Industrial equipment fault diagnosis system based on multi-class time series data analysis

Abstract

The invention relates to the technical field of industrial equipment monitoring and management, in particular to an industrial equipment fault diagnosis system based on multi-class time series data analysis, which comprises a data acquisition unit used for acquiring operation data of multi-class industrial equipment, the operation data comprising a plurality of to-be-trained operation time series data and a plurality of to-be-diagnosed operation time series data; the model training unit is used for analyzing a plurality of data types corresponding to the to-be-trained operation time sequence data and transmitting the to-be-trained operation time sequence data corresponding to each data type to a pre-constructed initial fault diagnosis model for training to obtain a target fault diagnosis model corresponding to each data type; and the fault diagnosis unit is used for analyzing a plurality of data types corresponding to the to-be-diagnosed operation time sequence data and transmitting the to-be-diagnosed operation time sequence data to the corresponding target fault diagnosis model according to the data types. According to the invention, when the industrial equipment is monitored and managed, the efficiency and accuracy of fault diagnosis are improved through analysis of multiple types of time series data.

Description

Technical Field

[0001] The present invention relates to the technical field of industrial equipment monitoring and management, and in particular to an industrial equipment fault diagnosis system based on multi-type time series data analysis. Background Art

[0002] With the development of industrial technology, the types of industrial equipment are becoming more and more diverse. In industrial scenarios, a factory often includes various industrial equipment, such as wind turbines, port cranes, chemical equipment, and power transmission equipment. As the time of industrial equipment operation increases, the performance of the equipment will gradually decline, and the probability of failure will gradually increase. Failure of industrial equipment often triggers a chain reaction. If it is not discovered and eliminated in time, it will affect the operation of industrial production and even cause serious economic losses and casualties. Therefore, fault diagnosis of industrial equipment is very important.

[0003] At present, the fault diagnosis of industrial equipment mainly involves manual labeling of data, and then anomaly detection and fault prediction through machine learning and other methods. However, there are many types of industrial equipment. For small and medium-sized enterprises, each type of equipment needs to have a large amount of anomaly data labeled. The lack of versatility of the fault diagnosis system will lead to low efficiency of equipment fault diagnosis. On the other hand, the failure modes of equipment are usually diverse and not fixed, and the sampling data of different sensors have different characteristics. At this time, if the same fault diagnosis standard is used, the accuracy of the diagnosis result will be reduced.

[0004] Therefore, how to improve the efficiency and accuracy of industrial equipment fault diagnosis is an urgent problem that needs to be solved. Summary of the invention

[0005] In order to solve the technical problem of how to improve the efficiency and accuracy of industrial equipment fault diagnosis, the purpose of the present invention is to provide an industrial equipment fault diagnosis system based on multi-type time series data analysis, and the technical solution adopted is as follows:

[0006] In one embodiment of the present application, an industrial equipment fault diagnosis system based on multi-type time series data analysis is provided, and the system includes:

[0007] A data acquisition unit, used to acquire operation data of various types of industrial equipment, wherein the operation data includes various operation time series data to be trained and various operation time series data to be diagnosed;

[0008] A model training unit, used for analyzing a plurality of data types corresponding to the plurality of runtime data to be trained, and transmitting the runtime data to be trained corresponding to each of the data types to a pre-built initial fault diagnosis model for training, so as to obtain a target fault diagnosis model corresponding to each of the data types;

[0009] The fault diagnosis unit is used to analyze multiple data types corresponding to the multiple types of operating sequence data to be diagnosed, and transmit the operating sequence data to be diagnosed to the corresponding target fault diagnosis model according to the data types to obtain a fault diagnosis result.

[0010] In one embodiment of the present application, the system further includes:

[0011] The preprocessing unit is used to perform windowing and differential processing on the multiple types of runtime data to be trained to obtain preprocessed data to be trained, wherein the preprocessed data to be trained is used to be transmitted to the model training unit to analyze and obtain the multiple data types.

[0012] In one embodiment of the present application, the pre-processing unit includes:

[0013] A frequency domain conversion subunit, used for performing frequency domain conversion on the plurality of to-be-trained running time series data to obtain a plurality of to-be-trained running frequency domain data;

[0014] A differential operation subunit is used to analyze the periodic performance of the multiple types of frequency domain data to be trained, determine the periodic significance, determine the periodic data and the non-periodic data according to the periodic significance, perform windowing processing on the periodic data and the non-periodic data, and perform differential operation on the periodic data and the non-periodic data after the windowing processing;

[0015] The normalization subunit is used to perform normalization processing on the data after the difference operation to obtain the pre-processed data to be trained.

[0016] In one embodiment of the present application, analyzing the periodicity performance of the plurality of frequency domain data to be trained and determining the periodicity significance includes:

[0017] Obtaining extreme values ​​of the plurality of frequency domain data to be trained;

[0018] Determining an envelope according to the extreme value;

[0019] The periodic significance degree is determined according to the frequency domain distribution of the plurality of frequency domain data to be trained and the distance distribution of the extreme values ​​corresponding to the envelope.

[0020] In one embodiment of the present application, determining the periodic data and the non-periodic data according to the periodic significance degree, and performing windowing processing on the periodic data and the non-periodic data, includes:

[0021] The data whose periodicity significance is greater than a preset periodicity significance threshold is taken as the periodic data, and the data whose periodicity significance is less than or equal to the preset periodicity significance threshold is taken as the non-periodic data;

[0022] Taking the distance mean of all extreme values ​​on the envelope as a first window size, and performing windowing processing on the periodic data based on the first window size;

[0023] The non-periodic data is clustered to obtain a plurality of clusters, the non-periodic data is divided into data corresponding to a plurality of states according to the distances between the data points of the plurality of clusters, and the data corresponding to each state is windowed.

[0024] In one embodiment of the present application, performing a differential operation on the non-periodic data after the windowing process includes:

[0025] For multiple data segments in the data corresponding to any of the states, when forward differencing is performed on the data segments and the data volume of the data segments is greater than the data volume of the adjacent previous data segments, the data segments are divided into data segments with equal lengths to the adjacent previous data segments for differential operations, and the remaining parts are aligned for differential operations.

[0026] In one embodiment of the present application, the initial fault diagnosis model includes a long short-term memory network model, and the model training unit includes:

[0027] A data classification subunit, configured to analyze the noise manifestation and fault manifestation of the plurality of training running time series data, and determine the plurality of data types according to the analysis results;

[0028] The model training subunit is used to transmit the to-be-trained running time sequence data corresponding to each of the data types to the pre-built long short-term memory network model for training, so as to obtain the target fault diagnosis model corresponding to each of the data types.

[0029] In one embodiment of the present application, the noise manifestation and fault manifestation of the plurality of training running time series data are analyzed, and the plurality of data types are determined according to the analysis results, including:

[0030] For any of the training running time series data, a first classification window is obtained according to a preset second window size, the first classification window is moved, a second classification window is obtained based on new data, and data changes in the first classification window and the second classification window are analyzed to determine data change consistency;

[0031] Based on the consistency of the data changes, analyzing the data changes in the first classification window to determine the degree of fault manifestation;

[0032] Perform similarity comparison on the data in the window with the largest degree of fault manifestation in the two types of running time series data to be trained, and determine a combined training index;

[0033] The multiple types of running time series data to be trained are merged according to the size of the merged training index to determine the multiple types of data.

[0034] In one embodiment of the present application, the to-be-trained running time series data corresponding to each of the data types are transmitted to the pre-built long short-term memory network model for training to obtain the target fault diagnosis model corresponding to each of the data types, including:

[0035] Extracting a training data set from the runtime data to be trained corresponding to each of the data types;

[0036] The training data set is transmitted to the pre-built long short-term memory network model, the possibility of the fault state is output, and the fault state with the greatest possibility of the fault state is taken as the fault state corresponding to the training data set, wherein the fault state includes a normal state and fault states corresponding to multiple fault types.

[0037] In one embodiment of the present application, the analyzing of the multiple data types corresponding to the multiple types of running time series data to be diagnosed includes:

[0038] Performing windowing and differential processing on the plurality of to-be-diagnosed running time series data to obtain pre-processed to-be-diagnosed data;

[0039] For any of the data to be diagnosed, determine the consistency of data changes, analyze the data changes based on the consistency of data changes, determine the degree of fault manifestation, perform similarity comparison on the data in the window with the largest degree of fault manifestation in the two data to be diagnosed, and determine the combined training index;

[0040] The multiple types of data to be diagnosed are merged according to the size of the merged training index to determine the multiple types of data.

[0041] The present invention has the following beneficial effects:

[0042] First, the data acquisition unit acquires the operating data of multiple types of industrial equipment, and the operating data includes multiple operating sequence data to be trained and multiple operating sequence data to be diagnosed; then, the model training unit analyzes multiple data types corresponding to the multiple operating sequence data to be trained, and transmits the operating sequence data to be trained corresponding to each of the data types to a pre-built initial fault diagnosis model for training, so as to obtain a target fault diagnosis model corresponding to each of the data types; finally, the fault diagnosis unit analyzes multiple data types corresponding to the multiple operating sequence data to be diagnosed, and transmits the operating sequence data to be diagnosed to the corresponding target fault diagnosis model according to the data type, so as to obtain a fault diagnosis result. In the present invention, the data acquisition unit acquires the operating data of multiple types of industrial equipment and can receive the operating data from different types of industrial equipment, rather than being limited to a specific type of equipment, so that the system has a wide range of applicability and is suitable for various industrial scenarios such as wind turbines, port cranes, chemical equipment, and power transmission equipment; the model training unit analyzes multiple data types corresponding to multiple operating sequence data to be trained, and can identify and process different types of data (such as vibration data, temperature data, current data, etc.), so as to adapt to the data characteristics collected by different sensors, further enhance the versatility of the system, and avoid the need for separate modeling for each device; the model training unit automatically analyzes the data type and transmits the corresponding data to the initial fault diagnosis model for training, which reduces manual intervention, significantly improves the degree of automation of the fault diagnosis system, and reduces the need for small and medium-sized enterprises to manually label a large number of data. Pay attention to data dependency; by generating a corresponding target fault diagnosis model for each data type, it can quickly adapt to different equipment types and data characteristics without redesigning or adjusting the entire diagnosis process, thereby improving diagnosis efficiency; because the system generates corresponding target fault diagnosis models according to different data types, each target fault diagnosis model can better capture the characteristics and failure modes of a specific type of data. Compared with the method of processing all data with a single model, this method can significantly improve the accuracy of the diagnosis results; the fault diagnosis unit will select the corresponding target fault diagnosis model according to the data type of the runtime data to be diagnosed, ensuring that the diagnosis process is always based on the model that best suits the data type, thereby avoiding misjudgment problems caused by the use of a unified standard; by integrating the operating data of multiple types of industrial equipment, the system can analyze the equipment status from multiple dimensions, further improving the reliability of the diagnosis results. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] In order to more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings required for use in the embodiments or the prior art descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0044] Figure 1 A schematic diagram of an implementation environment of an industrial equipment fault diagnosis system based on multi-type time series data analysis provided by an embodiment of the present invention;

[0045] Figure 2 A schematic diagram of the structure of an industrial equipment fault diagnosis system based on multi-type time series data analysis provided by an embodiment of the present invention;

[0046] Figure 3 A schematic diagram of a training run time data provided by an embodiment of the present invention;

[0047] Figure 4 A schematic diagram of another type of running time series data to be trained provided by an embodiment of the present invention;

[0048] Figure 5 A schematic diagram of the structure of a long short-term memory network provided by an embodiment of the present invention;

[0049] Figure 6 A flowchart of an industrial equipment fault diagnosis method based on multi-type time series data analysis is provided in accordance with an embodiment of the present invention. DETAILED DESCRIPTION

[0050] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the following is a detailed description of the specific implementation, structure, features and effects of an industrial equipment fault diagnosis system based on multi-type time series data analysis proposed by the present invention in combination with the accompanying drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" does not necessarily refer to the same embodiment. In addition, specific features, structures or characteristics in one or more embodiments may be combined in any suitable form.

[0051] It should be noted that the terms "first", "second", etc. in the specification of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0052] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

[0053] A specific scheme of an industrial equipment fault diagnosis system based on multi-type time series data analysis provided by the present invention is described in detail below in conjunction with the accompanying drawings.

[0054] See also Figure 1 , Figure 1 A schematic diagram of an implementation environment of an industrial equipment fault diagnosis system based on multi-type time series data analysis provided by an embodiment of the present invention is shown in FIG. Figure 1 As shown, the implementation environment includes a fault diagnosis terminal 101, a data acquisition terminal 102, and an industrial device 103. The fault diagnosis terminal 101 may be a terminal device configured with a target client, including but not limited to a laptop computer, a tablet computer, a handheld computer, a PAD, a desktop computer, etc. with local computing capabilities. The target client may be a client such as a video client, an instant messaging client, a browser client, etc., which is installed with an industrial equipment fault diagnosis system based on multi-type time series data analysis; the fault diagnosis terminal 101 may, but is not limited to, communicate with the data acquisition terminal 102 through a network, and the network may include but is not limited to: a wired network, a wireless network, wherein the wired network includes: a local area network, a metropolitan area network, and a wide area network, and the wireless network includes: Bluetooth, WIFI, and other networks that implement wireless communication. The fault diagnosis terminal 101 may, but is not limited to, include a human-computer interaction screen, a processor, and a memory. The human-computer interaction screen may, but is not limited to, be used to display fault diagnosis results. The processor may, but is not limited to, be used to respond to human-computer interaction operations, perform corresponding operations, or generate corresponding instructions.

[0055] As an optional method, the operation data of the industrial equipment 103 can be collected through the data acquisition terminal 102. The data acquisition terminal 102 may include, for example: a sensor, a data acquisition module, and a spectrum analyzer. Among them, the sensors may include: a vibration sensor, which is used to measure the vibration signal of the equipment during operation, and is suitable for equipment such as wind turbines and port cranes; a temperature sensor, which is used to monitor the temperature changes of key parts of the equipment, and is suitable for scenes such as chemical equipment and power transmission equipment; a current / voltage sensor, which is used to collect current and voltage data of electrical equipment, and is suitable for equipment such as motors and transformers; a pressure sensor, which is used to measure the pressure value of the hydraulic or pneumatic system, and is suitable for port cranes, chemical equipment, etc.; an acceleration sensor, which is used to capture the acceleration changes of the equipment, and is suitable for high-speed mechanical equipment; an acoustic sensor.

[0056] As an optional manner, the industrial equipment 103 may be, for example: a wind turbine, a port crane, a chemical equipment, a power transmission equipment, a manufacturing production line, and an elevator system.

[0057] As an optional manner, the fault diagnosis terminal 101 may also be a server, which may be a single server, a server cluster composed of multiple servers, or a cloud server. The above is only an example, and this embodiment does not make any limitation to this.

[0058] As an optional method, an industrial equipment fault diagnosis system based on multi-type time series data analysis can be installed on the fault diagnosis terminal 101. The structure of the industrial equipment fault diagnosis system based on multi-type time series data analysis is as follows:

[0059] A data acquisition unit, used to acquire operation data of various types of industrial equipment, wherein the operation data includes various operation time series data to be trained and various operation time series data to be diagnosed;

[0060] A model training unit, used for analyzing a plurality of data types corresponding to the plurality of runtime data to be trained, and transmitting the runtime data to be trained corresponding to each of the data types to a pre-built initial fault diagnosis model for training, so as to obtain a target fault diagnosis model corresponding to each of the data types;

[0061] The fault diagnosis unit is used to analyze multiple data types corresponding to the multiple types of operating sequence data to be diagnosed, and transmit the operating sequence data to be diagnosed to the corresponding target fault diagnosis model according to the data types to obtain a fault diagnosis result.

[0062] Based on the above method, the data acquisition unit acquires the operating data of multiple types of industrial equipment and can receive operating data from different types of industrial equipment, rather than being limited to a specific type of equipment, making the system widely applicable and suitable for various industrial scenarios such as wind turbines, port cranes, chemical equipment, and power transmission equipment; the model training unit analyzes multiple data types corresponding to multiple operating sequence data to be trained, and can identify and process different types of data (such as vibration data, temperature data, current data, etc.), so as to adapt to the data characteristics collected by different sensors, further enhancing the versatility of the system and avoiding the need for separate modeling for each device; the model training unit will automatically analyze the data type and transmit the corresponding data to the initial fault diagnosis model for training, reducing manual intervention, significantly improving the degree of automation of the fault diagnosis system, and reducing the need for a large number of manual labor by small and medium-sized enterprises. Label data dependencies; by generating a corresponding target fault diagnosis model for each data type, it can quickly adapt to different equipment types and data characteristics without redesigning or adjusting the entire diagnosis process, thereby improving diagnosis efficiency; because the system generates corresponding target fault diagnosis models according to different data types, each target fault diagnosis model can better capture the characteristics and failure modes of a specific type of data. Compared with the method of processing all data with a single model, this method can significantly improve the accuracy of the diagnosis results; the fault diagnosis unit will select the corresponding target fault diagnosis model according to the data type of the runtime data to be diagnosed, ensuring that the diagnosis process is always based on the model that best suits the data type, thereby avoiding misjudgment problems caused by the use of a unified standard; by integrating the operating data of multiple types of industrial equipment, the system can analyze the equipment status from multiple dimensions, further improving the reliability of the diagnosis results.

[0063] As an optional example, this embodiment does not limit the host of the above-mentioned industrial equipment fault diagnosis system based on multi-type time series data analysis. The above-mentioned industrial equipment fault diagnosis system based on multi-type time series data analysis can be installed on the fault diagnosis terminal 101. For example, when the fault diagnosis terminal 101 is a desktop computer, some or all of the components of the above-mentioned industrial equipment fault diagnosis system based on multi-type time series data analysis can be installed on the desktop computer.

[0064] In one embodiment of the present application, an industrial equipment fault diagnosis system based on multi-type time series data analysis is provided. Figure 2 A schematic diagram of a structural diagram of an industrial equipment fault diagnosis system based on multi-type time series data analysis provided by an embodiment of the present invention, see Figure 2 , the industrial equipment fault diagnosis system based on multi-class time series data analysis includes the following units 210 to 230:

[0065] The data acquisition unit 210 is used to acquire operation data of multiple types of industrial equipment, where the operation data includes multiple types of operation time series data to be trained and multiple types of operation time series data to be diagnosed.

[0066] Among them, multiple types of industrial equipment means that the system can adapt to multiple types of industrial equipment, such as wind turbines, port cranes, chemical equipment, power transmission equipment, etc. The operating data of these devices have different characteristics and patterns, so a universal method is needed to process them.

[0067] Among them, the various operating time series data to be trained are operating status information collected from various types of industrial equipment, usually including data in the form of time series, such as vibration signals, temperature changes, current fluctuations, etc. These data are used to train fault diagnosis models.

[0068] Among them, a variety of operating time series data to be diagnosed come from various types of industrial equipment, but fault diagnosis is required in actual applications. Similar to the training data, they also contain operating status information in the form of time series.

[0069] The model training unit 220 is used to analyze the multiple data types corresponding to the multiple types of running time series data to be trained, and transmit the running time series data to be trained corresponding to each data type to a pre-built initial fault diagnosis model for training to obtain a target fault diagnosis model corresponding to each data type.

[0070] The data type refers to the result of classifying the operation data according to the characteristics of the data (such as source, physical meaning, statistical characteristics, etc.). For example, vibration data, temperature data, and current data can be regarded as different data types.

[0071] The initial fault diagnosis model is a pre-built basic model that receives training data of a specific data type and generates a more accurate target fault diagnosis model through training. It can be a model based on deep learning or other machine learning models.

[0072] When analyzing the multiple data types corresponding to the multiple types of running time series data to be trained, the system analyzes the input running time series data to be trained and identifies different types of data therein, for example, distinguishing vibration data, temperature data, and current data from a group of data.

[0073] According to the analysis results, the training run-time data of each data type are respectively input into the initial fault diagnosis model for training. After the training is completed, a target fault diagnosis model for each data type is generated. For example, a model specifically used for vibration data analysis is trained from vibration data.

[0074] The fault diagnosis unit 230 is used to analyze multiple data types corresponding to the multiple types of operating sequence data to be diagnosed, and transmit the operating sequence data to be diagnosed to the corresponding target fault diagnosis model according to the data types to obtain a fault diagnosis result.

[0075] In the actual diagnosis process, the system will analyze the received runtime data to be diagnosed and identify the different data types. According to the analysis results, the runtime data to be diagnosed of each data type will be input into the corresponding target fault diagnosis model to generate the fault diagnosis results.

[0076] For example, suppose there is a factory that contains the following three types of industrial equipment: wind turbines, which mainly collect vibration data and temperature data; port cranes, which mainly collect current data and pressure data; and power transmission equipment, which mainly collects voltage data and current data. The data acquisition unit collects operating data from the above equipment to obtain the following two types of data: operating sequence data to be trained: data used for training models; operating sequence data to be diagnosed: data used for actual diagnosis. The model training unit analyzes the operating sequence data to be trained and identifies the following data types: wind turbines: vibration data, temperature data; port cranes: current data, pressure data; power transmission equipment: voltage data, current data. The operating sequence data to be trained of each data type is input into the initial fault diagnosis model for training, and the following target fault diagnosis models are generated: vibration data model, temperature data model, current data model, pressure data model, voltage data model. The fault diagnosis unit analyzes the received operating sequence data to be diagnosed and identifies the data type. According to the data type, the operating sequence data to be diagnosed is input into the corresponding target fault diagnosis model to obtain the fault diagnosis result. For example, if the vibration data of a wind turbine is received, it is input into the vibration data model; if the current data of a port crane is received, it is input into the current data model.

[0077] In one embodiment of the present application, the system further includes:

[0078] The preprocessing unit is used to perform windowing and differential processing on the multiple types of runtime data to be trained to obtain preprocessed data to be trained, wherein the preprocessed data to be trained is used to be transmitted to the model training unit to analyze and obtain the multiple data types.

[0079] Taking the operation data of wind turbines as an example, it includes wind speed, generator speed, power output, temperature, current, frequency, pressure, unit vibration, etc. Figure 3 , Figure 3 A schematic diagram of a running time series data to be trained provided by an embodiment of the present invention, wherein the running time series data to be trained in the schematic diagram is the vibration time series data of the unit; see Figure 4 , Figure 4 This is a schematic diagram of another type of training time series data provided by an embodiment of the present invention, wherein the training time series data in the schematic diagram is wind speed time series data. Figure 3 and Figure 4 It can be seen that due to the influence of the outside world, mechanical noise, and mechanical operation cycle, different data have different sampling frequencies, and the periodic performance and change characteristics of the data are different. Therefore, if the original data is input into the model for learning, the detection accuracy will be insufficient, so preprocessing is required. The differential operation can reflect the changing characteristics of the data, but the effect of the differential operation is poor in data with large noise and weak regularity. By adding windows, treating part of the continuous data as a whole, and then performing differential processing, the processing effect of the data can be guaranteed.

[0080] Windowing refers to dividing the time series data into windows (or segments) of fixed length to analyze local data. The data in each window can be regarded as an independent data segment, which is convenient for subsequent processing and modeling. For example, for a vibration signal, the entire signal can be divided into multiple windows by setting the window size to 100 sampling points and the step size to 10 sampling points.

[0081] Among them, differential processing is a common time series data analysis method used to eliminate trend or periodic components in data and extract the change characteristics of data. Specifically, differential processing generates a new sequence by calculating the difference between adjacent data points. For example, for time series data x1, x2, x3, the first-order difference results are x2-x1, x3-x2.

[0082] Among them, when performing windowing and differential processing on the various types of training running time series data, firstly, windowing is performed on the various types of training running time series data to divide the data into multiple windows of fixed length; then, differential processing is performed on the data in each window to extract the data change characteristics. This preprocessing method can reduce noise interference, highlight the data change pattern, and provide more meaningful input data for subsequent model training.

[0083] For example, suppose a vibration signal is collected from a wind turbine, which contains the following time series data: [1.0, 1.2, 1.5, 1.8, 2.0, 2.2, 2.5, 2.8, 3.0, 3.2]. Set the window size to 4 and the step size to 2, then the original data is divided into the following windows: Window 1, [1.0, 1.2, 1.5, 1.8]; Window 2, [1.8, 2.0, 2.2, 2.5]; Window 3, [2.5, 2.8, 3.0, 3.2]. The data in each window is processed by first-order difference: the difference result of window 1 is: [1.2-1.0, 1.5-1.2, 1.8-1.5]=[0.2, 0.3, 0.3]; the difference result of window 2 is: [2.0-1.8, 2.2-2.0, 2.5-2.2]=[0.2, 0.2, 0.3]; the difference result of window 3 is: [2.8-2.5, 3.0-2.8, 3.2-3.0]=[0.3, 0.2, 0.2]. The data after windowing and difference processing (i.e., difference result) is used as the preprocessed training data and transmitted to the model training unit for analyzing the data type and generating the target fault diagnosis model.

[0084] In the above embodiment, windowing processing can divide long time series data into short windows, which is convenient for focusing on local features and reducing the impact of noise on the overall data. Differential processing can eliminate trend or periodic components in the data, highlight the changing characteristics of the data, and thus improve the sensitivity of the model to abnormal patterns. The preprocessed data is more stable and representative, which can effectively reduce noise interference and data redundancy, and enhance the robustness and generalization ability of the model. By windowing and differential processing the data, the system can more accurately capture the changing pattern of the equipment's operating status, thereby improving the accuracy of fault diagnosis. The preprocessed data structure is more regular, which is convenient for subsequent classification, feature extraction and model training, and significantly simplifies the entire fault diagnosis process.

[0085] In one embodiment of the present application, the preprocessing unit includes:

[0086] A frequency domain conversion subunit, used for performing frequency domain conversion on the plurality of to-be-trained running time series data to obtain a plurality of to-be-trained running frequency domain data;

[0087] A differential operation subunit is used to analyze the periodic performance of the multiple types of frequency domain data to be trained, determine the periodic significance, determine the periodic data and the non-periodic data according to the periodic significance, perform windowing processing on the periodic data and the non-periodic data, and perform differential operation on the periodic data and the non-periodic data after the windowing processing;

[0088] The normalization subunit is used to perform normalization processing on the data after the difference operation to obtain the pre-processed data to be trained.

[0089] Among them, frequency domain conversion is the process of converting time series data from the time domain to the frequency domain, which is usually achieved through Fourier Transform or Wavelet Transform. Frequency domain conversion can reveal the frequency components and periodic characteristics in the data and help analyze the oscillation mode of the signal.

[0090] The periodic performance refers to the periodic characteristics of time series data in the frequency domain. For example, during the operation of some equipment, there may be vibrations or fluctuations of fixed frequencies, which will be manifested as energy concentration at a specific frequency in the frequency domain.

[0091] The periodicity significance is used to quantify the strength of the periodic characteristics of the data. The periodicity significance can be determined by analyzing the energy distribution, envelope shape, or other statistical indicators of the frequency domain data. For example, if the energy proportion of a certain frequency is very high, it indicates that the data has strong periodicity.

[0092] Among them, periodic data refers to data with obvious periodic characteristics, which is usually manifested as energy concentration at a specific frequency in the frequency domain. For example, the rotating parts of a wind turbine may generate periodic vibration signals.

[0093] Among them, non-periodic data refers to data that does not have obvious periodic characteristics, and its frequency domain performance is usually continuous or scattered, such as irregular fluctuations caused by random noise or equipment failure.

[0094] Among them, normalization is a data preprocessing method used to scale data to a fixed range (such as [0, 1] or [-1, 1]) to eliminate the impact of dimensional differences and numerical ranges. Normalization can improve the convergence speed and stability of model training, while avoiding misjudgment caused by data scale differences.

[0095] For example, obtain n types of operating data of several types of industrial equipment ; Among them, the sampling frequency of each data is ; The number of samples for each data is . Periodicity is the basic characteristic of data. The window size can be determined according to the periodic performance of the running data. By drawing the upper and lower envelopes, the extreme value changes of the data can be obtained, and then the period can be obtained. Due to the existence of extreme noise, the periodic characteristics reflected by the envelope will be affected, but the extreme noise data is small. The impact of extreme noise data can be reduced by frequency domain conversion. Perform differential operations on all data, and then use the sigmoid function (an activation function) for normalization processing to convert all data to the [0, 1] range to complete data preprocessing.

[0096] In the above embodiment, frequency domain conversion can convert time series data into frequency domain representation, thereby more clearly revealing the periodic characteristics in the data. This is particularly important for analyzing the operating status of industrial equipment, because many equipment failures will show specific periodic changes. Periodic significance analysis further quantifies the periodic characteristics of the data, allowing the system to distinguish between periodic data and non-periodic data, providing a basis for subsequent processing. By performing windowing and differential operations on periodic data and non-periodic data respectively, the system can adopt different processing strategies for different types of data, thereby improving the accuracy of data classification. Normalization eliminates the dimensional differences and numerical range effects between data, making the model training process more stable and efficient. Normalized data is easier to be learned by models such as neural networks, thereby accelerating convergence and improving model performance.

[0097] In one embodiment of the present application, analyzing the periodic performance of the plurality of frequency domain data to be trained and determining the periodic significance includes:

[0098] Obtaining extreme values ​​of the plurality of frequency domain data to be trained;

[0099] Determining an envelope according to the extreme value;

[0100] The periodic significance degree is determined according to the frequency domain distribution of the plurality of frequency domain data to be trained and the distance distribution of the extreme values ​​corresponding to the envelope.

[0101] Among them, extreme values ​​refer to the locations where the energy distribution reaches a local maximum or minimum in the frequency domain data. These extreme points usually reflect important characteristics of the periodic components in the signal. For example, in the frequency domain graph, the energy peaks of certain frequencies may correspond to the vibration frequencies or harmonic frequencies in the operation of the equipment.

[0102] The envelope is a curve used to describe the range of signal amplitude variation, which can reflect the periodic or non-periodic characteristics of the signal. In frequency domain analysis, the envelope can be obtained by fitting the extreme values ​​of frequency domain data to further quantify the energy distribution law of the signal.

[0103] The frequency domain distribution describes the energy distribution of the signal at different frequencies. By analyzing the frequency domain distribution, the main frequency components of the signal and their strength can be identified. For example, the vibration signal of a wind turbine may show significant energy concentration in a certain frequency range.

[0104] The distance distribution refers to the statistical characteristics of the distances between adjacent extreme points on the envelope. By analyzing the distribution of these distances, we can determine the significance of the periodicity of the signal. If the distances between adjacent extreme points are relatively uniform, it indicates that the signal has a strong periodicity; conversely, if the distance distribution is chaotic, it indicates that the signal is non-periodic.

[0105] Exemplarily, when determining the degree of periodic significance, the following steps can be followed: obtain all the maximum and minimum values ​​of the data, connect them with straight lines, and obtain the upper and lower envelopes; for the upper and lower envelopes of any data, obtain the maximum and minimum values ​​of all envelopes, any two adjacent maximum or minimum values ​​reflect the period of the data, and the distance distribution between the maximum or minimum values ​​reflects the degree of periodic significance of the data; taking all the maximum values ​​of the upper envelope of any data as an example, obtain all the data between any two adjacent maximum values, and obtain the corresponding frequency domain distribution ; Get the frequency domain distribution of all adjacent maximum distances ; Based on the law of frequency domain distribution and the law of distance between envelope maxima, the periodic significance of any data is calculated; wherein, the distance between adjacent maxima is the interval between two adjacent maxima, which can be the interval in frequency; wherein, the value within the distance between adjacent maxima is the amplitude or energy size corresponding to the frequency component.

[0106] Exemplarily, the expression of the periodic significance degree may be:

[0107]

[0108] in, Indicates the significance of the periodicity of any data; Represents the exponential function in mathematical calculations; The number of types representing different adjacent maximum distances; is the number of types of adjacent maximum distances of the rth type; Represents the number of occurrences of all maximum distances; Indicates the number of maximum values; Indicates the minimum value in mathematical calculation; Represents the number of numerical distributions within the distance between adjacent maxima; It represents the frequency difference between the jth value of the cth maximum value segment (the data segment between the cth maximum value and the c+1th maximum value) and the sth maximum value segment.

[0109] in, The ratio of the number of types of different adjacent maximum value distances on the envelope to the number of occurrences of all maximum value distances characterizes the significance of the periodicity of the data. The larger the value, the wider the distribution of values, and the weaker the periodicity.

[0110] in, Represents the sum of the frequency differences of all values ​​of the c-th maximum segment and the s-th maximum segment.

[0111] in, It means obtaining the sum of the frequency differences between the cth maximum value segment and the minimum frequency difference among all maximum value segments. This value reflects the difference between the cth maximum value segment and the overall data.

[0112] in, It indicates the difference between all maximum value segments and the overall data. The larger the value is, the weaker the regularity and periodicity of the data are.

[0113] In the above embodiment, by extracting the extreme values ​​of the frequency domain data and fitting the envelope, the system can more accurately capture the periodic characteristics of the signal. This method is more intuitive and quantitative than directly analyzing the frequency domain distribution. The extreme value distance distribution corresponding to the envelope provides a quantitative indicator for evaluating the significance of the periodicity of the signal. This multi-dimensional analysis method can reduce the risk of misjudgment and improve the reliability of the diagnostic results. For complex industrial equipment signals, the frequency domain distribution may contain multiple frequency components and noise interference. Through envelope analysis and distance distribution statistics, the system can quickly screen out the main periodic characteristics and simplify the subsequent processing flow.

[0114] In one embodiment of the present application, determining the periodic data and the non-periodic data according to the periodic significance degree, and performing windowing processing on the periodic data and the non-periodic data, includes:

[0115] The data whose periodicity significance is greater than a preset periodicity significance threshold is taken as the periodic data, and the data whose periodicity significance is less than or equal to the preset periodicity significance threshold is taken as the non-periodic data;

[0116] Taking the distance mean of all extreme values ​​on the envelope as a first window size, and performing windowing processing on the periodic data based on the first window size;

[0117] The non-periodic data is clustered to obtain a plurality of clusters, the non-periodic data is divided into data corresponding to a plurality of states according to the distances between the data points of the plurality of clusters, and the data corresponding to each state is windowed.

[0118] Among them, clustering is an unsupervised learning method used to divide data points into several groups (clusters) so that the data points in the same group have high similarity, while the data points between different groups have large differences. In this scheme, clustering is applied to non-periodic data, with the purpose of dividing it into different states according to the distribution characteristics of the data. Clusters are the result of clustering, which represent a group of data points with similar characteristics. Each cluster corresponds to a certain pattern or state in the data. For example, in the operation of industrial equipment, there may be three states: "normal operation", "minor abnormality" and "serious failure", and these states can be obtained by clustering analysis. The corresponding clusters. Multiple states refer to different data categories divided by clustering, and each state corresponds to a specific situation in the operation of the equipment. For non-periodic data, multiple states can reflect the dynamic change characteristics of the equipment under different conditions.

[0119] For example, the higher the periodic significance, the stronger the regularity of the data, and the better the analysis effect of the data by increasing the window for differential operation, and vice versa. The preset periodic significance threshold is 0.7, and the data with a periodic significance greater than 0.7 is recorded as periodic data. The first window size is set to the average distance of all adjacent maximum values ​​of the upper envelope. For example, for non-periodic data, its changes are mostly related to the external environment, human operation, and natural wear and tear of equipment. Data can be classified through clustering, and different categories of data can be added with windows and then moved to perform differential operations, which can ensure the differential effect and distinguish the status of the operating data.

[0120] For example, when clustering non-periodic data, the following hierarchical clustering algorithm can be used for clustering: each data point is regarded as a separate cluster; for any data point, it is merged with the data point with the closest distance to obtain several new clusters; and it is iterated and merged continuously; until the distance between any cluster and all other clusters is greater than (maximum value - minimum value) or all data are merged into one cluster; the distance between two clusters is the minimum distance between any two data points; several clusters are obtained, and the minimum distance between any two data points between clusters is less than or equal to (maximum value - minimum value) is recorded as data in the same state, and data in several states of non-periodic data are obtained; because the time when the data is in the same state is not the same, the window size is not fixed.

[0121] In the above embodiment, the non-periodic data is divided into multiple clusters through clustering processing, which can more accurately capture the hidden patterns and state change characteristics in the data. This method is more sophisticated than the traditional unified processing method and avoids misjudgment caused by data diversity. Different windowing strategies are used for data in different states, so that the system can flexibly adapt to various data characteristics. For example, data in a low-load state may require a smaller window to capture rapid changes, while data in a high-load state requires a larger window to smooth noise interference.

[0122] In one embodiment of the present application, performing a differential operation on the non-periodic data after the windowing process includes:

[0123] For multiple data segments in the data corresponding to any of the states, when forward differencing is performed on the data segments and the data volume of the data segments is greater than the data volume of the adjacent previous data segments, the data segments are divided into data segments with equal lengths to the adjacent previous data segments for differential operations, and the remaining parts are aligned for differential operations.

[0124] The data segments refer to multiple subsequences obtained by windowing the non-periodic data. Each data segment contains a certain number of data points for subsequent differential operations or other analyses.

[0125] Among them, forward difference is a common difference calculation method used to extract the change trend of data. Forward difference is applied to each data segment to capture the change characteristics of the data.

[0126] Among them, alignment difference means that when the lengths of data segments are inconsistent, the length or position of the data segment is adjusted to align it with the adjacent data segment before performing the difference operation. Specifically, when the amount of data in a data segment is larger than the adjacent previous data segment, it is divided into parts of equal length to the previous data segment, and the difference operation is performed on the remaining parts separately.

[0127] For example, for any state of data segment ,in Indicates the vth data segment of the uth state. When performing forward difference, if The amount of data is greater than , then first Divide into The data segments of equal length are differentiated, and then the remaining parts are aligned and differentiated to complete the differential operation.

[0128] In the above embodiment, the aligned differential method is used to ensure that the differential operations between data segments of different lengths are comparable, avoiding misjudgments caused by inconsistent data segment lengths. When processing longer data segments, the information of all data points is retained through division and alignment operations, minimizing data loss. The differential operation can effectively extract the changing features of the data, and the aligned differential method further improves the accuracy and stability of feature extraction. For example, in industrial equipment fault diagnosis, the differential results can reflect the changing trend of the equipment operating status and help identify potential anomalies.

[0129] In one embodiment of the present application, the initial fault diagnosis model includes a long short-term memory network model, and the model training unit includes:

[0130] A data classification subunit, configured to analyze the noise manifestation and fault manifestation of the plurality of runtime data to be trained, and determine the plurality of data types according to the analysis results;

[0131] The model training subunit is used to transmit the to-be-trained running time sequence data corresponding to each of the data types to the pre-built long short-term memory network model for training, so as to obtain the target fault diagnosis model corresponding to each of the data types.

[0132] Among them, due to the variety of data changes, the data is divided into periodic data and non-periodic data, and the data noise performance is different, and the fault data performance is different. If all data are trained through the same model, it will lead to fault judgment errors. If each type of data is trained separately, the data volume will be insufficient, and the enterprise will still need to manage multiple data of multiple devices separately. Therefore, the data should also be classified according to the data noise performance and fault performance, and then each type of data should be trained separately.

[0133] Among them, the long short-term memory network model is a special recurrent neural network (RNN) designed for processing time series data. It can effectively capture the long-term dependencies in time series data by introducing cell states and gating mechanisms (such as input gates, forget gates, and output gates). In this solution, the LSTM model is used as an initial fault diagnosis model to analyze the complex patterns of industrial equipment operation data.

[0134] Noise manifestation refers to fluctuations or abnormal performance in the runtime data due to environmental interference, measurement errors or other non-fault factors. For example, the vibration signal of a wind turbine may be affected by changes in wind speed, and these changes do not necessarily reflect equipment failure.

[0135] Among them, fault manifestation refers to abnormal patterns or characteristics in the operating time series data caused by equipment failure. For example, bearing damage may cause harmonic components of specific frequencies in the vibration signal, which can be regarded as a fault manifestation.

[0136] The data classification subunit analyzes the noise manifestation and fault manifestation of the runtime data, identifies different characteristics in the data, and divides the data into multiple categories accordingly. For example, some data may mainly contain noise components, while other data exhibit obvious fault characteristics. Through this analysis, the system can more accurately determine the data type, thus providing a basis for subsequent modeling.

[0137] In the above embodiment, the noise is randomly distributed and will recur in the time series data. Fault data is special in that it is different from the range and variation of normal data. Therefore, the performance of the fault data can be obtained by calculating the repeatability of the variation of any data segment in the overall data. By analyzing the noise manifestation and fault manifestation, the system can more accurately distinguish data with different characteristics, avoid misjudging noise as a fault or ignoring potential fault signals. The target fault diagnosis model is trained separately for different types of data, so that each model can focus on the data features of a specific type, thereby improving the accuracy of the diagnosis results.

[0138] In one embodiment of the present application, the noise manifestation and fault manifestation of the plurality of training running time series data are analyzed, and the plurality of data types are determined according to the analysis results, including:

[0139] For any of the training running time series data, a first classification window is obtained according to a preset second window size, the first classification window is moved, a second classification window is obtained based on new data, and data changes in the first classification window and the second classification window are analyzed to determine data change consistency;

[0140] Based on the consistency of the data changes, analyzing the data changes in the first classification window to determine the degree of fault manifestation;

[0141] Perform similarity comparison on the data in the window with the largest degree of fault manifestation in the two types of running time series data to be trained, and determine a combined training index;

[0142] The multiple types of running time series data to be trained are merged according to the size of the merged training index to determine the multiple types of data.

[0143] Among them, by setting a preset window size (second window size), a first classification window is intercepted from the training runtime data, and the window is moved to obtain new data to form a second classification window. Subsequently, the data changes in the two windows are analyzed to capture the change pattern of local data.

[0144] The data change consistency is used to measure the similarity of data changes between the first classification window and the second classification window. If the data change trends of the two windows are consistent, they are considered to have high data change consistency; otherwise, they are considered to have low data change consistency. For example, if the data fluctuation amplitude, frequency or direction of the two windows are similar, their consistency is high.

[0145] Among them, the degree of fault manifestation is based on the results of data change consistency analysis, and further evaluates whether there are potential fault features in the first classification window. The significance of fault features is usually quantified through statistical indicators (such as mean, variance, frequency domain distribution, etc.).

[0146] Among them, similarity comparison refers to comparing the data in the window with the largest degree of fault manifestation in the two training running time series data and calculating the similarity between them. Commonly used methods include Euclidean distance, cosine similarity or mutual information.

[0147] The merge training index is a quantitative indicator used to evaluate whether two types of training time series data can be merged into the same data type for training. If the similarity comparison results show that the two types of data have a high similarity, the merge training index is larger; otherwise, it is smaller.

[0148] According to the size of the combined training index, it is determined whether to combine the multiple types of training time series data into the same data type. If the combined training index exceeds a preset threshold, the data are classified into the same category; otherwise, they are kept separately processed.

[0149] Exemplarily, for any of the training running time series data, the first classification window is obtained (10-100 sampling points, 1 is the step length), for the data segment in the first classification window, the data segment is shifted left and right to reach the second classification window , the first classification window Move up and down so that the second classification window As much data as possible appears in the first classification window middle.

[0150] Exemplary, the first classification window and the second classification window The consistency of data changes can be expressed as follows:

[0151]

[0152] in, Indicates the first classification window and the second classification window Consistency of data changes; Represents the exponential function in mathematical calculations; Indicates the window size; Indicates the first classification window The difference between the ath data and the previous data, where when a is equal to 1, the difference between the first data and the last data is obtained; Indicates the second classification window The difference between the ath data and the previous data in , where when a is equal to 1, the difference between the first data and the last data is obtained.

[0153] in, Indicates the first classification window and the second classification window The consistency of data changes at corresponding positions. If the difference is smaller, it means that the data trends in the windows are consistent, and thus the consistency of data changes is greater.

[0154] Exemplarily, by translating the first classification window up and down and left and right multiple times, multiple second classification windows can be obtained, and then the data change consistency between the first classification window and all the second classification windows that have been reached can be obtained, and all the obtained data change consistency are sorted from large to small, so that each first classification window corresponds to a sorting result of data change consistency. Among them, if there are more high data change consistency in the sorting result, it means that the more times the data changes in the first classification window appear in the overall data, the smaller the degree of fault manifestation, that is, the weaker the fault manifestation. On the contrary, if the data change consistency in the sorting result is relatively low, it means that the fault manifestation degree of the first classification window is greater, that is, the higher the fault manifestation. Among them, the level of data change consistency can be judged by a preset data change consistency threshold.

[0155] Exemplarily, the fault manifestation degree of the first classification window may be expressed as follows:

[0156]

[0157] in, Indicates the degree of fault manifestation of the first classification window; represents an exponential function in a mathematical calculation method; w is the number of second classification windows corresponding to the first w data change consistencies in the sorting result of the data change consistency, and in this embodiment, w can be 100; Indicates the first classification window and the second classification window Consistency of data changes; It means adding the data change consistency of the first classification window and w second classification windows.

[0158] It should be noted that in the embodiment of the present application, a first classification window is first intercepted according to the second window size, and then the above steps are used to determine the fault manifestation degree corresponding to the first classification window; then, a new first classification window is intercepted according to the second window size, and then the above steps are used to determine the fault manifestation degree corresponding to the new first classification window; and so on, the fault manifestation degrees corresponding to multiple first classification windows can be obtained.

[0159] Exemplarily, for any data, e first classification windows with the largest fault manifestation degree are obtained. If the data changes in the windows with the largest fault manifestation degree of two data are relatively similar, the two data can be trained together. The combined training index of any two data can be expressed as:

[0160]

[0161] in, Represents the combined training index of the i-th data and the h-th data; Represents the exponential function in mathematical calculations; Indicates the difference between the a-th data and the previous data in the e-th first classification window with the largest degree of fault manifestation in the i-th data, wherein when a is equal to 1, the difference between the first data and the last data is obtained; The difference between the a-th data and the previous data of the e-th first classification window with the greatest degree of fault manifestation in the h-th data is obtained, wherein when a is equal to 1, the difference between the first data and the last data is obtained; Indicates the total amount of data in the e-th first classification window with the greatest degree of fault manifestation.

[0162] in, It means adding the changes of all data in the e-th first classification window with the largest fault performance in the two data. The smaller the difference, the larger the combined training index of the data. Since the window size may be different, The smallest difference between the two windows can be taken for calculation.

[0163] Then, the two data with the largest combined training index are merged until there is no data with a combined training index greater than 0.5, thereby obtaining several types of data.

[0164] In the above embodiment, by analyzing the consistency of data changes within the classification window, the system can more accurately identify noise and fault characteristics in the data, thereby improving the accuracy of data classification. Based on similarity comparison and merged training index, the system can automatically determine which data can be merged into the same data type for training, reducing the need for manual intervention and improving training efficiency. The target fault diagnosis model generated after data merging can cover more diverse data characteristics, thereby enhancing the generalization ability of the model and improving its adaptability to unknown data. By merging similar data types, the system avoids multiple modeling of duplicate data and significantly reduces the consumption of computing resources.

[0165] In one embodiment of the present application, the to-be-trained running time series data corresponding to each of the data types are transmitted to the pre-built long short-term memory network model for training to obtain the target fault diagnosis model corresponding to each of the data types, including:

[0166] Extracting a training data set from the runtime data to be trained corresponding to each of the data types;

[0167] The training data set is transmitted to the pre-built long short-term memory network model, the possibility of the fault state is output, and the fault state with the greatest possibility of the fault state is taken as the fault state corresponding to the training data set, wherein the fault state includes a normal state and fault states corresponding to multiple fault types.

[0168] The training data set is a set of samples extracted from the training time series data corresponding to each data type, which is used for model training. After preprocessing (such as windowing, difference, etc.), these data can better reflect the changing characteristics of the equipment operation status.

[0169] Among them, the probability of fault state is a probability value output by the long short-term memory network (LSTM) model, which indicates the probability that the input data belongs to a certain fault state. For example, for a certain input data, the model may output the following results: the probability of normal state, 0.8; the probability of fault type A, 0.15; the probability of fault type B, 0.05.

[0170] Among them, the fault state refers to the state category that may occur during the operation of the equipment, including the normal state and the fault states corresponding to various fault types. The normal state means that the equipment is running normally without obvious abnormalities; the fault state corresponds to different fault modes of the equipment.

[0171] The fault states corresponding to the multiple fault types refer to various specific fault modes that may occur in the equipment. For example, a wind turbine may have the following fault types: bearing damage, gearbox wear, and generator overheating.

[0172] For example, see Figure 5 , Figure 5 A schematic diagram of the structure of a long short-term memory network provided by an embodiment of the present invention. Figure 5 As shown, the LSTM network contains three gating mechanisms (from left to right The processes represent forget gate, input gate, and output gate respectively). is the cell state, x is the input, y is the output, and tanh is the activation function. 70% of the data corresponding to the training runtime data to be trained is extracted as the training data set, and the training data set is input into the LSTM network model. The output is the possibility of each state of each data, and the state with the greatest possibility is recorded as the state of the data. The state includes normal and fault types.

[0173] In the above embodiment, through the possibility of fault status output by the LSTM model, the system can more accurately identify the operating status of the equipment and avoid misjudgment caused by a single threshold judgment. The model can not only distinguish between normal status and fault status, but also further subdivide them into multiple specific fault types, thereby providing more detailed diagnostic results.

[0174] In one embodiment of the present application, the analyzing of the multiple data types corresponding to the multiple types of runtime data to be diagnosed includes:

[0175] Performing windowing and differential processing on the plurality of to-be-diagnosed running time series data to obtain pre-processed to-be-diagnosed data;

[0176] For any of the data to be diagnosed, determine the consistency of data changes, analyze the data changes based on the consistency of data changes, determine the degree of fault manifestation, perform similarity comparison on the data in the window with the largest degree of fault manifestation in the two data to be diagnosed, and determine the combined training index;

[0177] The multiple types of data to be diagnosed are merged according to the size of the merged training index to determine the multiple types of data.

[0178] It should be noted that, in this embodiment, when analyzing multiple data types corresponding to multiple types of operating time series data to be diagnosed, the method used is consistent with the above-mentioned method for analyzing multiple data types corresponding to multiple types of operating time series data to be trained, that is, first pre-processing the multiple types of operating time series data to be diagnosed, and then determining the merged training index, and merging the multiple types of data to be diagnosed by merging the multiple types of data through the merged training index. Specifically, the steps of analyzing multiple types of data corresponding to multiple types of operating time series data to be diagnosed may be, for example:

[0179] Performing frequency domain conversion on a plurality of operation time series data to be diagnosed to obtain a plurality of operation frequency domain data to be diagnosed;

[0180] Analyze the periodic performance of various frequency domain data to be diagnosed, determine the periodic significance, determine periodic data and non-periodic data according to the periodic significance, perform windowing on the periodic data and non-periodic data, and perform differential operation on the periodic data and non-periodic data after windowing;

[0181] Normalizing the data after the differential operation to obtain the pre-processed data to be diagnosed;

[0182] For any pre-processed data to be diagnosed, a first classification window is obtained according to a preset second window size, the first classification window is moved, and a second classification window is obtained based on new data, and data changes in the first classification window and the second classification window are analyzed to determine the consistency of data changes, and based on the consistency of data changes, data changes in the first classification window are analyzed to determine the degree of fault manifestation;

[0183] The data in the window with the largest degree of fault manifestation in the two types of data to be diagnosed are similarly compared to determine the merged training index. According to the size of the merged training index, the pre-processed data to be diagnosed are merged to determine multiple data types.

[0184] Based on the above embodiments, it can be seen that the data acquisition unit acquires the operating data of multiple types of industrial equipment and can receive operating data from different types of industrial equipment, rather than being limited to a specific type of equipment, so that the system has a wide range of applicability and is suitable for various industrial scenarios such as wind turbines, port cranes, chemical equipment, and power transmission equipment; the model training unit analyzes multiple data types corresponding to multiple operating sequence data to be trained, and can identify and process different types of data (such as vibration data, temperature data, current data, etc.), so as to adapt to the data characteristics collected by different sensors, further enhance the versatility of the system, and avoid the need for separate modeling for each device; the model training unit will automatically analyze the data type and transmit the corresponding data to the initial fault diagnosis model for training, which reduces manual intervention, significantly improves the degree of automation of the fault diagnosis system, and reduces the dependence of small and medium-sized enterprises on a large number of people. By generating a corresponding target fault diagnosis model for each data type, it can quickly adapt to different equipment types and data characteristics without redesigning or adjusting the entire diagnosis process, thereby improving the diagnosis efficiency; because the system generates corresponding target fault diagnosis models according to different data types, each target fault diagnosis model can better capture the characteristics and failure modes of a specific type of data. Compared with the method of processing all data with a single model, this method can significantly improve the accuracy of the diagnosis results; the fault diagnosis unit will select the corresponding target fault diagnosis model according to the data type of the runtime data to be diagnosed, ensuring that the diagnosis process is always based on the model that best suits the data type, thereby avoiding misjudgment problems caused by the use of a unified standard; by integrating the operating data of multiple types of industrial equipment, the system can analyze the equipment status from multiple dimensions, further improving the reliability of the diagnosis results.

[0185] In one embodiment of the present application, an industrial equipment fault diagnosis method based on multi-type time series data analysis is also provided. Figure 6 A flowchart of an industrial equipment fault diagnosis method based on multi-type time series data analysis provided by an embodiment of the present invention is shown in FIG. Figure 6 , the method comprises the following steps:

[0186] S610, obtaining operation data of multiple types of industrial equipment, wherein the operation data includes multiple types of operation time series data to be trained and multiple types of operation time series data to be diagnosed;

[0187] S620, analyzing multiple data types corresponding to the multiple types of running time series data to be trained, and transmitting the running time series data to be trained corresponding to each type of data to a pre-built initial fault diagnosis model for training, to obtain a target fault diagnosis model corresponding to each type of data;

[0188] S630, analyzing multiple data types corresponding to the multiple types of operating sequence data to be diagnosed, and transmitting the operating sequence data to be diagnosed to the corresponding target fault diagnosis model according to the data types to obtain a fault diagnosis result.

[0189] Through the industrial equipment fault diagnosis method based on multi-category time series data analysis in the embodiment of the present application, the operating data of multiple types of industrial equipment are obtained, and the operating data from different types of industrial equipment can be received, rather than being limited to a specific type of equipment. It has wide applicability and is suitable for various industrial scenarios such as wind turbines, port cranes, chemical equipment, and power transmission equipment; multiple data types corresponding to multiple types of operating time series data to be trained can be analyzed, and different types of data (such as vibration data, temperature data, current data, etc.) can be identified and processed, so as to adapt to the data characteristics collected by different sensors, further enhance the versatility, and avoid the need for separate modeling for each device; automatically analyze the data types, and transmit the corresponding data to the initial fault diagnosis model for training, which reduces manual intervention, significantly improves the degree of automation, and reduces the dependence of small and medium-sized enterprises on a large number of people. By generating a corresponding target fault diagnosis model for each data type, it can quickly adapt to different equipment types and data characteristics without redesigning or adjusting the entire diagnosis process, thereby improving the diagnosis efficiency; according to different data types, corresponding target fault diagnosis models are generated, so each target fault diagnosis model can better capture the characteristics and failure modes of a specific type of data. Compared with the method of processing all data with a single model, this method can significantly improve the accuracy of the diagnosis results; according to the data type of the running sequence data to be diagnosed, the corresponding target fault diagnosis model is selected to ensure that the diagnosis process is always based on the model that best suits the data type, thereby avoiding misjudgment problems caused by the use of a unified standard; by integrating the operation data of multiple types of industrial equipment, the equipment status can be analyzed from multiple dimensions, further improving the reliability of the diagnosis results.

[0190] The specific embodiments of the industrial equipment fault diagnosis method based on multi-type time series data analysis in the present application can refer to the example shown in the above-mentioned industrial equipment fault diagnosis system based on multi-type time series data analysis, which will not be repeated here in this example.

[0191] It should be noted that the sequence of the above embodiments of the present invention is only for description and does not represent the advantages and disadvantages of the embodiments. The processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0192] The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referenced to each other, and each embodiment focuses on the differences from other embodiments.

Claims

1. An industrial equipment fault diagnosis system based on multi-class time series data analysis, characterized in that: The system comprises: A data acquisition unit, used to acquire operation data of various types of industrial equipment, wherein the operation data includes various operation time series data to be trained and various operation time series data to be diagnosed; A model training unit, used for analyzing a plurality of data types corresponding to the plurality of runtime data to be trained, and transmitting the runtime data to be trained corresponding to each of the data types to a pre-built initial fault diagnosis model for training, so as to obtain a target fault diagnosis model corresponding to each of the data types; The fault diagnosis unit is used to analyze multiple data types corresponding to the multiple types of operating sequence data to be diagnosed, and transmit the operating sequence data to be diagnosed to the corresponding target fault diagnosis model according to the data types to obtain a fault diagnosis result.

2. The industrial equipment fault diagnosis system based on multi-type time series data analysis according to claim 1, characterized in that: The system further comprises: The preprocessing unit is used to perform windowing and differential processing on the multiple types of runtime data to be trained to obtain preprocessed data to be trained, wherein the preprocessed data to be trained is used to be transmitted to the model training unit to analyze and obtain the multiple data types.

3. The industrial equipment fault diagnosis system based on multi-type time series data analysis according to claim 2, characterized in that: The pre-processing unit comprises: A frequency domain conversion subunit, used for performing frequency domain conversion on the plurality of to-be-trained running time series data to obtain a plurality of to-be-trained running frequency domain data; A differential operation subunit is used to analyze the periodic performance of the multiple types of frequency domain data to be trained, determine the periodic significance, determine the periodic data and the non-periodic data according to the periodic significance, perform windowing processing on the periodic data and the non-periodic data, and perform differential operation on the periodic data and the non-periodic data after the windowing processing; The normalization subunit is used to perform normalization processing on the data after the difference operation to obtain the pre-processed data to be trained.

4. The industrial equipment fault diagnosis system based on multi-type time series data analysis according to claim 3, characterized in that: The step of analyzing the periodic performance of the plurality of frequency domain data to be trained and determining the periodic significance includes: Obtaining extreme values ​​of the plurality of frequency domain data to be trained; Determining an envelope according to the extreme value; The periodic significance degree is determined according to the frequency domain distribution of the plurality of frequency domain data to be trained and the distance distribution of the extreme values ​​corresponding to the envelope.

5. The industrial equipment fault diagnosis system based on multi-type time series data analysis according to claim 4, characterized in that: The determining of periodic data and non-periodic data according to the periodic significance degree, and performing windowing processing on the periodic data and the non-periodic data, comprises: The data whose periodicity significance is greater than a preset periodicity significance threshold is taken as the periodic data, and the data whose periodicity significance is less than or equal to the preset periodicity significance threshold is taken as the non-periodic data; Taking the distance mean of all extreme values ​​on the envelope as a first window size, and performing windowing processing on the periodic data based on the first window size; The non-periodic data is clustered to obtain a plurality of clusters, the non-periodic data is divided into data corresponding to a plurality of states according to the distances between the data points of the plurality of clusters, and the data corresponding to each state is windowed.

6. The industrial equipment fault diagnosis system based on multi-type time series data analysis according to claim 5, characterized in that: Performing a differential operation on the non-periodic data after the windowing process, including: For multiple data segments in the data corresponding to any of the states, when forward differencing is performed on the data segments and the data volume of the data segments is greater than the data volume of the adjacent previous data segments, the data segments are divided into data segments with equal lengths to the adjacent previous data segments for differential operations, and the remaining parts are aligned for differential operations.

7. The industrial equipment fault diagnosis system based on multi-type time series data analysis according to claim 1, characterized in that: The initial fault diagnosis model includes a long short-term memory network model, and the model training unit includes: A data classification subunit, configured to analyze the noise manifestation and fault manifestation of the plurality of training running time series data, and determine the plurality of data types according to the analysis results; The model training subunit is used to transmit the to-be-trained running time sequence data corresponding to each of the data types to the pre-built long short-term memory network model for training, so as to obtain the target fault diagnosis model corresponding to each of the data types.

8. The industrial equipment fault diagnosis system based on multi-type time series data analysis according to claim 7, characterized in that: The analyzing the noise manifestation and fault manifestation of the plurality of training running time series data, and determining the plurality of data types according to the analysis results, includes: For any of the training running time series data, a first classification window is obtained according to a preset second window size, the first classification window is moved, a second classification window is obtained based on new data, and data changes in the first classification window and the second classification window are analyzed to determine data change consistency; Based on the consistency of the data changes, analyzing the data changes in the first classification window to determine the degree of fault manifestation; Perform similarity comparison on the data in the window with the largest degree of fault manifestation in the two types of running time series data to be trained, and determine a combined training index; The multiple types of running time series data to be trained are merged according to the size of the merged training index to determine the multiple types of data.

9. The industrial equipment fault diagnosis system based on multi-type time series data analysis according to claim 8, characterized in that: The step of transmitting the to-be-trained running time series data corresponding to each of the data types to the pre-built long short-term memory network model for training to obtain the target fault diagnosis model corresponding to each of the data types includes: Extracting a training data set from the runtime data to be trained corresponding to each of the data types; The training data set is transmitted to the pre-built long short-term memory network model, the possibility of the fault state is output, and the fault state with the greatest possibility of the fault state is taken as the fault state corresponding to the training data set, wherein the fault state includes a normal state and fault states corresponding to multiple fault types.

10. The industrial equipment fault diagnosis system based on multi-type time series data analysis according to claim 8, characterized in that: The analyzing of the multiple data types corresponding to the multiple types of running time series data to be diagnosed includes: Performing windowing and differential processing on the plurality of to-be-diagnosed running time series data to obtain pre-processed to-be-diagnosed data; For any of the data to be diagnosed, determine the consistency of data changes, analyze the data changes based on the consistency of data changes, determine the degree of fault manifestation, perform similarity comparison on the data in the window with the largest degree of fault manifestation in the two data to be diagnosed, and determine the combined training index; The multiple types of data to be diagnosed are merged according to the size of the merged training index to determine the multiple types of data.

Images