obtained.
[0017] As a further solution of the present invention, for each abnormal sequence in the continuous abnormal sequence, the maximum abnormal score of the abnormal sequence is first obtained, and then it is sorted in descending order, and the decline percentage is calculated. When the first decline percentage does not exceed the decline threshold, the remaining sequences in the continuous abnormal sequence are eliminated.
[0018] As a further solution of the present invention, the step of constructing an abnormality scoring function for representing the difference between the time series to be detected and the normal sequence model includes:
[0019] For each time point t, in the sequence to be tested and the expected value of the normal sequence of the anomaly detection model Calculate the residual between ;
[0020] The anomaly scoring function is expressed as: .
[0021] As a further solution of the present invention, in the step of finding abnormal sequences by using a local adaptive threshold, a sliding window W is defined, and for each sliding window position w, the average value of the abnormal scores of all time points in the window is calculated: ;
[0022] Compute the standard deviation within a window: ;
[0023] For each sliding window, the anomaly threshold The anomaly threshold is the average anomaly score plus 4 times the standard deviation. It is expressed as: .
[0024] As a further solution of the present invention, the step of identifying a sequence having an anomaly score greater than an anomaly threshold as an anomaly to obtain a continuous anomaly sequence includes:
[0025] For each time point t, if the anomaly score Greater than the abnormal threshold of the corresponding sliding window , then the sequence at that time point is considered abnormal;
[0026] The sequences corresponding to multiple time points identified as abnormal are sorted to form a continuous abnormal sequence.
[0027] As a further solution of the present invention, the input of the prediction model is an abnormal sequence of abnormal detection results; the output of the prediction model is a state prediction sequence of the electromechanical equipment corresponding to the future time period;
[0028] The structure of the prediction model includes an input layer, a convolutional layer, a pooling layer, an LSTM layer, a fully connected layer and an output layer;
[0029] The convolution layer includes three one-dimensional convolution layers, which are used to process time series data and capture data to obtain local features. Each convolution layer is followed by a RELU activation function for introducing nonlinearity. After each convolution layer, the local features of the input sequence are converted into 32 different feature representations.
[0030] The pooling layer performs a one-dimensional maximum pooling operation, the pooling window size is 2, and the maximum value of every two consecutive time points on the sequence time axis is taken as the output;
[0031] The LSTM layer includes 64 LSTM units;
[0032] The fully connected layer is arranged after the LSTM layer and is used to map the features learned by the LSTM layer to the final output layer.
[0033] In another embodiment of the present invention, a data analysis system for monitoring health of electromechanical equipment is provided, the data analysis system comprising the following steps:
[0034] An anomaly detection module is used to train multiple anomaly detection models based on time series data using a sample-balanced database, and the multiple anomaly detection models based on time series data are respectively configured in corresponding device subsystems according to their device identification information;
[0035] A data transmission module, used to construct multiple data transmission channels between the monitoring main system and multiple device subsystems;
[0036] The analysis and early warning module is used to respond to the abnormal detection results sent to the monitoring main system in the data transmission channel during the maintenance cycle of the electromechanical equipment, use the abnormal data corresponding to the abnormal detection results as the input of the prediction model, and output the output of the prediction model together with the abnormal detection results in the monitoring main system for early warning.
[0037] Compared with the prior art, the electromechanical equipment health monitoring data analysis method and system of the present invention have the following beneficial effects:
[0038] First, the present invention uses a sample-balanced database to train multiple time series data anomaly detection models. Multiple time series data anomaly detection models are respectively configured in corresponding device subsystems according to their device identification information, and independent abnormal data detection is performed in each device subsystem, thereby avoiding the pressure on the data transmission channel caused by excessive data passing through the data transmission channel;
[0039] Second, based on the time series data anomaly detection model, the present invention constructs an anomaly scoring function for representing the difference between the time series to be detected and the normal series model, and calculates the anomaly score using the anomaly scoring function; uses a local adaptive threshold method to find the abnormal sequence, thereby improving the accuracy of abnormal data detection;
[0040] Third, the present invention uses the abnormal data corresponding to the abnormal detection results as the input of the prediction model, and outputs the output of the prediction model together with the abnormal detection results in the monitoring main system for early warning. Not only can the abnormal data of each equipment subsystem be seen in the monitoring system, but also the state prediction results of the equipment corresponding to the abnormal data in a future time period can be intuitively known, providing data support for further processing of the electromechanical equipment; among them, the prediction model provided by the present invention can capture the complex patterns of the health status sequence of the substation equipment, and predict the accurate future health trend of the electromechanical equipment. BRIEF DESCRIPTION OF THE DRAWINGS
[0041] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present invention.
[0042] Figure 1 is a system architecture diagram of the present invention;
[0043] Figure 2 A flowchart of a method for analyzing electromechanical equipment health monitoring data according to the present invention;
[0044] Figure 3 A sub-flow chart of the electromechanical equipment health monitoring data analysis method of the present invention;
[0045] Figure 4 This is a structural block diagram of a mechatronic equipment health monitoring data analysis system of the present invention;
[0046] Figure 5 The present invention is a structural block diagram of a computer device. DETAILED DESCRIPTION
[0047] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by technicians in the technical field to which this application belongs; the terms used in the specification of the application are only for the purpose of describing specific embodiments and are not intended to limit this application; the terms "including" and "having" in the specification and claims of this application and the above-mentioned figure descriptions and any variations thereof are intended to cover non-exclusive inclusions.
[0048] Reference to "embodiments" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present application. The appearance of the phrase in various locations in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.
[0049] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0050] like Figure 1 As shown, a schematic diagram of a system architecture for implementing the electromechanical equipment health monitoring data analysis method of the present invention is provided;
[0051] The system architecture 100 of the embodiment of the present invention may include electromechanical devices 101, 102, 103, a network 104 and a server 105. The network 104 is used to provide a medium for communication links between the electromechanical devices 101, 102, 103 and the server 105.
[0052] The network 104 may include various connection types, such as wired, wireless communication links or fiber optic cables, etc.; the electromechanical devices 101, 102, 103 may be used to perform data transmission with the server 105 via the network 104.
[0053] The electromechanical equipment 101, 102, 103 may be station house electromechanical equipment with a data entry module;
[0054] The server 105 may be a server that provides various services, such as a server that processes data sent by the electromechanical devices 101 , 102 , 103 .
[0055] It should be understood that Figure 1 The number of electromechanical devices, networks, and servers in the embodiment is only for illustration. Any number of electromechanical devices, networks, and servers may be provided according to implementation requirements.
[0056] The specific implementation of the electromechanical equipment health monitoring data analysis method and system of the present invention is described in detail below in conjunction with specific embodiments.
[0057] like Figure 2 FIG. 1 is a flow chart of a method for analyzing data of electromechanical equipment health monitoring provided by an embodiment of the present invention. The method for analyzing data includes the following steps:
[0058] Step S201: training multiple time series data anomaly detection models using a sample-balanced database, and configuring the multiple time series data anomaly detection models in corresponding device subsystems according to their device identification information;
[0059] The present invention trains the detection model through a sample-balanced database, so that the number of samples in each category in the database is relatively balanced, thereby improving the generalization ability and accuracy of the model.
[0060] Among them, the sample-balanced database adopts an oversampling minority class strategy on the one hand, and an undersampling majority class strategy on the other hand; the former balances the database by increasing the number of abnormal samples, and increasing abnormal samples can be achieved through random oversampling and clustering-based oversampling methods; the latter achieves balance by reducing the number of normal samples, and reducing normal samples can be achieved through random undersampling methods.
[0061] Based on this, the present invention uses a sample-balanced database to train multiple time series data anomaly detection models. Multiple time series data anomaly detection models are respectively configured in corresponding device subsystems according to their device identification information, and independent abnormal data detection is performed in each device subsystem, avoiding the pressure on the data transmission channel caused by excessive data passing through the data transmission channel.
[0062] like Figure 3As shown, it is a sub-flow chart of the electromechanical equipment health monitoring data analysis method of the present invention; preferably, the present invention uses a sample-balanced database to train multiple steps based on time series data anomaly detection models, including:
[0063] Step S301: Acquire a plurality of sample-balanced databases containing device identification information;
[0064] Step S302: using the sample-balanced database to train a time series data anomaly detection model containing device identification information.
[0065] In the disclosed embodiments, for each electromechanical device, a data acquisition sensor is installed on the device to collect key operating parameters such as temperature, vibration, and current; a data acquisition system is configured to ensure that the data collected from the sensor can be transmitted to a central database in real time; a unique identifier is assigned to each device, and this identification information is associated with the collected data, so that multiple databases containing device identification information can be obtained. In order to avoid imbalance in the acquired database samples, an oversampling minority class strategy or an undersampling majority class strategy is adopted to balance the database samples.
[0066] In addition, this embodiment also includes the step of preprocessing the collected data, and the preprocessing step includes: cleaning the collected data, including removing noise, processing missing values, etc.
[0067] Outputting the anomaly detection result sent to the monitoring main system in the data transmission channel through a trained anomaly detection model based on time series data, wherein the anomaly detection model based on time series data includes a data processing module and an anomaly detection module;
[0068] In the data processing module, the newly acquired device time series data is preprocessed, and the frequency components and the corresponding number of channels, number of sequences and length of the detection time window are obtained by decomposition to form a time series to be detected;
[0069] Equipment time series data is a data sequence collected from electromechanical equipment in chronological order using data acquisition sensors. This data reflects the status of the equipment during operation and is crucial for predicting equipment failures, optimizing maintenance plans, and ensuring production safety. By analyzing these data, the operating status of the equipment can be evaluated and corresponding decisions can be made in a timely manner.
[0070] In the step of decomposing the frequency components of the present invention, the embodiment of the present invention uses a wavelet transform decomposition method to split the original time series data into multiple components. These components can reflect the characteristics of the data at different frequencies, and then can extract the trend, cycle and random components of the data, which is helpful for more in-depth analysis of the characteristics of the data.
[0071] Different data streams are collected from electromechanical equipment using data acquisition sensors, and each data stream is called a channel. In addition, the sequence number of the time series is the number of observation points in the time series data, that is, the length of the time series, which affects the time range and level of detail of the analysis. The detection time window length refers to the time span of the data selected during the analysis.
[0072] The present invention constructs a time series to be detected through the obtained frequency components, number of channels, number of sequences and appropriate detection time window length. The time series to be detected is the input of the anomaly detection model. The model will be trained and predicted based on these data to identify abnormal conditions that are inconsistent with the normal operating mode.
[0073] Furthermore, in the anomaly detection module, an anomaly scoring function is constructed to represent the difference between the time series to be detected and the normal sequence model, and the anomaly score is calculated using the anomaly scoring function; the abnormal sequence is found using a local adaptive threshold method, wherein the anomaly threshold for each sliding window is defined as 4 standard deviations different from the average value of the anomaly score in the sliding window; the sequence with an anomaly score greater than the anomaly threshold is identified as an anomaly, and a continuous anomaly sequence is obtained.
[0074] The present invention is based on a time series data anomaly detection model, constructs an anomaly scoring function for representing the difference between the time series to be detected and the normal sequence model, and calculates the anomaly score using the anomaly scoring function; uses a local adaptive threshold method to find the abnormal sequence, thereby improving the accuracy of abnormal data detection.
[0075] Please continue reading Figure 2 , the implementation steps of the electromechanical equipment health monitoring data analysis method of the present invention also include: step S202: constructing multiple data transmission channels between the monitoring main system and multiple equipment subsystems;
[0076] Step S203: During the maintenance cycle of the electromechanical equipment, in response to the abnormal detection results sent to the monitoring main system in the data transmission channel, the abnormal data corresponding to the abnormal detection results are used as the input of the prediction model, and the output of the prediction model is output together with the abnormal detection results in the monitoring main system for early warning.
[0077] In the embodiment of the present invention, the time series database InfluxDB is used to store the time series data sent by the device subsystem to facilitate query and analysis.
[0078] The present invention uses the abnormal data corresponding to the abnormal detection result as the input of the prediction model, and outputs the output of the prediction model together with the abnormal detection result in the monitoring main system for early warning. Not only can the abnormal data of each equipment subsystem be seen in the monitoring system, but also the state prediction result of the equipment corresponding to the abnormal data in a future time period can be intuitively known, providing data support for the further processing of electromechanical equipment.
[0079] Furthermore, in an embodiment of the present invention, for each abnormal sequence in the continuous abnormal sequence, the maximum abnormal score of the abnormal sequence is first obtained, and then the sequences are sorted in descending order, and the decline percentage is calculated. When the first decline percentage does not exceed the decline threshold, the remaining sequences in the continuous abnormal sequence are eliminated.
[0080] It is understandable that in the embodiment of the present disclosure, when the percentage of decrease is less than or equal to the preset decrease threshold, it indicates that the trend of the abnormal sequence has significantly decreased, and the abnormal state can be considered to have ended. At this time, the remaining part of the continuous abnormal sequence can be eliminated and no longer treated as an abnormality; eliminating sequences that are no longer important helps save resources for subsequent analysis and processing.
[0081] Furthermore, in an embodiment of the present invention, the step of constructing an anomaly scoring function for representing the difference between the time series to be detected and the normal sequence model includes:
[0082] For each time point t, in the sequence to be tested and the expected value of the normal sequence of the anomaly detection model Calculate the residual between ;
[0083] The anomaly scoring function is expressed as: .
[0084] Further, in the embodiment of the present invention, in the step of finding the abnormal sequence by using the local adaptive threshold, a sliding window W is defined, and for each sliding window position w, the average value of the abnormal scores of all time points in the window is calculated: ;
[0085] Compute the standard deviation within a window: ;
[0086] For each sliding window, the anomaly threshold The anomaly threshold is the average anomaly score plus 4 times the standard deviation. It is expressed as: .
[0087] In a preferred embodiment of the present disclosure, the step of identifying a sequence having an anomaly score greater than an anomaly threshold as an anomaly to obtain a continuous anomaly sequence includes:
[0088] For each time point t, if the anomaly score Greater than the abnormal threshold of the corresponding sliding window , then the sequence at that time point is considered abnormal;
[0089] The sequences corresponding to multiple time points identified as abnormal are sorted to form a continuous abnormal sequence.
[0090] Furthermore, the input of the prediction model is an abnormal sequence of abnormal detection results; the output of the prediction model is a state prediction sequence of the electromechanical equipment corresponding to the future time period;
[0091] The structure of the prediction model includes an input layer, a convolutional layer, a pooling layer, an LSTM layer, a fully connected layer and an output layer;
[0092] The convolution layer includes three one-dimensional convolution layers, which are used to process time series data and capture data to obtain local features. Each convolution layer is followed by a RELU activation function for introducing nonlinearity. After each convolution layer, the local features of the input sequence are converted into 32 different feature representations.
[0093] The pooling layer performs a one-dimensional maximum pooling operation, the pooling window size is 2, and the maximum value of every two consecutive time points on the sequence time axis is taken as the output;
[0094] The LSTM layer includes 64 LSTM units;
[0095] The fully connected layer is arranged after the LSTM layer and is used to map the features learned by the LSTM layer to the final output layer;
[0096] Therefore, the prediction model provided by the present invention can capture the complex patterns of the health status sequence of the substation equipment and predict the accurate future health trend of the electromechanical equipment.
[0097] like Figure 4 As shown, in another embodiment of the present invention, a data analysis system for monitoring health of electromechanical equipment is provided, and the data analysis system 400 includes the following steps:
[0098] Anomaly detection module 401, used to train multiple anomaly detection models based on time series data using a sample-balanced database, and multiple anomaly detection models based on time series data are respectively configured in corresponding device subsystems according to their device identification information;
[0099] The data transmission module 402 is used to construct multiple data transmission channels between the monitoring main system and multiple device subsystems;
[0100] The analysis and early warning module 403 is used to respond to the abnormal detection results sent to the monitoring main system in the data transmission channel during the maintenance cycle of the electromechanical equipment, use the abnormal data corresponding to the abnormal detection results as the input of the prediction model, and output the output of the prediction model together with the abnormal detection results in the monitoring main system for early warning.
[0101] In order to solve the above technical problems, an embodiment of the present application also provides a computer device.
[0102] Please refer to Figure 5 , Figure 5 This is a basic structural block diagram of the computer device in this embodiment.
[0103] The computer device 500 includes a memory 501, a processor 502, and a network interface 503 which are interconnected and communicated with each other via a system bus.
[0104] It should be noted that the figure only shows a computer device 500 having a component connection memory 501, a processor 502, and a network interface 503, but it should be understood that it is not required to implement all the components shown, and more or fewer components may be implemented instead.
[0105] Those skilled in the art can understand that a computer device is a device that can automatically perform numerical calculations and / or information processing according to pre-set or stored instructions, and its hardware includes but is not limited to microprocessors, application specific integrated circuits (ASICs), programmable gate arrays (FPGAs), digital signal processors (DSPs), embedded devices, etc.
[0106] The memory 501 includes at least one type of readable storage medium, and the readable storage medium includes flash memory, hard disk, multimedia card, card-type memory (for example, SD or D interface display memory, etc.), random access memory (RAM), static random access memory (SRAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), programmable read-only memory (PROM), magnetic memory, disk, optical disk, etc.
[0107] In some embodiments, the memory 501 may be an internal storage unit of the computer device 500, such as a hard disk or memory of the computer device 500. In other embodiments, the memory 501 may also be an external storage device of the computer device 500, such as a plug-in hard disk, a Smart Media Card (SMC), a Secure Digital (SD) card, a Flash Card, etc., equipped on the computer device 500.
[0108] Of course, the memory 501 may also include both the internal storage unit and the external storage device of the computer device 500. In this embodiment, the memory 501 is generally used to store the operating system and various application software installed on the computer device 500, such as the program code of the electromechanical equipment health monitoring data analysis method.
[0109] In addition, the memory 501 can also be used to temporarily store various types of data that have been output or are to be output; the processor 502 can also be a central processing unit (CPU), a controller, a microcontroller, a microprocessor, or other data processing chips in some embodiments. The processor 502 is generally used to control the overall operation of the computer device 500. In this embodiment, the processor 502 is used to run the program code stored in the memory 501 or process data, such as running the program code of the electromechanical equipment health monitoring data analysis method.
[0110] The network interface 503 may include a wireless network interface or a wired network interface. The network interface 503 is generally used to establish a communication connection between the computer device 500 and other electronic devices.
[0111] The present application also provides another embodiment, namely, providing a computer-readable storage medium, wherein the computer-readable storage medium stores an interface display program, and the interface display program can be executed by at least one processor so that the at least one processor performs the steps of the above-mentioned electromechanical equipment health monitoring data analysis method.
[0112] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above embodiment methods can be implemented by means of software plus a necessary general hardware platform, and of course by hardware, but in many cases the former is a better implementation method.
[0113] Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, can essentially be embodied in the form of a software product. The computer software product is stored in a storage medium (such as ROM / RAM, disk, or CD), and includes a number of instructions for enabling a terminal device (which can be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the electromechanical equipment health monitoring data analysis method described in each embodiment of the present application.
[0114] Obviously, the embodiments described above are only some embodiments of the present application, rather than all embodiments. The preferred embodiments of the present application are given in the accompanying drawings, but they do not limit the patent scope of the present application. The present application can be implemented in many different forms. On the contrary, the purpose of providing these embodiments is to make the understanding of the disclosure of the present application more thorough and comprehensive. Although the present application is described in detail with reference to the aforementioned embodiments, for those skilled in the art, it is still possible to modify the technical solutions recorded in the aforementioned specific implementation methods, or to replace some of the technical features therein with equivalents. Any equivalent structure made using the contents of the specification and drawings of this application, directly or indirectly used in other related technical fields, is similarly within the scope of patent protection of this application.
Claims
1. A method for analyzing electromechanical equipment health monitoring data, It is characterized in that The data analysis method includes the following steps: A sample-balanced database is used to train multiple time series data anomaly detection models, and the multiple time series data anomaly detection models are respectively configured in corresponding device subsystems according to their device identification information; Construct multiple data transmission channels between the monitoring main system and multiple equipment subsystems; During the maintenance cycle of the electromechanical equipment, in response to the abnormal detection results sent to the monitoring main system in the data transmission channel, the abnormal data corresponding to the abnormal detection results are used as the input of the prediction model, and the output of the prediction model is outputted in the monitoring main system together with the abnormal detection results for early warning; Outputting the anomaly detection result sent to the monitoring main system in the data transmission channel through a trained anomaly detection model based on time series data, wherein the anomaly detection model based on time series data includes a data processing module and an anomaly detection module; In the data processing module, the newly acquired device time series data is preprocessed, and the frequency components and the corresponding number of channels, number of sequences and length of the detection time window are obtained by decomposition to form a time series to be detected; In the anomaly detection module, an anomaly scoring function is constructed to represent the difference between the expected value of the time series to be detected and the normal series, and the anomaly score is calculated using the anomaly scoring function; the abnormal sequence is found by using a local adaptive threshold method, wherein the anomaly threshold for each sliding window is defined as 4 standard deviations different from the average value of the anomaly score in the sliding window; the sequence with an anomaly score greater than the anomaly threshold is identified as an anomaly, and a continuous anomaly sequence is obtained; The step of constructing an abnormality scoring function for representing the difference between the expected value of the time series to be detected and the normal series includes: For each time point t, in the sequence to be tested and the expected value of the normal sequence of the anomaly detection model Calculate the residual between ; The anomaly scoring function is expressed as: ; In the step of finding abnormal sequences by using a local adaptive threshold, a sliding window W is defined, and for each sliding window position w, the average value of the abnormal scores of all time points in the window is calculated: ; Compute the standard deviation within a window: ; For each sliding window, the anomaly threshold The anomaly threshold is the average anomaly score plus 4 times the standard deviation. It is expressed as: ; The step of identifying a sequence whose anomaly score is greater than an anomaly threshold as an anomaly and obtaining a continuous anomaly sequence includes: For each time point t, if the anomaly score Greater than the abnormal threshold of the corresponding sliding window , then the sequence at that time point is considered abnormal; The sequences corresponding to multiple time points identified as abnormal are sorted to form a continuous abnormal sequence.
2. The electromechanical equipment health monitoring data analysis method according to claim 1, It is characterized in that The step of using a sample-balanced database to train multiple time series data anomaly detection models includes: Obtain a plurality of sample-balanced databases containing device identification information; A sample-balanced database is used to train a time series data anomaly detection model containing device identification information.
3. The method for analyzing electromechanical equipment health monitoring data according to claim 2, It is characterized in that For each abnormal sequence in the continuous abnormal sequence, first obtain the maximum abnormal score of the abnormal sequence, then sort it in descending order, calculate the decline percentage, and when the first decline percentage does not exceed the decline threshold, the remaining sequences in the continuous abnormal sequence are removed.
4. The method for analyzing electromechanical equipment health monitoring data according to claim 3, It is characterized in that The input of the prediction model is an abnormal sequence of abnormal detection results; the output of the prediction model is a state prediction sequence of the electromechanical equipment corresponding to the future time period; The structure of the prediction model includes an input layer, a convolutional layer, a pooling layer, an LSTM layer, a fully connected layer and an output layer; The convolution layer includes three one-dimensional convolution layers, which are used to process time series data and capture data to obtain local features. Each convolution layer is followed by a RELU activation function for introducing nonlinearity. After each convolution layer, the local features of the input sequence are converted into 32 different feature representations. The pooling layer performs a one-dimensional maximum pooling operation, the pooling window size is 2, and the maximum value of every two consecutive time points on the sequence time axis is taken as the output; The LSTM layer includes 64 LSTM units; The fully connected layer is arranged after the LSTM layer and is used to map the features learned by the LSTM layer to the final output layer.
5. A data analysis system for implementing the electromechanical equipment health monitoring data analysis method according to any one of claims 1 to 4, It is characterized in that The data analysis system includes the following modules: An anomaly detection module is used to train multiple anomaly detection models based on time series data using a sample-balanced database, and the multiple anomaly detection models based on time series data are respectively configured in corresponding device subsystems according to their device identification information; A data transmission module, used to construct multiple data transmission channels between the monitoring main system and multiple device subsystems; The analysis and early warning module is used to respond to the abnormal detection results sent to the monitoring main system in the data transmission channel during the maintenance cycle of the electromechanical equipment, use the abnormal data corresponding to the abnormal detection results as the input of the prediction model, and output the output of the prediction model together with the abnormal detection results in the monitoring main system for early warning.
