Sensor circuit failure diagnosis method, system, medium, and apparatus

Abstract

This invention belongs to the field of sensor circuit fault diagnosis, and discloses a method, system, medium, and device for sensor circuit fault diagnosis. The method includes acquiring a continuous-time voltage signal output by a sensor circuit; performing time-domain and frequency-domain feature extraction on the continuous-time voltage signal to obtain time-domain feature data and frequency-domain feature data; and calling a preset sensor circuit fault diagnosis model based on the continuous-time voltage signal, time-domain feature data, and frequency-domain feature data to obtain a sensor circuit fault diagnosis result. This invention achieves joint processing of feature extraction and time-series signals, considering not only the correlation between the continuous-time voltage signal itself and sensor circuit faults, but also utilizing feature engineering methods to perform dual feature extraction in both the time and frequency domains of the continuous-time voltage signal. These data are then input into the sensor circuit fault diagnosis model for fusion to obtain the sensor circuit fault diagnosis result, significantly improving accuracy.

Description

Technical Field

[0001] This invention belongs to the field of sensor circuit fault diagnosis, and relates to a sensor circuit fault diagnosis method, system, medium and device. Background Technology

[0002] With the rapid development of IoT technology and the electronic circuit industry, sensor circuits containing various components are being widely used in various fields of industrial production and daily life. The requirements for the reliability of sensor circuits are also increasing. Most faults in sensor circuits originate from component failures. Process deviations in actual production, poor contact during soldering, and various non-ideal factors in the external environment can all lead to component failures in sensor circuits, thus causing sensor circuit failures, affecting equipment operation, and in severe cases, causing significant economic losses or even dangerous accidents. Furthermore, as the complexity of sensor circuit components increases, traditional fault diagnosis methods are insufficient to meet current diagnostic needs. How to quickly locate faulty components in sensor circuits has gradually become a research hotspot in both academia and industry.

[0003] Fault diagnosis of hardware circuits such as sensor circuits involves processing and analyzing the circuit's output signals to accurately pinpoint the fault location. Currently, there are two main methods for fault diagnosis of hardware circuits such as sensor circuits. One method involves extracting features from the circuit's output signals to obtain a small number of features related to the circuit fault, and then using machine learning methods such as SVM for fault classification. The other method directly utilizes the time-series or frequency-domain signals output by the circuit, inputting the signals into a neural network for high-dimensional data processing and outputting classification information to complete the circuit fault diagnosis. However, both methods have relatively low diagnostic accuracy. Summary of the Invention

[0004] The purpose of this invention is to overcome the shortcomings of the prior art and provide a method, system, medium and device for diagnosing sensor circuit faults.

[0005] To achieve the above objectives, the present invention employs the following technical solution:

[0006] In a first aspect, the present invention provides a method for diagnosing sensor circuit faults, comprising:

[0007] Acquire the continuous voltage signal in the time domain output by the sensor circuit; extract time-domain features and frequency-domain features from the continuous voltage signal to obtain time-domain feature data and frequency-domain feature data; based on the continuous voltage signal, time-domain feature data, and frequency-domain feature data, call the preset sensor circuit fault diagnosis model to obtain the sensor circuit fault diagnosis result.

[0008] Optionally, the step of extracting time-domain features and frequency-domain features from the continuous voltage signal in the time domain to obtain time-domain feature data and frequency-domain feature data includes: obtaining one or more of the maximum value, minimum value, average value, standard deviation, kurtosis, and skewness of the continuous voltage signal in the time domain to obtain time-domain feature data; performing time-frequency conversion on the continuous voltage signal in the time domain to obtain a continuous voltage signal in the frequency domain; and obtaining one or more of the bandwidth and center frequency of the continuous voltage signal in the frequency domain to obtain frequency-domain feature data.

[0009] Optionally, the step of calling a preset sensor circuit fault diagnosis model based on the time-domain continuous voltage signal, time-domain feature data, and frequency-domain feature data to obtain sensor circuit fault diagnosis results includes: combining time-domain feature data and frequency-domain feature data to obtain time-frequency mixed data; normalizing both the time-domain continuous voltage signal and the time-frequency mixed data to obtain normalized time-frequency voltage data and normalized time-frequency mixed data; and inputting the normalized time-frequency voltage data and normalized time-frequency mixed data into the preset sensor circuit fault diagnosis model to obtain sensor circuit fault diagnosis results.

[0010] Optionally, the preset sensor circuit fault diagnosis model includes a DNN neural network, an LSTM neural network, a fully connected layer network, and a SoftMax layer network. The DNN neural network is used to input normalized time-frequency mixed data and outputs DNN neural network output data to the fully connected layer network. The LSTM neural network is used to input normalized time-frequency voltage data and outputs LSTM neural network output data to the fully connected layer network. The fully connected layer network is used to perform fully connected processing on the DNN neural network output data and the LSTM neural network output data to obtain fully connected layer network output data and output it to the SoftMax layer network. The SoftMax layer network is used to obtain and output the sensor circuit fault diagnosis result based on the input fully connected layer network output data.

[0011] Optionally, the SoftMax function of the SoftMax layer network is:

[0012]

[0013] Where fc_out(m)[i] is the i-th data point of the fully connected layer network output data of the m-th group of time-domain continuous voltage signals, fc_out(m)[j] is the j-th data point of the fully connected layer network output data of the m-th group of time-domain continuous voltage signals, m is the number of acquisition groups of time-domain continuous voltage signals, and K is the number of fault types of the sensor circuit; s(fc_out(m)[i]) is the i-th data point of the fault diagnosis result of the sensor circuit, i>0 indicates the probability that sensor circuit element i will fail, and i=0 indicates the probability that the sensor circuit will not fail.

[0014] A second aspect of the present invention provides a sensor circuit fault diagnosis system, comprising:

[0015] The data acquisition module is used to acquire the time-domain continuous voltage signal output by the sensor circuit;

[0016] The feature extraction module is used to extract time-domain and frequency-domain features from the continuous voltage signal in the time domain, and obtain time-domain feature data and frequency-domain feature data.

[0017] The diagnostic module is used to call a preset sensor circuit fault diagnosis model based on the time-domain continuous voltage signal to obtain the sensor circuit fault diagnosis result.

[0018] Optionally, the diagnostic module is specifically used for: combining time-domain feature data and frequency-domain feature data to obtain time-frequency mixed data; normalizing both the time-domain continuous voltage signal and the time-frequency mixed data to obtain normalized time-frequency voltage data and normalized time-frequency mixed data; and inputting the normalized time-frequency voltage data and normalized time-frequency mixed data into a preset sensor circuit fault diagnosis model to obtain sensor circuit fault diagnosis results.

[0019] Optionally, the preset sensor circuit fault diagnosis model includes a DNN neural network, an LSTM neural network, a fully connected layer network, and a SoftMax layer network. The DNN neural network is used to input normalized time-frequency mixed data and outputs DNN neural network output data to the fully connected layer network. The LSTM neural network is used to input normalized time-frequency voltage data and outputs LSTM neural network output data to the fully connected layer network. The fully connected layer network is used to perform fully connected processing on the DNN neural network output data and the LSTM neural network output data to obtain fully connected layer network output data and output it to the SoftMax layer network. The SoftMax layer network is used to obtain and output the sensor circuit fault diagnosis result based on the input fully connected layer network output data.

[0020] In a third aspect, the present invention provides a sensor circuit fault diagnosis device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the sensor circuit fault diagnosis method described above.

[0021] In a fourth aspect, the present invention provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the above-described sensor circuit fault diagnosis method.

[0022] Compared with the prior art, the present invention has the following beneficial effects:

[0023] This invention provides a sensor circuit fault diagnosis method that extracts time-domain and frequency-domain features from a continuous voltage signal to obtain time-domain and frequency-domain feature data. Based on these data, a pre-defined sensor circuit fault diagnosis model is invoked to obtain the sensor circuit fault diagnosis result. This method achieves joint processing of feature extraction and time-series signals during sensor circuit fault diagnosis. It not only considers the correlation between the continuous voltage signal itself and the sensor circuit fault but also utilizes feature engineering methods to extract dual features from the continuous voltage signal in both the time and frequency domains. Finally, these data are input into the pre-defined sensor circuit fault diagnosis model for fusion, ultimately obtaining the sensor circuit fault diagnosis result, thus significantly improving the accuracy of the sensor circuit fault diagnosis result. Attached Figure Description

[0024] Figure 1 This is a flowchart of a sensor circuit fault diagnosis method according to an embodiment of the present invention;

[0025] Figure 2 This is a block diagram of the sensor circuit fault diagnosis model according to an embodiment of the present invention;

[0026] Figure 3 This is a schematic diagram illustrating the principle of the LSTM neural network in an embodiment of the present invention.

[0027] Figure 4 This is a block diagram of the sensor circuit fault diagnosis system according to an embodiment of the present invention. Detailed Implementation

[0028] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0029] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0030] As introduced in the background section, there are currently two main methods for fault diagnosis of sensor circuits. One method is to extract features from the circuit output signal to obtain a small number of features related to the circuit fault, and then use machine learning methods such as SVM for fault classification. The other method is to directly use the time-series signal or frequency domain signal output by the circuit, input the signal into a neural network for high-level data processing, and output classification information to complete the diagnosis of circuit faults. However, both methods suffer from low diagnostic accuracy.

[0031] To address the aforementioned problems, this invention provides a sensor circuit fault diagnosis method, comprising: acquiring a continuous-time voltage signal output by a sensor circuit; performing time-domain feature extraction and frequency-domain feature extraction on the continuous-time voltage signal to obtain time-domain feature data and frequency-domain feature data; and, based on the continuous-time voltage signal, time-domain feature data, and frequency-domain feature data, calling a preset sensor circuit fault diagnosis model to obtain a sensor circuit fault diagnosis result. This method achieves joint processing of feature extraction and time-series signals, considering not only the correlation between the continuous-time voltage signal itself and sensor circuit faults, but also utilizing feature engineering methods to perform dual feature extraction in both the time and frequency domains of the continuous-time voltage signal. Finally, these data are input together into the sensor circuit fault diagnosis model for fusion, ultimately obtaining the sensor circuit fault diagnosis result, significantly improving accuracy. The invention will now be described in further detail with reference to the accompanying drawings:

[0032] See Figure 1 In one embodiment of the present invention, a sensor circuit fault diagnosis method is provided, which proposes the idea of ​​integrating feature extraction and time-series feature processing to improve the accuracy of sensor circuit fault diagnosis and ultimately achieve accurate classification and location of sensor circuit fault types.

[0033] Specifically, the fault diagnosis method for this sensor circuit includes the following steps:

[0034] S1: Acquire the time-domain continuous voltage signal output by the sensor circuit.

[0035] S2: Extract time-domain and frequency-domain features from the continuous voltage signal in the time domain to obtain time-domain feature data and frequency-domain feature data.

[0036] S3: Based on the time-domain continuous voltage signal, time-domain feature data, and frequency-domain feature data, call the preset sensor circuit fault diagnosis model to obtain the sensor circuit fault diagnosis result.

[0037] Optionally, in step S1, when acquiring the time-domain continuous voltage signal output by the sensor circuit, the target sensor circuit can be simulated to obtain a target analog circuit. The output terminal of the target analog circuit is used as a test point (i.e., a data acquisition point) to obtain the time-domain continuous voltage signal. Sampling is performed at time intervals ΔT, and a total of N voltage value points are acquired, represented as V(m)=[v1,v2,...,v...]. N ], where m represents the number of acquisition groups of continuous time-domain voltage signals, and the total number of acquisition groups of continuous time-domain voltage signals is M. For each group of continuous time-domain voltage signals, its fault type can be represented as {F0, F1, ..., F K}, where F i ,i∈{1,2,...,K} indicates that component i in the sensor circuit has failed, and F0 indicates that the sensor circuit has no faults.

[0038] Optionally, the preset sensor circuit fault diagnosis model can be constructed using a neural network. Based on the different characteristics of the time-domain continuous voltage signal, time-domain feature data, and frequency-domain feature data, high-dimensional features are extracted, processed, and fused to ultimately obtain the location and classification prediction of sensor circuit faults as the diagnostic result.

[0039] In summary, the sensor circuit fault diagnosis method of the present invention extracts time-domain and frequency-domain features from a continuous voltage signal to obtain time-domain feature data and frequency-domain feature data. Based on the continuous voltage signal, time-domain feature data, and frequency-domain feature data, a preset sensor circuit fault diagnosis model is invoked to obtain the sensor circuit fault diagnosis result. In the sensor circuit fault diagnosis, the method achieves joint processing of feature extraction and time-series signals. It not only considers the correlation between the continuous voltage signal itself and the sensor circuit fault, but also uses feature engineering methods to extract dual features from the continuous voltage signal in both the time and frequency domains. Finally, these data are input together into the preset sensor circuit fault diagnosis model for fusion, ultimately obtaining the sensor circuit fault diagnosis result, which greatly improves the accuracy of the sensor circuit fault diagnosis result.

[0040] In one possible implementation, the step of extracting time-domain features and frequency-domain features from the continuous voltage signal in the time domain to obtain time-domain feature data and frequency-domain feature data includes: obtaining one or more of the maximum value, minimum value, average value, standard deviation, kurtosis, and skewness of the continuous voltage signal in the time domain to obtain time-domain feature data; performing time-frequency conversion on the continuous voltage signal in the time domain to obtain a continuous voltage signal in the frequency domain; and obtaining one or more of the bandwidth and center frequency of the continuous voltage signal in the frequency domain to obtain frequency-domain feature data.

[0041] Specifically, regarding the time-domain feature extraction of a continuous voltage signal, this embodiment uses the maximum value, minimum value, average value, standard deviation, kurtosis, and skewness as time-domain feature indicators to extract the time-domain features of the continuous voltage signal. This means obtaining the maximum value, minimum value, average value, standard deviation, kurtosis, and skewness of the continuous voltage signal, which can be represented as td(m):

[0042] td(m)=[v max (m),v min (m),v avg (m),v std (m),v peak (m),v ske (m)]

[0043] Among them, v max (m) = max(V(m)) is the maximum value of the continuous voltage signal in the time domain, v min (m) = min(V(m)) is the minimum value of the continuous voltage signal in the time domain. The average value of the continuous voltage signal in the time domain. The standard deviation of a continuous voltage signal in the time domain. Kurtosis is the kurtosis of a continuous voltage signal in the time domain, used to characterize the peak value of the probability density distribution curve at the average value. The skewness is the skewness of a continuous voltage signal in the time domain, used to characterize the degree of asymmetry of the probability distribution density curve relative to the mean value.

[0044] Regarding the frequency domain feature extraction of a continuous voltage signal in the time domain, a time-frequency conversion is first performed to obtain the continuous voltage signal in the frequency domain. Then, feature extraction is performed on the continuous voltage signal in the frequency domain. In this embodiment, bandwidth and center frequency are used as frequency domain feature indicators to obtain the bandwidth and center frequency of the continuous voltage signal in the frequency domain, thus obtaining frequency domain feature data. This frequency domain feature data can be represented as fd(m):

[0045] fd(m) = [band(m), freq(m)]

[0046] Where band(m) is the bandwidth of the frequency domain continuous voltage signal, and freq(m) is the center frequency of the frequency domain continuous voltage signal.

[0047] In one possible implementation, the step of calling a preset sensor circuit fault diagnosis model based on the time-domain continuous voltage signal, time-domain feature data, and frequency-domain feature data to obtain sensor circuit fault diagnosis results includes: combining time-domain feature data and frequency-domain feature data to obtain time-frequency mixed data; normalizing both the time-domain continuous voltage signal and the time-frequency mixed data to obtain normalized time-frequency voltage data and normalized time-frequency mixed data; and inputting the normalized time-frequency voltage data and normalized time-frequency mixed data into the preset sensor circuit fault diagnosis model to obtain sensor circuit fault diagnosis results.

[0048] Specifically, the time-domain feature data and the frequency-domain feature data are first combined to form time-frequency hybrid data, which is represented as tf(m): tf(m) = [fd(m), td(m)].

[0049] Then, normalization is performed on each group of time-frequency mixed data to obtain the corresponding normalized time-frequency mixed data norm_tf(m), and normalization is performed on each group of time-domain continuous voltage signals to obtain the corresponding normalized time-frequency voltage data norm_V(m).

[0050] Optionally, when normalizing each set of time-frequency mixed data, the min-max normalization method is used, and the normalization of each set of time-frequency mixed data is achieved by the following formula:

[0051]

[0052] Among them, tf i (m) represents the i-th data point in the m-th group of time-frequency mixed data, tf i * (m) represents tf i (m) The normalized numerical value, tf min (m) represents the minimum value of tf(m), tf max (m) represents the maximum value of tf(m).

[0053] Optionally, each set of continuous voltage signals in the time domain can be normalized. Alternatively, the min-max normalization method can be used, and the normalization of each set of continuous voltage signals in the time domain can be achieved by the following formula:

[0054]

[0055] Among them, v i (m) represents the i-th data in the m-th group of time-domain continuous voltage signals. Indicates vi (m) The normalized value.

[0056] In one possible implementation, see Figure 2 The sensor circuit fault diagnosis model includes DNN neural network, LSTM neural network, fully connected layer network and SoftMax (classification network) layer network.

[0057] Fully connected layers (FC) act as a "classifier" in the entire convolutional neural network. If convolutional layers, pooling layers, and activation function layers map the raw data to the hidden feature space, then fully connected layers map the learned distributed feature representations to the sample label space. In practice, fully connected layers can be implemented using convolutional operations. The Softmax layer maps several real numbers (-∞, +∞) to the same number of real numbers (0, 1) (representing probabilities), while ensuring that their sum is 1.

[0058] The DNN neural network is used as input for normalized time-frequency mixed data and outputs its data to the fully connected layer network. The LSTM neural network is used as input for normalized time-frequency voltage data and outputs its data to the fully connected layer network. The fully connected layer network is used to perform fully connected processing on the output data of the DNN neural network and the LSTM neural network to obtain the output data of the fully connected layer network and output it to the SoftMax layer network. The SoftMax layer network is used to obtain the sensor circuit fault diagnosis result based on the input fully connected layer network output data and output it.

[0059] The specific process of inputting normalized time-frequency voltage data and normalized time-frequency mixed data into a preset sensor circuit fault diagnosis model to obtain sensor circuit fault diagnosis results is as follows:

[0060] Step 1: Input the normalized time-frequency mixed data norm_tf(m) into the DNN neural network and obtain the DNN neural network output data d_out(m).

[0061] Step 2: Input the normalized time-frequency voltage data norm_V(m) into the LSTM neural network and obtain the output data l_out(m) of the LSTM neural network.

[0062] Step 3: Combine the DNN neural network output data d_out(m) from Step 1 and the LSTM neural network output data l_out(m) from Step 2, and input them into the fully connected layer network to obtain the fully connected layer network output data fc_out(m). fc_out(m) is a K+1 dimensional vector, where K is the number of fault types in the sensor circuit. The number of fault types in the sensor circuit is related to the number of components in the sensor circuit. The sensor circuit has a total of... K If there are multiple components, the fault types are {component 1 fault, component 2 fault, ..., component K fault}.

[0063] Step 4: Use the output data fc_out(m) of the fully connected layer network as the input of the SoftMax layer network. After calculation by the SoftMax layer network, a vector of dimension K+1 is obtained, which is the sensor circuit fault diagnosis result.

[0064] Specifically, in this embodiment, based on the different characteristics of time-frequency mixed data and time-frequency voltage data, feature extraction is performed using DNN (Deep Neural Networks) decision neural networks and LSTM (Long Short Term Memory) neural networks, respectively. Finally, the data from the two networks are combined to obtain the final predicted value, thereby improving the predictive ability of the sensor circuit fault diagnosis model.

[0065] Optionally, the DNN neural network is a fully connected neural network comprising one input layer, one output layer, and two hidden layers. The forward propagation function of the hidden layers is:

[0066]

[0067] Among them, y i For the i-th output of this hidden layer, x j For the j-th input of this hidden layer, w i,j b is the weight of the j-th input corresponding to the i-th output. i The bias corresponds to the i-th input. It should be noted that the forward propagation formula for the fully connected layer network in this embodiment is the same as the above formula.

[0068] Optional, see Figure 3 Each cell of the LSTM neural network performs forward propagation through an input gate, a forget gate, and an output gate.

[0069] The update of the forget gate can be represented as:

[0070] f t =σ·(W f h t-1 +U f xt +b f )

[0071] Where σ is the sigmoid activation function, W f U f and b f Here, h represents the coefficients and bias of the forget gate, both of which are trainable parameters. t-1 Let x be the hidden output state of the (t-1)th cell. t For the t-th input of this sequence, f represents the value of the t-th element of the normalized time-frequency voltage data norm_V(m) for this round. t This represents the updated state of the t-th forget gate.

[0072] The update of the input gate can be represented as:

[0073] i t =σ·(W i h t-1 +U i x t +b i )

[0074]

[0075] C t =f t ·C t-1 +i t ·C t

[0076] Among them, W i U i b i W c U c and b c Let C be the coefficients and bias of the input gate, both of which are trainable parameters. t-1 C represents the long-term state of the previous moment. t Let i be the state after the t-th input gate is updated, and let i represent the long-term state at the current time step. t and These are all intermediate state quantities of the input gate. Indicates the current memory state, i t Indicates to The ability to forget.

[0077] The update of the output gate can be represented as:

[0078] o t =σ·(W o h t-1 +U o x t +b o)

[0079] h t =o t ·tanh(C t )

[0080] Among them, W o U o and b o The output gate coefficients and biases are trainable parameters, h. t Let O be the hidden output state of the t-th cell. t It represents the ability to forget long-term states at the current moment.

[0081] At this point, one cell of the LSTM neural network has completed the full forward propagation.

[0082] In one possible implementation, the SoftMax function of the SoftMax layer network is:

[0083]

[0084] Where fc_out(m)[i] is the i-th data point of the fully connected layer network output data of the m-th group of time-domain continuous voltage signals, fc_out(m)[j] is the j-th data point of the fully connected layer network output data of the m-th group of time-domain continuous voltage signals, m is the number of acquisition groups of time-domain continuous voltage signals, and K is the number of fault types of the sensor circuit; s(fc_out(m)[i]) is the i-th data point of the sensor circuit fault diagnosis result, i>0 indicates the probability that sensor circuit element i will fail, and i=0 indicates the probability that the sensor circuit will not fail.

[0085] Finally, s(m) = [s(fc_out(m)[0]), s(fc_out(m)[1]), ..., s(fc_out(m)[K])] is output as the fault diagnosis result of the sensor circuit, so as to reflect the probability of faults in the sensor circuit and its internal components.

[0086] The following are embodiments of the apparatus of the present invention, which can be used to execute embodiments of the method of the present invention. For details not disclosed in the apparatus embodiments, please refer to the embodiments of the method of the present invention.

[0087] See Figure 4 In another embodiment of the present invention, a sensor circuit fault diagnosis system is provided, which can be used to implement the above-mentioned sensor circuit fault diagnosis system method. Specifically, the sensor circuit fault diagnosis system includes a data acquisition module, a feature extraction module, and a diagnosis module.

[0088] The data acquisition module is used to acquire the time-domain continuous voltage signal output by the sensor circuit; the feature extraction module is used to extract time-domain features and frequency-domain features from the time-domain continuous voltage signal to obtain time-domain feature data and frequency-domain feature data; the diagnosis module is used to call a preset sensor circuit fault diagnosis model based on the time-domain continuous voltage signal to obtain the sensor circuit fault diagnosis result.

[0089] In one possible implementation, the diagnostic module is specifically used to: combine time-domain feature data and frequency-domain feature data to obtain time-frequency mixed data; normalize both the time-domain continuous voltage signal and the time-frequency mixed data to obtain normalized time-frequency voltage data and normalized time-frequency mixed data; input the normalized time-frequency voltage data and normalized time-frequency mixed data into a preset sensor circuit fault diagnosis model to obtain sensor circuit fault diagnosis results.

[0090] In one possible implementation, the sensor circuit fault diagnosis model includes a DNN neural network, an LSTM neural network, a fully connected layer network, and a SoftMax layer network. The DNN neural network is used to input normalized time-frequency mixed data and outputs DNN neural network output data to the fully connected layer network. The LSTM neural network is used to input normalized time-frequency voltage data and outputs LSTM neural network output data to the fully connected layer network. The fully connected layer network is used to perform fully connected processing on the DNN neural network output data and the LSTM neural network output data to obtain fully connected layer network output data, which is then output to the SoftMax layer network. The SoftMax layer network is used to obtain and output the sensor circuit fault diagnosis result based on the input fully connected layer network output data.

[0091] In one possible implementation, the feature extraction module is specifically used to: obtain one or more of the maximum value, minimum value, average value, standard deviation, kurtosis and skewness of the time-domain continuous voltage signal to obtain time-domain feature data; perform time-frequency conversion on the time-domain continuous voltage signal to obtain a frequency-domain continuous voltage signal; and obtain one or more of the bandwidth and center frequency of the frequency-domain continuous voltage signal to obtain frequency-domain feature data.

[0092] In one possible implementation, the SoftMax function of the SoftMax layer network is:

[0093]

[0094] Where fc_out(m)[i] is the i-th data point of the fully connected layer network output data of the m-th group of time-domain continuous voltage signals, fc_out(m)[j] is the j-th data point of the fully connected layer network output data of the m-th group of time-domain continuous voltage signals, m is the number of acquisition groups of time-domain continuous voltage signals, and K is the number of fault types of the sensor circuit; s(fc_out(m)[i]) is the i-th data point of the sensor circuit fault diagnosis result, i>0 indicates the probability that sensor circuit element i will fail, and i=0 indicates the probability that the sensor circuit will not fail.

[0095] All relevant content of each step involved in the aforementioned embodiments of the sensor circuit fault diagnosis method can be referenced to the functional description of the corresponding functional module of the sensor circuit fault diagnosis system in the embodiments of the present invention, and will not be repeated here.

[0096] The module division in this embodiment of the invention is illustrative and represents only one logical functional division. In actual implementation, other division methods may be used. Furthermore, the functional modules in the various embodiments of the invention can be integrated into a single processor, exist as separate physical entities, or be integrated into a single module. The integrated modules described above can be implemented as sensors or as software functional modules.

[0097] In another embodiment of the present invention, a sensor circuit fault diagnosis device is provided. This device includes a processor and a memory. The memory stores a computer program, which includes program instructions. The processor executes the program instructions stored in the computer storage medium. The processor may be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. It is the computing and control core of the terminal, suitable for implementing one or more instructions, specifically suitable for loading and executing one or more instructions from the computer storage medium to achieve a corresponding method flow or function. The processor described in this embodiment can be used for the operation of a sensor circuit fault diagnosis method.

[0098] In another embodiment of the present invention, a storage medium is provided, specifically a computer-readable storage medium (Memory), which is a memory device in a computer device used to store programs and data. It is understood that the computer-readable storage medium here can include both the built-in storage medium in the computer device and extended storage media supported by the computer device. The computer-readable storage medium provides storage space that stores the terminal's operating system. Furthermore, the storage space also stores one or more instructions suitable for loading and execution by a processor. These instructions can be one or more computer programs (including program code). It should be noted that the computer-readable storage medium here can be high-speed RAM or non-volatile memory, such as at least one disk storage device. The processor can load and execute one or more instructions stored in the computer-readable storage medium to implement the corresponding steps of the sensor circuit fault diagnosis method in the above embodiments.

[0099] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0100] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0101] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0102] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0103] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the protection scope of the claims of the present invention.

Claims

1. A sensor circuit failure diagnosis method characterized by, include: Acquire the continuous voltage signal in the time domain output by the sensor circuit; Time-domain and frequency-domain features are extracted from the continuous voltage signal in the time domain to obtain time-domain feature data and frequency-domain feature data. Based on the time-domain continuous voltage signal, time-domain feature data, and frequency-domain feature data, a preset sensor circuit fault diagnosis model is invoked to obtain the sensor circuit fault diagnosis result. The extraction of time-domain and frequency-domain features from the continuous voltage signal in the time domain to obtain time-domain feature data and frequency-domain feature data includes: Obtain one or more of the maximum, minimum, average, standard deviation, kurtosis, and skewness of the continuous voltage signal in the time domain to obtain time-domain characteristic data; The time-domain continuous voltage signal is converted to frequency to obtain the frequency-domain continuous voltage signal. One or more of the bandwidth and center frequency of the frequency-domain continuous voltage signal are obtained to obtain frequency domain characteristic data. The step of calling a preset sensor circuit fault diagnosis model based on the time-domain continuous voltage signal, time-domain feature data, and frequency-domain feature data to obtain sensor circuit fault diagnosis results includes: By combining time-domain feature data and frequency-domain feature data, time-frequency hybrid data is obtained; Both the continuous voltage signal in the time domain and the time-frequency mixed data are normalized to obtain normalized time-frequency voltage data and normalized time-frequency mixed data. Normalized time-frequency voltage data and normalized time-frequency mixed data are input into a preset sensor circuit fault diagnosis model to obtain sensor circuit fault diagnosis results. The preset sensor circuit fault diagnosis model includes DNN neural network, LSTM neural network, fully connected layer network and SoftMax layer network; DNN neural networks are used as inputs for normalized time-frequency mixed data and output data from DNN neural networks to fully connected layer networks. The LSTM neural network is used as input to normalized time-frequency voltage data and outputs the LSTM neural network output data to the fully connected layer network. Fully connected layer networks are used to process the output data of DNN neural networks and LSTM neural networks in a fully connected manner, and then output the fully connected layer network output data to the SoftMax layer network. The SoftMax layer network is used to obtain and output sensor circuit fault diagnosis results based on the input data from the fully connected layer network. The SoftMax function of the SoftMax layer network is: in, [ i ] is the first m The first group of time-domain continuous voltage signals output data from the fully connected layer network i One data point, [ j ] is the first m The first group of time-domain continuous voltage signals output data from the fully connected layer network j One data point, The number of groups of continuous voltage signals acquired in the time domain. K This represents the number of fault types in the sensor circuit. s ( [ i The first result of the sensor circuit fault diagnosis is... i One data point, i >0 indicates sensor circuit elements i The probability of failure. i =0 indicates the probability that the sensor circuit will not malfunction.

2. A sensor circuit failure diagnosis system based on the sensor circuit failure diagnosis method according to claim 1, characterized by include: The data acquisition module is used to acquire the time-domain continuous voltage signal output by the sensor circuit; The feature extraction module is used to extract time-domain and frequency-domain features from the continuous voltage signal in the time domain, and obtain time-domain feature data and frequency-domain feature data. The diagnostic module is used to call a preset sensor circuit fault diagnosis model based on the time-domain continuous voltage signal to obtain the sensor circuit fault diagnosis result.

3. A sensor circuit failure diagnosis apparatus comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized by, When the processor executes the computer program, it implements the steps of the sensor circuit fault diagnosis method as described in claim 1.

4. A computer-readable storage medium storing a computer program, the computer program comprising instructions that, when executed by a computer, cause the computer to perform the method of any one of claims 1 to 3. When the computer program is executed by the processor, it implements the steps of the sensor circuit fault diagnosis method as described in claim 1.

Images