Distributed sensor fault processing method and device, electronic equipment and storage medium

Abstract

The application provides a distributed sensor fault processing method and device, electronic equipment and storage medium, comprising: acquiring sensor multi-source data collected by a distributed sensor in a preset time interval; determining a fault sensor from the distributed sensor based on the sensor multi-source data through a pre-trained fault positioning model, and determining a sensor location identifier and a sensor type identifier of the fault sensor; predicting reconstruction data corresponding to the fault sensor based on the sensor multi-source data, the sensor location identifier and the sensor type identifier through a pre-trained fault data reconstruction model. The application realizes closed-loop fault diagnosis of the distributed sensor, and has better generalization ability and accuracy for fault diagnosis of the distributed sensor.

Description

Technical Field

[0001] This invention relates to the field of aircraft technology, and in particular to a method, apparatus, electronic device, and storage medium for handling distributed sensor faults. Background Technology

[0002] With the advancement of science and technology in the field of aircraft research, aircraft acquire useful information by configuring multiple IMU (Inertial Measurement Unit) sensors on both wings. Compared to a single IMU sensor, a distributed IMU sensor not only acquires the effective information measured by each IMU sensor, but the information measured by the distributed IMU sensors can also supervise each other, thereby generating more accurate measurement information.

[0003] IMU sensors are electronic components, and they can be damaged or malfunction due to vibration and frequent use on aircraft wings. Existing sensor fault diagnosis methods often focus on fault location and isolation, without forming a closed-loop fault diagnosis solution, which can potentially affect subsequent data processing and results. Summary of the Invention

[0004] In view of this, the purpose of the present invention is to provide a method, apparatus, electronic device and storage medium for handling distributed sensor faults, which realizes closed-loop fault diagnosis of distributed sensors, and has better generalization ability and accuracy for fault diagnosis of distributed sensors.

[0005] In a first aspect, embodiments of the present invention provide a distributed sensor fault handling method, including:

[0006] Acquire multi-source sensor data collected by distributed sensors within a preset time interval;

[0007] Using a pre-trained fault location model, the faulty sensor is identified from the distributed sensors based on the multi-source data of the sensors, and the sensor location identifier and sensor type identifier of the faulty sensor are also determined.

[0008] Using a pre-trained fault data reconstruction model, the reconstructed data corresponding to the faulty sensor is predicted based on the multi-source data of the sensor, the sensor location identifier, and the sensor type identifier.

[0009] In one embodiment, the fault location model includes a first encoder and a first decoder;

[0010] Using a pre-trained fault location model, the system identifies faulty sensors from the distributed sensors based on multi-source sensor data, and determines the sensor location and sensor type identifiers of the faulty sensors, including:

[0011] The first high-dimensional feature of the multi-source data from the sensor is extracted using the first encoder;

[0012] The first decoder determines the failure probability of each sensor in the distributed sensors based on the first high-dimensional feature, identifies the faulty sensor from the distributed sensors based on the failure probability, and determines the sensor location identifier and sensor type identifier of the faulty sensor.

[0013] In one embodiment, the multi-source sensor data includes acceleration data and angular velocity data collected by each sensor in the distributed sensor system;

[0014] Using a pre-trained fault data reconstruction model, based on the multi-source data from the sensor, the sensor location identifier, and the sensor type identifier, the reconstructed data corresponding to the faulty sensor is predicted, including:

[0015] Based on the sensor type identifier, the measurement data collected by the target sensor corresponding to the sensor type identifier is extracted from the multi-source data of the sensor;

[0016] Furthermore, based on the sensor location identifier, the measurement data collected by the faulty sensor is extracted from the multi-source data of the sensor, and the measurement data collected by the faulty sensor is set to zero;

[0017] By using a pre-trained fault data reconstruction model, based on the measurement data collected by the target sensor and the measurement data collected by the fault sensor after being zeroed, the reconstructed data corresponding to the fault sensor is predicted.

[0018] In one implementation, the sensor type identifier includes an accelerometer identifier or a gyroscope identifier;

[0019] Based on the sensor type identifier, the measurement data collected by the target sensor corresponding to the sensor type identifier is extracted from the multi-source sensor data, including:

[0020] If the sensor type is identified as an accelerometer, then the accelerometer sensor in the distributed sensor is used as the target sensor, and the measurement data collected by the accelerometer sensor is extracted from the multi-source data of the sensor.

[0021] Alternatively, if the sensor type is identified as a gyroscope, then the gyroscope sensor in the distributed sensor is used as the target sensor, and the measurement data collected by the gyroscope sensor is extracted from the multi-source data of the sensor.

[0022] In one embodiment, the fault data reconstruction model includes a second encoder and a second decoder;

[0023] Using a pre-trained fault data reconstruction model, based on the measurement data collected by the target sensor and the measurement data collected by the fault sensor after being zeroed, the model predicts the reconstructed data corresponding to the fault sensor, including:

[0024] The second encoder extracts the second high-dimensional features of the measurement data collected by the target sensor and the measurement data collected by the fault sensor after being zeroed.

[0025] The second decoder predicts the reconstructed data corresponding to the fault sensor based on the second high-dimensional feature.

[0026] In one embodiment, the first encoder includes a first CNN layer, a first Transformer layer, a second CNN layer, a first pooling layer, and a third CNN layer connected in sequence, and the first decoder includes a flattening layer, a first fully connected layer, a first activation layer, a second fully connected layer, and a normalization layer connected in sequence.

[0027] The second encoder comprises a fourth CNN layer, a second Transformer layer, a fifth CNN layer, a second pooling layer, and a sixth CNN layer connected in sequence, and the second decoder comprises a third fully connected layer, a second activation layer, and a fourth fully connected layer connected in sequence.

[0028] In one implementation, after predicting the reconstructed data corresponding to the faulty sensor based on the sensor multi-source data, the sensor location identifier, and the sensor type identifier using a pre-trained fault data reconstruction model, the method further includes:

[0029] The reconstructed data corresponding to the faulty sensor is used to replace the measurement data collected by the faulty sensor in the multi-source sensor data.

[0030] The replaced sensor multi-source data is sent to a designated processing device for data processing.

[0031] Secondly, embodiments of the present invention also provide a distributed sensor fault handling device, comprising:

[0032] The data acquisition module is used to acquire multi-source sensor data collected by distributed sensors within a preset time interval;

[0033] The fault location module is used to determine the faulty sensor from the distributed sensors based on the multi-source data of the sensors using a pre-trained fault location model, and to determine the sensor location identifier and sensor type identifier of the faulty sensor.

[0034] The data reconstruction module is used to predict the reconstructed data corresponding to the faulty sensor based on the multi-source data of the sensor, the sensor location identifier, and the sensor type identifier through a pre-trained fault data reconstruction model.

[0035] Thirdly, embodiments of the present invention also provide an electronic device, including a processor and a memory, the memory storing computer-executable instructions executable by the processor, the processor executing the computer-executable instructions to implement the method described in any of the first aspects.

[0036] Fourthly, embodiments of the present invention also provide a computer-readable storage medium storing computer-executable instructions, which, when invoked and executed by a processor, cause the processor to implement the method described in any of the first aspects.

[0037] This invention provides a distributed sensor fault handling method, apparatus, electronic device, and storage medium. First, it acquires multi-source sensor data collected by distributed sensors within a preset time interval. Then, using a pre-trained fault location model, it identifies the faulty sensor from the distributed sensors based on the multi-source sensor data, as well as the sensor location and type identifiers of the faulty sensor. Finally, using a pre-trained fault data reconstruction model, it predicts the reconstructed data corresponding to the faulty sensor based on the multi-source sensor data, sensor location identifier, and sensor type identifier. After acquiring the multi-source sensor data within the preset time interval, the above method can use the fault location model to locate the faulty sensor and its sensor location and type identifiers from the distributed sensors. Then, using the fault data reconstruction model, combined with the multi-source sensor data, sensor location identifier, and sensor type identifier, it reconstructs the measurement data of the faulty sensor to obtain the corresponding reconstructed data, thereby achieving closed-loop fault diagnosis of distributed sensors. Furthermore, this invention does not require accurate mathematical modeling of the distributed sensors, and has better generalization ability and accuracy for fault diagnosis of distributed sensors.

[0038] Other features and advantages of the invention will be set forth in the description which follows, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention are realized and obtained in accordance with the structures particularly pointed out in the description, claims and drawings.

[0039] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, preferred embodiments are described below in detail with reference to the accompanying drawings. Attached Figure Description

[0040] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0041] Figure 1 A flowchart illustrating a distributed sensor fault handling method provided in an embodiment of the present invention;

[0042] Figure 2 This invention provides a schematic diagram of the location distribution of distributed sensors on an aircraft.

[0043] Figure 3 This is a schematic diagram of multi-source sensor data provided in an embodiment of the present invention;

[0044] Figure 4 This is a schematic diagram of the structure of a fault location model provided in an embodiment of the present invention;

[0045] Figure 5 This is a schematic diagram of a data processing procedure for fault isolation and reconstruction provided in an embodiment of the present invention;

[0046] Figure 6 This is a schematic diagram of the structure of a fault data reconstruction model provided in an embodiment of the present invention;

[0047] Figure 7 This is a schematic diagram of the structure of a distributed sensor fault handling device provided in an embodiment of the present invention;

[0048] Figure 8 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention. Detailed Implementation

[0049] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the present invention will be clearly and completely described below in conjunction with the embodiments. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0050] IMU sensors are electronic components, and they can be damaged or malfunction due to vibration and frequent use on aircraft wings. Therefore, an effective method for IMU sensor fault diagnosis is needed to detect faults promptly and prevent erroneous measurements from being used in calculations. Existing methods typically focus on diagnosing single IMU sensors, with limited research on fault diagnosis methods for distributed IMU sensors. Compared to diagnosing single IMU sensor faults, which only requires determining whether an IMU has malfunctioned, diagnosing distributed IMU sensor faults is more complex. It requires locating faults in multiple IMU sensors. After determining the location of the faulty IMU sensor, data reconstruction is necessary because distributed IMU sensor measurement data often participates in data fusion processes to improve the accuracy of the measurement information.

[0051] Existing IMU sensor fault diagnosis schemes have at least the following problems: (1) It is difficult to extract fault features for distributed IMU sensors; (2) They are limited to fault location and fault isolation and have not formed a closed-loop fault diagnosis scheme. There is little research on the reconstruction and recovery of measurement data after a sensor fault occurs; (3) Accurate mathematical modeling of distributed IMU sensors leads to weak generalization ability and poor accuracy in fault diagnosis of distributed IMU sensors.

[0052] Based on this, the present invention provides a method, apparatus, electronic device and storage medium for handling distributed sensor faults, realizing closed-loop fault diagnosis of distributed sensors, and having better generalization ability and accuracy for fault diagnosis of distributed sensors.

[0053] To facilitate understanding of this embodiment, a detailed description of a distributed sensor fault handling method disclosed in this embodiment of the invention will be provided first, see [link to relevant documentation]. Figure 1 The diagram shows a flowchart of a distributed sensor fault handling method, which mainly includes the following steps S102 to S106:

[0054] Step S102: Acquire multi-source sensor data collected by the distributed sensors within a preset time interval. The distributed sensors, which can be installed on the aircraft, are also known as distributed IMU sensors and may include several IMUs. The multi-source sensor data includes acceleration and angular velocity data collected by each IMU in the distributed sensors.

[0055] In one implementation, acceleration and angular velocity data collected by each IMU can be acquired by a sliding time window, wherein the length of the sliding time window is Δt(s) and the step size is Δt.

[0056] Step S104 involves using a pre-trained fault location model to identify the faulty sensor from distributed sensors based on multi-source sensor data, and to determine the sensor location identifier and sensor type identifier of the faulty sensor. The sensor type identifier includes an accelerometer identifier or a gyroscope identifier. The fault location model can employ an improved Transformer-CNN (Convolutional Neural Networks) network. In one embodiment, the input to the fault location model is multi-source sensor data, and the output is the probability of sensor failure, as well as the sensor location identifier and sensor type identifier; this sensor is the faulty sensor.

[0057] Step S106: Using a pre-trained fault data reconstruction model, the reconstructed data corresponding to the faulty sensor is predicted based on multi-source sensor data, sensor location identifiers, and sensor type identifiers. The fault data reconstruction model is another improved Transformer-CNN network. In one implementation, the input data for the fault data reconstruction model can be constructed based on the sensor location identifiers and sensor type identifiers. This input data includes measurement data collected by each IMU corresponding to the sensor type identifier, and measurement data collected by the faulty sensor, wherein the measurement data collected by the faulty sensor has been set to 0. The output of the fault data reconstruction model is the reconstructed data corresponding to the faulty sensor.

[0058] The distributed sensor fault handling method provided in this invention, after acquiring multi-source sensor data within a preset time interval, can use a fault location model to locate the faulty sensor, its sensor location identifier, and sensor type identifier from the distributed sensors. Then, using a fault data reconstruction model, combined with the multi-source sensor data, sensor location identifier, and sensor type identifier, the measurement data of the faulty sensor is reconstructed to obtain the corresponding reconstructed data, thereby realizing closed-loop fault diagnosis of distributed sensors. In addition, this invention does not require accurate mathematical modeling of distributed sensors, and has better generalization ability and accuracy for fault diagnosis of distributed sensors.

[0059] The purpose of this invention is to provide a method for fault location and isolation reconstruction of distributed sensors in aircraft, which realizes the fault location of distributed sensors and the reconstruction and recovery of measurement data after a sensor failure. This invention improves the reliability of distributed sensors in aircraft through a closed-loop fault diagnosis scheme.

[0060] To facilitate understanding of the foregoing embodiments, this invention provides a specific implementation of a distributed sensor fault handling method. First, see... Figure 2The diagram shows the location distribution of distributed sensors on an aircraft. The main IMU is installed at the center of mass, and the sub-IMUs are evenly distributed along both sides of the wing. The output frequency of each IMU is f (Hz).

[0061] exist Figure 1 Based on this, this embodiment of the invention provides an implementation method for step S102, which uses a sliding time window with a length of Δt(s) and a step size of Δt to collect measurement data (i.e., multi-source sensor data) output by n IMUs and use it for fault diagnosis. The measurement data of each IMU includes six types of data: acceleration measured by a 3-axis accelerometer and angular velocity measured by a gyroscope. Therefore, the shape of the measurement data of each IMU collected by the sliding time window is (f×Δt)×6, a 2D data structure. Thus, the shape of all the measurement data of the n IMUs collected by the sliding time window is n×(f×Δt)×6, a 3D data structure, such as... Figure 3 The diagram shows a multi-source sensor data set, which is used as input data for a fault location model to detect the location of faulty sensors.

[0062] The above-mentioned measurement data acquisition scheme is designed for distributed sensors. The multi-source sensor data obtained through this scheme is more easily subjected to feature extraction and feature fusion using intelligent methods.

[0063] Furthermore, this embodiment of the invention provides an implementation method for step S104, designing a fault feature extraction encoder based on an improved Transformer-CNN to achieve fault localization. By improving the Transformer and CNN networks, it is easier to extract the temporal and spatial nonlinear features of the aircraft's distributed sensor data. In a specific implementation, see... Figure 4 The diagram shows a structural schematic of a fault location model, which includes a first encoder (also known as a fault location encoder) and a first decoder (also known as a fault location decoder).

[0064] Specifically, the first encoder includes a first CNN layer, a first Transformer layer, a second CNN layer, a first pooling layer, and a third CNN layer connected in sequence, and the first decoder includes a flattening layer, a first fully connected layer, a first activation layer, a second fully connected layer, and a normalization layer connected in sequence.

[0065] exist Figure 4 Based on this, embodiments of the present invention provide a specific implementation method for determining faulty sensors from distributed sensors based on multi-source sensor data using a pre-trained fault location model, and for determining the sensor location identifier and sensor type identifier of the faulty sensor, as detailed in steps a to b below:

[0066] Step a: Extract the first high-dimensional features from the multi-source sensor data using the first encoder. In one implementation, nonlinear features are obtained through a CNN layer, spatiotemporal correlation features are obtained through a Transformer layer, and fault location features of the distributed sensor are obtained by fusing the multi-source data. Pooling layers are added between the CNN layers to further reduce the dimensionality of the extracted fault features and decrease computational cost. In the first CNN layer, positional encoding is added to the original features to add a vector representing the positional order, enhancing the correlation between features in subsequent Transformer layers. Positional encoding can be understood as using non-repeating values ​​between 0 and 1 obtained from sine and cosine functions to distinguish each nonlinear feature.

[0067] In the specific implementation, the first CNN layer extracts features from the multi-source data of the sensor to obtain nonlinear features; the first Transformer layer extracts features from the nonlinear features output by the first CNN layer to obtain spatiotemporal correlation features; the second CNN layer extracts features from the spatiotemporal correlation features to obtain nonlinear features; the first pooling layer reduces the dimensionality of the nonlinear features output by the second CNN layer; and the third CNN layer extracts features from the dimensionality-reduced nonlinear features to obtain the first high-dimensional features.

[0068] Step b involves determining the failure probability of each sensor in the distributed sensor array based on the first high-dimensional features using a first decoder, identifying the faulty sensor from the distributed sensor array based on the failure probability, and determining the sensor location identifier and sensor type identifier of the faulty sensor. In one embodiment, the first decoder can be designed based on a multilayer perceptron. The first decoder decodes and reduces the dimensionality of the first high-dimensional features to obtain the location identifier and category identifier of the faulty sensor. In the first decoder, a flattening layer merges the feature dimensions, a fully connected layer performs the feature decoding process, an activation layer increases the nonlinear correlation of features, and finally, a normalization layer obtains the probability of the faulty sensor location, thus decoding the fault location number and sensor type.

[0069] For example, if the failure probability of a certain sensor is greater than a preset threshold, the sensor can be identified as a faulty sensor. At this time, the sensor location identifier and sensor type identifier of the faulty sensor can be output. The sensor location identifier can be used to mark the position of the measurement data collected by the faulty sensor in the multi-source data of the sensor, and the sensor type identifier can be used to identify the category of the faulty sensor, such as whether the faulty sensor is an accelerometer or a gyroscope.

[0070] In one implementation, the fault location model can be pre-trained, wherein the training data can be measurement data collected by multiple IMUs. The training data carries labels, which are used to mark the faulty IMU and its fault probability, sensor location identifier and sensor type identifier, so that the fault location model learns the mapping relationship between the training data and the fault probability, sensor location identifier and sensor type identifier. When the loss of the fault location model converges, the training can be stopped.

[0071] Furthermore, this embodiment of the invention provides an implementation method for step S106, which can be specifically described in steps 1 to 3 below:

[0072] Step 1: Based on the sensor type identifier, extract the measurement data collected by the target sensor corresponding to the sensor type identifier from the multi-source sensor data.

[0073] In one implementation, see Figure 5 The diagram illustrates a data processing procedure for fault isolation and reconstruction. Based on the fault sensor location number and sensor type (accelerometer or gyroscope) obtained from the fault location model, the corresponding triaxial data (i.e., the aforementioned measurement data) of all IMUs can be selected. Specifically: (1) If the sensor type is identified as an accelerometer, the accelerometer sensor in the distributed sensor is used as the target sensor, and the measurement data collected by the accelerometer sensor is extracted from the multi-source sensor data; (2) If the sensor type is identified as a gyroscope, the gyroscope sensor in the distributed sensor is used as the target sensor, and the measurement data collected by the gyroscope sensor is extracted from the multi-source sensor data. For example, assuming that IMU1, IMU2, and IMU3 are accelerometers in the distributed sensor, and IMU4, IMU5, and IMU6 are gyroscopes, and IMU4 is the fault sensor, that is, the sensor type is identified as a gyroscope, then the measurement data collected by IMU5 and IMU6 are extracted from the multi-source sensor data.

[0074] Step 2: Based on the sensor location identifier, extract the measurement data collected by the faulty sensor from the multi-source sensor data, and set the measurement data collected by the faulty sensor to zero. In one implementation, please refer to... Figure 5 The measurement data at the fault location is set to 0 so that the normal data at other locations can be extracted and fused using the fault data reconstruction model, thereby restoring the normal measurement data of the faulty sensor (i.e., reconstructing the data).

[0075] In this embodiment of the invention, the fault is isolated by setting the measurement data collected by the faulty sensor to 0, and the normal measurement data of the faulty sensor can be predicted and restored by inputting the remaining normal measurement data into the fault data reconstruction model.

[0076] Step 3: Using a pre-trained fault data reconstruction model, based on the measurement data collected by the target sensor and the measurement data collected by the fault sensor after zeroing, predict the reconstructed data corresponding to the fault sensor. In one implementation, a fault feature extraction encoder based on an improved Transformer-CNN (i.e., a fault data reconstruction model) can be designed to achieve fault recovery.

[0077] In one specific implementation method, see Figure 6 The diagram shows a structural schematic of a fault data reconstruction model, which includes a second encoder (also known as a fault isolation reconstruction encoder) and a second decoder (also known as a fault isolation reconstruction decoder).

[0078] Specifically, the second encoder includes a fourth CNN layer, a second Transformer layer, a fifth CNN layer, a second pooling layer, and a sixth CNN layer connected in sequence, and the second decoder includes a third fully connected layer, a second activation layer, and a fourth fully connected layer connected in sequence.

[0079] exist Figure 6 Based on this, the present invention provides an implementation method for predicting the reconstructed data corresponding to the fault sensor by using a pre-trained fault data reconstruction model based on the measurement data collected by the target sensor and the measurement data collected by the fault sensor after being zeroed: (1) extracting the second high-dimensional features of the measurement data collected by the target sensor and the measurement data collected by the fault sensor after being zeroed by the second encoder; (2) predicting the reconstructed data corresponding to the fault sensor based on the second high-dimensional features by the second decoder.

[0080] In practical applications, data from normal sensors and zero-set data from faulty sensors are input into the second encoder. The second encoder correlates the spatiotemporal relationship between the normal and faulty data, and the reconstructed data of the faulty sensor is obtained by decoding and predicting the features through the second decoder. In the second decoder, the flattening layer and normalization layer of the first decoder are removed, and the normal data of the faulty sensor is directly predicted through the fully connected output.

[0081] In one implementation, after determining the reconstructed data corresponding to the faulty sensor, the reconstructed data can be used to replace the measurement data collected by the faulty sensor in the multi-source sensor data. The replaced multi-source sensor data is then sent to a designated processing device for data processing. This embodiment of the invention replaces the data collected by the faulty sensor with the predicted reconstructed data, thereby achieving closed-loop fault diagnosis of distributed sensors and effectively avoiding negative impacts on subsequent data processing caused by the data collected by the faulty sensor.

[0082] In summary, the distributed sensor fault handling method provided by this invention designs an improved Transformer-CNN (Convolutional Neural Networks) fault feature extraction encoder to extract the temporal and spatial features of multi-source data from distributed IMU sensors. Then, a decoder is constructed using a multilayer perceptron to achieve fault IMU location detection and fault IMU data reconstruction, thereby realizing closed-loop fault diagnosis for distributed sensors. The advantage of this invention is that the improved Transformer-CNN encoder can effectively extract high-dimensional features from multi-source data of distributed IMU sensors, solving the problem of difficult fault feature extraction for distributed IMU sensors. Furthermore, existing sensor fault diagnosis methods often focus on fault location and isolation, without forming a closed-loop fault diagnosis scheme. Research on data reconstruction and recovery after sensor failure is scarce. This invention designs a fault sensor data reconstruction method based on the data redundancy characteristics of distributed sensors, thereby achieving closed-loop fault diagnosis. Additionally, this invention does not require accurate mathematical modeling of the distributed IMU sensor, resulting in better generalization ability and accuracy for fault diagnosis of distributed IMU sensors.

[0083] Regarding the distributed sensor fault handling method provided in the foregoing embodiments, this invention provides a distributed sensor fault handling device, see [link to related documentation]. Figure 7 The diagram shows a structural schematic of a distributed sensor fault handling device, which mainly includes the following parts:

[0084] The data acquisition module 702 is used to acquire multi-source sensor data collected by distributed sensors within a preset time interval;

[0085] The fault location module 704 is used to identify the faulty sensor from distributed sensors based on multi-source sensor data using a pre-trained fault location model, and to identify the sensor location identifier and sensor type identifier of the faulty sensor.

[0086] The data reconstruction module 706 is used to predict the reconstructed data corresponding to the faulty sensor based on multi-source sensor data, sensor location identifier, and sensor type identifier through a pre-trained fault data reconstruction model.

[0087] The distributed sensor fault handling device provided in this embodiment of the invention can locate the faulty sensor, its sensor location identifier, and sensor type identifier from the distributed sensors by using a fault location model after acquiring multi-source sensor data within a preset time interval. Then, it can use the fault data reconstruction model to reconstruct the measurement data of the faulty sensor by combining the multi-source sensor data, sensor location identifier, and sensor type identifier to obtain the corresponding reconstructed data, thereby realizing closed-loop fault diagnosis of distributed sensors. In addition, this embodiment of the invention does not require accurate mathematical modeling of distributed sensors, and has better generalization ability and accuracy for fault diagnosis of distributed sensors.

[0088] In one implementation, the fault location model includes a first encoder and a first decoder;

[0089] The fault location module 704 is also used for:

[0090] The first high-dimensional feature of the multi-source data from the sensor is extracted using the first encoder;

[0091] The first decoder determines the failure probability of each sensor in the distributed sensor based on the first high-dimensional feature, and identifies the faulty sensor from the distributed sensor based on the failure probability, as well as the sensor location identifier and sensor type identifier of the faulty sensor.

[0092] In one implementation, the sensor multi-source data includes acceleration data and angular velocity data collected by each sensor in the distributed sensor system;

[0093] The data reconstruction module 706 is also used for:

[0094] Based on sensor type identifiers, the measurement data collected by the target sensor corresponding to the sensor type identifier is extracted from multi-source sensor data;

[0095] Additionally, based on sensor location identifiers, the measurement data collected by the faulty sensor is extracted from multi-source sensor data, and the measurement data collected by the faulty sensor is set to zero.

[0096] By using a pre-trained fault data reconstruction model, the model predicts the reconstructed data corresponding to the fault sensor based on the measurement data collected by the target sensor and the measurement data collected by the fault sensor after being zeroed.

[0097] In one implementation, the sensor type identifier includes an accelerometer identifier or a gyroscope identifier;

[0098] The data reconstruction module 706 is also used for:

[0099] If the sensor type is identified as accelerometer, then the accelerometer sensor in the distributed sensor is used as the target sensor, and the measurement data collected by the accelerometer sensor is extracted from the multi-source sensor data.

[0100] Alternatively, if the sensor type is identified as a gyroscope, then the gyroscope sensor in the distributed sensor is used as the target sensor, and the measurement data collected by the gyroscope sensor is extracted from the multi-source sensor data.

[0101] In one implementation, the fault data reconstruction model includes a second encoder and a second decoder;

[0102] The data reconstruction module 706 is also used for:

[0103] The second encoder extracts the second high-dimensional features of the measurement data collected by the target sensor and the measurement data collected by the fault sensor after being zeroed.

[0104] The second decoder predicts the reconstructed data corresponding to the faulty sensor based on the second high-dimensional features.

[0105] In one embodiment, the first encoder includes a first CNN layer, a first Transformer layer, a second CNN layer, a first pooling layer, and a third CNN layer connected in sequence, and the first decoder includes a flattening layer, a first fully connected layer, a first activation layer, a second fully connected layer, and a normalization layer connected in sequence.

[0106] The second encoder comprises a fourth CNN layer, a second Transformer layer, a fifth CNN layer, a second pooling layer, and a sixth CNN layer connected in sequence, and the second decoder comprises a third fully connected layer, a second activation layer, and a fourth fully connected layer connected in sequence.

[0107] In one implementation, it further includes: a data replacement module, used for:

[0108] The reconstructed data corresponding to the faulty sensor is used to replace the measurement data collected by the faulty sensor in the multi-source sensor data.

[0109] The replaced sensor multi-source data is sent to a designated processing device for data processing.

[0110] The device provided in this embodiment of the invention has the same implementation principle and technical effect as the aforementioned method embodiment. For the sake of brevity, any parts not mentioned in the device embodiment can be referred to the corresponding content in the aforementioned method embodiment.

[0111] This invention provides an electronic device, specifically, the electronic device includes a processor and a storage device; the storage device stores a computer program, and the computer program, when run by the processor, executes the method described in any of the above embodiments.

[0112] Figure 8 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention. The electronic device 100 includes: a processor 80, a memory 81, a bus 82, and a communication interface 83. The processor 80, the communication interface 83, and the memory 81 are connected through the bus 82. The processor 80 is used to execute executable modules, such as computer programs, stored in the memory 81.

[0113] The memory 81 may include high-speed random access memory (RAM) or non-volatile memory, such as at least one disk storage device. Communication between this system network element and at least one other network element is achieved through at least one communication interface 83 (which can be wired or wireless), such as the Internet, wide area network, local area network, metropolitan area network, etc.

[0114] Bus 82 can be an ISA bus, PCI bus, or EISA bus, etc. The bus can be divided into address bus, data bus, control bus, etc. For ease of representation, Figure 8 The symbol is represented by a single double-headed arrow, but this does not mean that there is only one bus or one type of bus.

[0115] The memory 81 is used to store programs. After receiving an execution instruction, the processor 80 executes the program. The method executed by the device for defining the flow process disclosed in any of the foregoing embodiments of the present invention can be applied to the processor 80 or implemented by the processor 80.

[0116] The processor 80 may be an integrated circuit chip with signal processing capabilities. In implementation, each step of the above method can be completed by the integrated logic circuitry in the hardware of the processor 80 or by software instructions. The processor 80 may be a general-purpose processor, including a Central Processing Unit (CPU), a Network Processor (NP), etc.; it may also be a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), a Field-Programmable Gate Array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. It can implement or execute the methods, steps, and logic block diagrams disclosed in the embodiments of this invention. The general-purpose processor may be a microprocessor or any conventional processor. The steps of the methods disclosed in the embodiments of this invention can be directly embodied in the execution of a hardware decoding processor, or executed by a combination of hardware and software modules in the decoding processor. The software modules may reside in random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, registers, or other mature storage media in the art. The storage medium is located in memory 81. The processor 80 reads the information in memory 81 and, in conjunction with its hardware, completes the steps of the above method.

[0117] The computer program product of the readable storage medium provided in the embodiments of the present invention includes a computer-readable storage medium storing program code. The instructions included in the program code can be used to execute the methods described in the foregoing method embodiments. For specific implementation, please refer to the foregoing method embodiments, which will not be repeated here.

[0118] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0119] Finally, it should be noted that the above-described embodiments are merely specific implementations of the present invention, used to illustrate the technical solutions of the present invention, and not to limit it. The scope of protection of the present invention is not limited thereto. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that any person skilled in the art can still modify or easily conceive of changes to the technical solutions described in the foregoing embodiments within the technical scope disclosed in the present invention, or make equivalent substitutions for some of the technical features; and these modifications, changes, or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be covered within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for handling distributed sensor faults, characterized in that, include: Acquire multi-source sensor data collected by distributed sensors within a preset time interval. The multi-source sensor data includes acceleration data and angular velocity data collected by each sensor in the distributed sensors. Using a pre-trained fault location model, the faulty sensor is identified from the distributed sensors based on the multi-source data of the sensors, and the sensor location identifier and sensor type identifier of the faulty sensor are also determined. Using a pre-trained fault data reconstruction model, based on the multi-source sensor data, the sensor location identifier, and the sensor type identifier, the model predicts the reconstructed data corresponding to the faulty sensor. This includes: extracting measurement data collected by the target sensor corresponding to the sensor type identifier from the multi-source sensor data based on the sensor type identifier; and extracting the measurement data collected by the faulty sensor from the multi-source sensor data based on the sensor location identifier, and setting the measurement data collected by the faulty sensor to zero. The model then uses the pre-trained fault data reconstruction model to predict the reconstructed data corresponding to the faulty sensor based on the measurement data collected by the target sensor and the zeroed measurement data collected by the faulty sensor.

2. The distributed sensor fault handling method according to claim 1, characterized in that, The fault location model includes a first encoder and a first decoder; Using a pre-trained fault location model, the system identifies faulty sensors from the distributed sensors based on multi-source sensor data, and determines the sensor location and sensor type identifiers of the faulty sensors, including: The first high-dimensional feature of the multi-source data from the sensor is extracted using the first encoder; The first decoder determines the failure probability of each sensor in the distributed sensors based on the first high-dimensional feature, identifies the faulty sensor from the distributed sensors based on the failure probability, and determines the sensor location identifier and sensor type identifier of the faulty sensor.

3. The distributed sensor fault handling method according to claim 1, characterized in that, The sensor type identifier includes an accelerometer identifier or a gyroscope identifier; Based on the sensor type identifier, the measurement data collected by the target sensor corresponding to the sensor type identifier is extracted from the multi-source sensor data, including: If the sensor type is identified as an accelerometer, then the accelerometer sensor in the distributed sensor is used as the target sensor, and the measurement data collected by the accelerometer sensor is extracted from the multi-source data of the sensor. Alternatively, if the sensor type is identified as a gyroscope, then the gyroscope sensor in the distributed sensor is used as the target sensor, and the measurement data collected by the gyroscope sensor is extracted from the multi-source data of the sensor.

4. The distributed sensor fault handling method according to claim 1, characterized in that, The fault data reconstruction model includes a second encoder and a second decoder; Using a pre-trained fault data reconstruction model, based on the measurement data collected by the target sensor and the measurement data collected by the fault sensor after being zeroed, the model predicts the reconstructed data corresponding to the fault sensor, including: The second encoder extracts the second high-dimensional features of the measurement data collected by the target sensor and the measurement data collected by the fault sensor after being zeroed. The second decoder predicts the reconstructed data corresponding to the fault sensor based on the second high-dimensional feature.

5. The distributed sensor fault handling method according to claim 2 or 4, characterized in that, The first encoder includes a first CNN layer, a first Transformer layer, a second CNN layer, a first pooling layer, and a third CNN layer connected in sequence; the first decoder includes a flattening layer, a first fully connected layer, a first activation layer, a second fully connected layer, and a normalization layer connected in sequence. The second encoder comprises a fourth CNN layer, a second Transformer layer, a fifth CNN layer, a second pooling layer, and a sixth CNN layer connected in sequence, and the second decoder comprises a third fully connected layer, a second activation layer, and a fourth fully connected layer connected in sequence.

6. The distributed sensor fault handling method according to claim 1, characterized in that, After predicting the reconstructed data corresponding to the faulty sensor based on the multi-source data of the sensor, the sensor location identifier, and the sensor type identifier using a pre-trained fault data reconstruction model, the method further includes: The reconstructed data corresponding to the faulty sensor is used to replace the measurement data collected by the faulty sensor in the multi-source sensor data. The replaced sensor multi-source data is sent to a designated processing device for data processing.

7. A distributed sensor fault handling device, characterized in that, include: The data acquisition module is used to acquire multi-source sensor data collected by the distributed sensors within a preset time interval. The multi-source sensor data includes acceleration data and angular velocity data collected by each sensor in the distributed sensors. The fault location module is used to determine the faulty sensor from the distributed sensors based on the multi-source data of the sensors using a pre-trained fault location model, and to determine the sensor location identifier and sensor type identifier of the faulty sensor. The data reconstruction module is used to predict the reconstructed data corresponding to the faulty sensor based on the multi-source sensor data, the sensor location identifier, and the sensor type identifier using a pre-trained fault data reconstruction model. This includes: extracting measurement data collected by the target sensor corresponding to the sensor type identifier from the multi-source sensor data based on the sensor type identifier; and extracting the measurement data collected by the faulty sensor from the multi-source sensor data based on the sensor location identifier, and setting the measurement data collected by the faulty sensor to zero. The module then uses the pre-trained fault data reconstruction model to predict the reconstructed data corresponding to the faulty sensor based on the measurement data collected by the target sensor and the zeroed measurement data collected by the faulty sensor.

8. An electronic device, characterized in that, The method includes a processor and a memory, the memory storing computer-executable instructions executable by the processor, the processor executing the computer-executable instructions to implement the method of any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer-executable instructions that, when invoked and executed by a processor, cause the processor to perform the method according to any one of claims 1 to 6.

Images