Method and system for device health assessment based on population data and autoencoder

Abstract

The application provides a device health evaluation method and system based on population data and an autoencoder, comprising: obtaining running state signals of multiple same devices under the same working condition; preprocessing all signals, calculating the difference between the signals and generating a difference matrix, and splicing the difference matrix of multiple time points and multiple time spans into a tensor; using the tensor spliced from the difference matrix as training data to train a multi-scale autoencoder to minimize the reconstruction error and generate a machine learning model; inputting the processed device running data into the multi-scale autoencoder, reflecting the degree of data deviation from the normal state through the reconstruction error, and evaluating the device health condition according to the data anomaly degree. The application judges the anomaly of individual data through the difference between population data, thereby reflecting the health condition of the device, and avoiding the acquisition of high-frequency signals and the control of complex working conditions.

Description

Technical Field

[0001] This invention relates to the field of equipment anomaly detection technology, and more specifically, to a method and system for equipment health assessment based on population data and autoencoders. Background Technology

[0002] Equipment performance gradually declines over time, so predicting equipment health status is essential for preventing equipment failures and improving equipment reliability.

[0003] Currently, repairs are typically only dispatched after equipment malfunctions. However, because the type of malfunction cannot be determined in advance, repairs are time-consuming and inefficient, and the equipment becomes unusable after a malfunction, resulting in significant losses during the repair process.

[0004] The main methods for equipment health assessment currently include the health factor method, the historical data method, and the degradation model method. However, health factors vary for different equipment, and their construction is closely related to operating conditions, equipment model, and other information. Currently, most equipment lacks widely applicable health factors to describe its health status.

[0005] Historical data analysis is suitable for equipment of the same type, operating conditions, and model. Given a large amount of relevant historical data, the most similar historical equipment curve can be selected from the historical database based on the current equipment's operating status and its historical status to obtain the current equipment's health assessment curve. However, this method requires a large amount of data and is difficult to adapt to changes in the equipment's operating environment. Furthermore, many devices lack sufficient historical data, and in actual production processes, the amount of fault data is relatively rare.

[0006] The degradation model method, based on a large amount of historical data, analyzes and fuses the time-frequency domain characteristics of various signals to form a degradation index that can represent the changes in system performance. This degradation index is used as a health factor for evaluating the health status of the equipment. Its characteristics are that it is not always possible to find a suitable degradation index and it requires certain data support, but its applicability is wider than simply using health factors.

[0007] Existing equipment evaluation methods have many limitations. For example, they require high-quality raw signals, including high sampling frequencies and signal-to-noise ratios, which are difficult to achieve under complex operating conditions. The diagnostic results of existing methods are all verified in laboratory environments, resulting in significant discrepancies with conclusions drawn from field data. Furthermore, the algorithms cannot cover most potential operating conditions. For instance, for motors operating in noisy environments for extended periods, the vibration signal has a low signal-to-noise ratio, making analysis extremely difficult. Moreover, due to the harsh working environment, it is difficult to install appropriate sensors in suitable locations. Summary of the Invention

[0008] To address the shortcomings of existing technologies, the present invention aims to provide a method and system for device health assessment based on population data and autoencoders.

[0009] A device health assessment method based on population data and an autoencoder, provided by the present invention, includes:

[0010] Data acquisition steps: Acquire the operating status signals of multiple identical devices under the same operating conditions;

[0011] Data discrepancy analysis steps: preprocess all signals, calculate the differences between signals and generate a discrepancy matrix, and concatenate the discrepancy matrices from multiple time points and time spans into a tensor;

[0012] Model training steps: Use the tensor formed by concatenating the difference matrices as training data to train a multi-scale autoencoder to minimize the reconstruction error and generate a machine learning model;

[0013] Health assessment steps: Input the processed equipment operation data into a multi-scale self-encoder. The reconstruction error reflects the degree of data deviation from the normal state. The health status of the equipment is assessed based on the degree of data anomaly.

[0014] Preferably, the operating status signal includes the device's basic control signals, feedback signals, and acquisition signals.

[0015] Preferably, the preprocessing of the operating status signals includes: performing outlier processing on all signals, and then normalizing all signals.

[0016] Preferably, the process of calculating the difference between the information is as follows: calculate the difference between each pair of signals, and then calculate the root mean square of the difference.

[0017] Preferably, the multi-scale autoencoder includes an encoder, a decoder, and feature fusion. The tensor data is processed by the encoder to extract temporal and convolutional features to generate a feature vector. The feature vector is fused with the data in the encoding process and then processed by the decoder to generate reconstructed output data. The training is performed with the minimum reconstruction error as the loss function.

[0018] Preferably, the encoder and decoder employ multi-layer convolutional neural networks; the feature fusion employs temporal neural networks and recurrent long short-term memory neural networks.

[0019] Preferably, in the health assessment step, the degree of data abnormality is judged by setting a threshold or using a convolutional neural network.

[0020] According to the present invention, a device health assessment system based on population data and self-encoding includes:

[0021] Data acquisition module: Acquires operating status signals from multiple identical devices under the same operating conditions;

[0022] Data difference analysis module: preprocesses all signals, calculates the differences between signals and generates a difference matrix, and concatenates the difference matrices from multiple time points and time spans into a tensor;

[0023] Model training module: Using tensors formed by concatenating dissimilarity matrices as training data, a multi-scale autoencoder is trained to minimize reconstruction error and generate a machine learning model;

[0024] Health assessment module: The processed equipment operation data is input into the multi-scale self-encoder. The reconstruction error reflects the degree of data deviation from the normal state, and the health status of the equipment is assessed based on the degree of data anomaly.

[0025] According to the present invention, a computer-readable storage medium storing a computer program is provided, which, when executed by a processor, implements the steps of the above-described device health assessment method based on population data and autoencoders.

[0026] According to the present invention, an apparatus includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When executed by the processor, the computer program implements the steps of the above-described device health assessment method based on population data and autoencoders.

[0027] Compared with the prior art, the present invention has the following beneficial effects:

[0028] 1. This invention judges the anomalies of individual data by the differences between group data, thereby reflecting the health status of the equipment and avoiding the acquisition of high-frequency signals and control of complex operating conditions.

[0029] 2. This invention employs a multi-scale autoencoder algorithm. Under the condition of missing abnormal fault data in actual use, it completes the abnormal detection and health assessment of the equipment based on historical health data, and has a high accuracy rate. Attached Figure Description

[0030] Other features, objects, and advantages of the present invention will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings:

[0031] Figure 1 This is a flowchart illustrating the construction method of the difference matrix in an embodiment of the present invention;

[0032] Figure 2 This is a model framework diagram of the multi-scale autoencoder in an embodiment of the present invention;

[0033] Figure 3 This is a diagram showing the anomaly detection results in actual testing during an embodiment of the present invention;

[0034] Figure 4 This is a performance variation curve of the main drive motor set in an embodiment of the present invention. Detailed Implementation

[0035] The present invention will now be described in detail with reference to specific embodiments. These embodiments will help those skilled in the art to further understand the present invention, but do not limit the invention in any way. It should be noted that those skilled in the art can make several changes and improvements without departing from the concept of the present invention. These all fall within the protection scope of the present invention.

[0036] This invention proposes a device health assessment method based on population data and autoencoders, comprising:

[0037] Data acquisition steps: Acquire operating status signals from multiple identical devices under the same working conditions. These can be multiple components within the same device performing the same function and operating condition, such as the main drive motors in a tunnel boring machine, which are typically 8-14 in number. Operating status signals can include basic control signals, feedback signals, acquisition signals, and other circuit signals related to the device's operating status.

[0038] In one specific implementation, the main drive motor current signal in good operating condition is collected by the PLC, with a sampling frequency of 1Hz.

[0039] In another specific implementation, the main drive motor speed signal in good operating condition is collected by PLC, with a sampling frequency of 1Hz.

[0040] In other specific implementations, the sampling frequency is not limited to 1 Hz.

[0041] Data discrepancy analysis steps: Refer to Figure 1 As shown, all signals are preprocessed, and outlier handling is performed on all signals, such as removing data segments with all zeros, deleting obviously abnormal data, and filtering working status data.

[0042] Then, all signals are normalized to ensure that the order of magnitude of the results in subsequent calculations will not be too large for signals with different amplitudes under different operating conditions.

[0043] The differences between each pair of signals are calculated, and a difference matrix is ​​generated. For n signals, the difference needs to be calculated n(n-1) / 2 times. The difference can be calculated by taking the root mean square of the difference signal, the absolute mean of the difference signal, the inner product of the difference signal and the original signal, or the inner product of the original signals, etc. The dimension of the difference matrix is ​​the number of devices. Finally, the difference matrices from multiple time points and time spans are concatenated into a tensor.

[0044] Model training steps: Using a tensor formed by concatenating dissimilarity matrices as training data, a multi-scale autoencoder is trained to minimize reconstruction error, generating a machine learning model. The machine learning model is as follows: Figure 2 As shown, the model includes an encoder, a decoder, and feature fusion. Tensor data is processed by the encoder to extract temporal and convolutional features, generating feature vectors. These feature vectors are then fused with data from the encoding process and processed by the decoder to generate reconstructed output data. The model is trained using the minimization of reconstruction error as the loss function, and the training data is derived from health data of the device during normal operation. The encoder and decoder are structured as multi-layer convolutional neural networks, while feature fusion utilizes a temporal neural network and a recurrent long short-term memory neural network.

[0045] Health assessment steps: The tensor generated from the processed equipment operating data is input into a multi-scale autoencoder. The reconstruction error reflects the degree to which the data deviates from the normal state. An example of anomaly results is shown in the figure below. Figure 3 As shown, a threshold is set or a simple convolutional neural network is used to determine the degree of anomaly at a given moment, assessing the health of the equipment and issuing timely warnings based on the assessment results. The equipment's performance status change curve is shown in the figure. Figure 4 As shown.

[0046] The present invention also provides a device health assessment system based on population data and self-encoding. The device health assessment system based on population data and self-encoding can be implemented by executing the process steps of the device health assessment method based on population data and self-encoding. That is, those skilled in the art can understand the device health assessment method based on population data and self-encoding as a preferred embodiment of the device health assessment system based on population data and self-encoding.

[0047] The present invention also provides a device health assessment system based on population data and self-encoding, comprising:

[0048] Data acquisition module: Acquires operating status signals from multiple identical devices under the same operating conditions;

[0049] Data difference analysis module: preprocesses all signals, calculates the differences between signals and generates a difference matrix, and concatenates the difference matrices from multiple time points and time spans into a tensor;

[0050] Model training module: Using tensors formed by concatenating dissimilarity matrices as training data, a multi-scale autoencoder is trained to minimize reconstruction error and generate a machine learning model;

[0051] Health assessment module: The processed equipment operation data is input into the multi-scale self-encoder. The reconstruction error reflects the degree of data deviation from the normal state, and the health status of the equipment is assessed based on the degree of data anomaly.

[0052] The present invention also provides a computer-readable storage medium storing a computer program, which may be a portable hard drive, a USB flash drive, etc., wherein the computer program, when executed by a processor, implements the steps of the above-described device health assessment method based on population data and autoencoder.

[0053] The present invention also provides an apparatus including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the computer program, when executed by the processor, implements the steps of the above-described device health assessment method based on population data and autoencoders.

[0054] Those skilled in the art will understand that, besides implementing the system and its various devices, modules, and units provided by this invention in the form of purely computer-readable program code, the same functions can be achieved entirely through logical programming of the method steps, making the system and its various devices, modules, and units of this invention function in the form of logic gates, switches, application-specific integrated circuits, programmable logic controllers, and embedded microcontrollers. Therefore, the system and its various devices, modules, and units provided by this invention can be considered as a hardware component, and the devices, modules, and units included therein for implementing various functions can also be considered as structures within the hardware component; alternatively, the devices, modules, and units for implementing various functions can be considered as both software modules implementing the method and structures within the hardware component.

[0055] In the description of this application, it should be understood that the terms "upper", "lower", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this application and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this application.

[0056] Specific embodiments of the present invention have been described above. It should be understood that the present invention is not limited to the specific embodiments described above, and those skilled in the art can make various changes or modifications within the scope of the claims, which do not affect the essence of the present invention. Unless otherwise specified, the embodiments and features described in this application can be arbitrarily combined with each other.

Claims

1. A device health assessment method based on population data and autoencoders, characterized in that, include: Data acquisition steps: Acquire the operating status signals of multiple identical devices under the same operating conditions; The operating status signals include the equipment's basic control signals, feedback signals, and acquisition signals; Data discrepancy analysis steps: preprocess all signals, calculate the differences between signals and generate a discrepancy matrix, and concatenate the discrepancy matrices from multiple time points and time spans into a tensor; The process of calculating the difference between the signals is as follows: calculate the difference between each pair of signals, and then calculate the root mean square of the difference. Model training steps: Use the tensor formed by concatenating the difference matrices as training data to train a multi-scale autoencoder to minimize the reconstruction error and generate a machine learning model; Health assessment steps: Input the processed equipment operating data into a multi-scale self-encoder, use the reconstruction error to reflect the degree of data deviation from the normal state, and assess the health status of the equipment based on the degree of data anomaly; The multi-scale autoencoder includes an encoder, a decoder, and feature fusion. Tensor data is processed by the encoder to extract temporal and convolutional features. Feature vectors are generated based on the temporal and convolutional features. The feature vectors are fused with the temporal features from the encoding process. The fused features are then input into the decoder to generate reconstructed output data. Training is performed with the minimum reconstruction error as the loss function.

2. The device health assessment method based on population data and autoencoder according to claim 1, characterized in that, The preprocessing of the operating status signals includes: handling outliers for all signals, and then normalizing all signals.

3. The device health assessment method based on population data and self-encoding according to claim 1, characterized in that: The encoder and decoder employ a multi-layer convolutional neural network.

4. The device health assessment method based on population data and self-encoding according to claim 1, characterized in that: In the health assessment step, the degree of abnormality of the data is judged by setting a threshold or using a convolutional neural network.

5. A system for implementing the device health assessment method based on population data and autoencoder as described in claim 1, characterized in that, include: Data acquisition module: Acquires operating status signals from multiple identical devices under the same operating conditions; Data difference analysis module: preprocesses all signals, calculates the differences between signals and generates a difference matrix, and concatenates the difference matrices from multiple time points and time spans into a tensor; Model training module: Using tensors formed by concatenating dissimilarity matrices as training data, a multi-scale autoencoder is trained to minimize reconstruction error and generate a machine learning model; Health assessment module: The processed equipment operation data is input into the multi-scale self-encoder. The reconstruction error reflects the degree of data deviation from the normal state, and the health status of the equipment is assessed based on the degree of data anomaly.

6. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the steps of the device health assessment method based on population data and autoencoder as described in any one of claims 1 to 4.

7. An apparatus comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the computer program is executed by the processor, it implements the steps of the device health assessment method based on population data and autoencoder as described in any one of claims 1 to 4.

Images