Fault prediction and early warning system and method based on multi-sensor monitoring

Abstract

The invention discloses a fault prediction and early warning system and method based on multi-sensor monitoring, and belongs to the field of big data, and the method comprises the steps: deploying various types of sensors, collecting equipment operation data in real time, and forming a multi-dimensional original data set; preprocessing the data to remove abnormal values and missing values, removing noise in the sensor data by adopting a filter, converting the data of different dimensions into a unified range, and extracting time domain features and frequency domain features from the preprocessed data; fusing the features of different sensors to form a comprehensive feature vector; training a fault prediction model by using a machine learning algorithm based on the historical data; and inputting sensor data collected in real time into the trained model, calculating a fault occurrence probability, setting an early warning threshold value, and triggering early warning when the predicted fault probability exceeds the threshold value. The device operation data is collected in real time, potential faults are found in time, early warning is carried out in advance, the device shutdown time is shortened, and the production efficiency is improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of big data, and more specifically, particularly relates to a fault prediction and early warning system and method based on multi-sensor monitoring. Background Art

[0002] Data loss or anomalies caused by sensor failures or environmental interference affect the accuracy of the model. Sensor data may be affected by factors such as electromagnetic interference, resulting in noise and affecting the reliability of the data. Data from different types of sensors may have different time scales, frequencies, and accuracies. How to effectively fuse these multi-source data is a challenge. The data acquisition times of multi-sensors may be asynchronous. How to align and synchronize the data for effective analysis is a difficult point.

[0003] Multi-sensor data may lead to a too high dimensionality of the feature space, increasing the complexity of model training and inference. Complex models may perform well on training data, but may overfit in actual applications, resulting in poor generalization ability. Real-time fault prediction requires fast data processing and model inference, which may pose high requirements on computing resources. Delays in data acquisition, processing, and prediction may affect the timeliness of early warning and reduce the effectiveness of the system. The hardware of different sensors may have compatibility problems, affecting the stability and reliability of the system. The multi-sensor system needs regular maintenance and calibration to ensure the accuracy of the data and the long-term stable operation of the system. Summary of the Invention

[0004] The purpose of this part is to outline some aspects of the embodiments of the present invention and briefly introduce some preferred embodiments. Some simplifications or omissions may be made in this part, as well as in the abstract and title of the specification of this application, to avoid obscuring the purpose of this part, the abstract, and the title. However, such simplifications or omissions shall not be used to limit the scope of the present invention.

[0005] In view of the above problems of the existing fault prediction and early warning methods based on multi-sensor monitoring, the present invention is proposed.

[0006] To solve the above technical problems, the present invention provides the following technical solutions:

[0007] An embodiment of the present invention provides a fault prediction and early warning method based on multi-sensor monitoring, including: deploying various types of sensors to collect device operation data in real time to form a multi-dimensional original data set;

[0008] Preprocessing the data to remove outliers and missing values, using a filter to remove noise in the sensor data, converting data with different dimensions to a unified range, and extracting time-domain features and frequency-domain features from the preprocessed data;

[0009] Fuse the features of different sensors to form a comprehensive feature vector;

[0010] Based on historical data, use machine learning algorithms to train a fault prediction model;

[0011] Input the sensor data collected in real time into the trained model, calculate the probability of a fault occurring, set a warning threshold, and trigger a warning when the predicted fault probability exceeds the threshold.

[0012] As a preferred solution of the fault prediction and warning method based on multi-sensor monitoring according to the present invention, wherein: deploy multiple types of sensors to collect device operation data in real time to form a multi-dimensional original data set, including:

[0013] Install temperature sensors at overheated parts and vibration sensors at rotating component positions, design a multi-channel data acquisition system to process data from different sensors simultaneously, support the simultaneous access of multiple sensors and each channel independently acquires data, ensuring that the data of all sensors is acquired at the same time point for data fusion and comparison.

[0014] As a preferred solution of the fault prediction and warning method based on multi-sensor monitoring according to the present invention, wherein: preprocess the data to remove outliers and missing values, use a filter to remove noise in the sensor data, and convert data with different dimensions to a unified range, including:

[0015] For data points that deviate from other observations, calculate the IQR based on the quartiles of the data and determine the outlier range:

[0016] IQR = Q3 - Q1

[0017] Lower bound = Q1 - 1.5 × IQR

[0018] Upper bound = Q3 + 1.5 × IQR

[0019] Wherein, Q1 and Q3 are the first and third quartiles respectively, and data points outside the upper and lower bounds are regarded as outliers;

[0020] When there are missing observations in the data set, fill the missing values with the mean value of this feature:

[0021]

[0022] Wherein, X new is the mean value used to fill the missing values, n is the number of non-missing values in the feature, and X i is the i-th non-missing value in the feature;

[0023] For noise, perform state estimation of the dynamic system through Kalman filtering:

[0024] Prediction step:

[0025] X k|k-1 = AX k-1|k-1 + Bu k-1

[0026] P k|k-1 = AP k-1|k-1 A T + Q

[0027] Update step:

[0028] K k = P k|k-1 H T (HP k|k-1 H T + R) -1

[0029] X k|k = X k|k-1 + K k (Z k - HX k|k-1 )

[0030] P k|k = (I - K k H)P k|k-1

[0031] Wherein, X is the state estimate, P is the estimation error covariance, A is the state transition matrix, B is the control matrix, u is the control input, Q is the process noise covariance, K is the Kalman gain, H is the observation matrix, R is the observation noise covariance, and Z is the observed value.

[0032] As a preferred solution of the fault prediction and early warning method based on multi - sensor monitoring according to the present invention, wherein: extracting time - domain features and frequency - domain features from the pre - processed data includes:

[0033] The time - domain features are calculated from time - series data. The standard deviation measures the dispersion of the signal amplitude relative to the mean, reflecting the volatility of the signal. Its expression is as follows:

[0034]

[0035] Wherein, N is the total number of samples, x i is the i - th sample value, and Mean is the mean of the signal;

[0036] The frequency - domain features are analyzed by converting the time - series signal to the frequency domain. Based on the Fourier transform, the time - domain signal is converted into a frequency - domain representation to analyze the amplitudes and phases of each frequency component:

[0037]

[0038] where X(f) is a complex value at frequency f, representing the amplitude and phase of that frequency component, N is the total number of samples, x n is the nth time-domain sample value, and j is the imaginary unit.

[0039] As a preferred solution of the fault prediction and early warning method based on multi-sensor monitoring according to the present invention, wherein: the step of fusing the features of different sensors to form a comprehensive feature vector includes:

[0040] Extract features such as mean, standard deviation, and spectral features from the preprocessed data of each sensor, align the features of different sensors in time and space, and perform normalization on the features to eliminate the influence of dimensional differences. Min-max normalization:

[0041]

[0042] where x is the original feature value, x min is the minimum value of the feature, x max is the maximum value of the feature, and x' is the normalized feature value;

[0043] Standardization:

[0044]

[0045] where x is the original feature value, μ is the mean of the feature, σ is the standard deviation of the feature, and x' is the standardized feature value;

[0046] Fuse the aligned and normalized features to form a comprehensive feature vector, and perform weighted averaging on the features of each sensor according to the weight ratio:

[0047]

[0048] where F is the fused feature vector, n is the number of sensors, wi is the weight of the ith sensor, and F i is the feature vector of the ith sensor;

[0049] Reduce the dimension of the features, extract the main components, and form a new feature vector:

[0050] F fused = W T F

[0051] where F fused is the feature vector after dimensionality reduction, W is the projection matrix of PCA, obtained by eigenvalue decomposition, and F is the original feature matrix.

[0052] As a preferred solution of the fault prediction and early warning method based on multi-sensor monitoring according to the present invention, wherein: based on historical data, a machine learning algorithm is used to train a fault prediction model:

[0053] According to historical records, the data is divided into two categories: normal and faulty, and labels are generated. If a specific type of fault is to be predicted, different fault types are classified and labeled;

[0054] The data set is divided into a training set, a validation set, and a test set, and a long short-term memory network is selected for model training. The forward propagation process of the model is as follows:

[0055] Input gate:

[0056] i t = σ(W i · [h t-1 , x t + b i )

[0057] wherein, i t is the input gate vector at time t, σ is the activation function, W i is the weight matrix of the input gate, h t-1 is the hidden state vector at the previous time, x t is the input vector at the current time, b i is the bias vector of the input gate;

[0058] Forget gate:

[0059] f t = σ(W f · [h t-1 , x t + b f )

[0060] wherein, f t is the forget gate vector at time t, W f is the weight matrix of the forget gate, b f is the bias vector of the forget gate;

[0061] Output gate:

[0062] o t = σ(W o · [h t-1 , x t + b o )

[0063] wherein, o t is the output gate vector at time t, Wo is the weight matrix of the output gate, and bo is the bias vector of the output gate.

[0064] As a preferred solution of the fault prediction and early warning method based on multi-sensor monitoring according to the present invention, wherein: inputting the sensor data collected in real time into the trained model, calculating the probability of a fault occurring, setting an early warning threshold, and triggering an early warning when the predicted fault probability exceeds the threshold, including:

[0065] Input the extracted feature vectors into the trained neural network model, set a fault probability threshold θ according to historical data and requirements, optimize the threshold through cross-validation, and trigger an early warning when the fault probability P output by the model exceeds the threshold θ.

[0066] A fault prediction and early warning system based on multi-sensor monitoring, including: a data acquisition module for deploying various types of sensors to collect device operation data in real time and form a multi-dimensional original data set; a feature extraction module for preprocessing the data to remove outliers and missing values, using a filter to remove noise in the sensor data, converting data with different dimensions to a unified range, and extracting time-domain features and frequency-domain features from the preprocessed data; a vector fusion module for fusing the features of different sensors to form a comprehensive feature vector; a model training module for training a fault prediction model based on historical data using a machine learning algorithm; a fault early warning module for inputting the sensor data collected in real time into the trained model, calculating the probability of a fault occurring, setting an early warning threshold, and triggering an early warning when the predicted fault probability exceeds the threshold.

[0067] A computing device, the computing device includes:

[0068] At least one processor, a memory, and an input / output unit;

[0069] Wherein, the memory is used to store a computer program, and the processor is used to call the computer program stored in the memory to execute the steps of the fault prediction and early warning method based on multi-sensor monitoring.

[0070] A computer-readable storage medium, which includes instructions that, when running on a computer, cause the computer to execute the steps of fault prediction and early warning based on multi-sensor monitoring.

[0071] The beneficial effects of the present invention are: The present invention collects device operation data in real time, discovers potential faults in time, gives early warnings, reduces device downtime, and improves production efficiency. By predicting faults in advance, maintenance can be carried out before the faults occur, avoiding high maintenance costs and production losses caused by sudden faults. It helps enterprises reasonably arrange maintenance personnel and resources, avoid over-maintenance or resource waste, and achieve the optimal allocation of resources. Through continuous monitoring and early warning, potential problems of the device can be discovered in time, targeted maintenance can be carried out, and the overall reliability and service life of the device can be improved. Description of the Drawings

[0072] To more clearly illustrate the technical solutions of the embodiments of the present invention, the following will briefly introduce the accompanying drawings required for the description of the embodiments. Obviously, the accompanying drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other accompanying drawings can be obtained based on these drawings.

[0073] Figure 1 It is a flowchart of a fault prediction and early warning method based on multi-sensor monitoring provided by an embodiment of the present invention.

[0074] Figure 2 It is a schematic structural diagram of a fault prediction and early warning system based on multi-sensor monitoring provided by an embodiment of the present invention.

[0075] Figure 3 Schematically shows a schematic structural diagram of a medium of an embodiment of the present invention.

[0076] Figure 4 Schematically shows a schematic structural diagram of a computing device of an embodiment of the present invention.

[0077] In the accompanying drawings, the same or corresponding reference numerals indicate the same or corresponding parts. Detailed implementation manners

[0078] To make the above objects, features, and advantages of the present invention more obvious and understandable, the following will make a detailed description of the specific implementation manners of the present invention in conjunction with the accompanying drawings of the specification.

[0079] In the following description, many specific details are set forth to facilitate a thorough understanding of the present invention. However, the present invention can also be implemented in other ways different from those described herein. Those skilled in the art can make similar extensions without departing from the connotation of the present invention. Therefore, the present invention is not limited by the specific embodiments disclosed below.

[0080] Secondly, the so-called "one embodiment" or "embodiment" herein refers to a specific feature, structure, or characteristic that can be included in at least one implementation manner of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment that excludes other embodiments.

[0081] Embodiment

[0082] The following refers to Figure 1 , Figure 1 It is a flowchart of a fault prediction and early warning method based on multi-sensor monitoring provided by an embodiment of the present invention. It should be noted that the implementation manners of the present invention can be applied to any applicable scenario.

[0083] Figure 1 The flow of the fault prediction and early warning method based on multi-sensor monitoring provided by an embodiment of the present invention shown in the figure includes:

[0084] S1: Deploy multiple types of sensors to collect device operation data in real time and form a multi-dimensional original data set.

[0085] Preferably, install temperature sensors at overheated parts and vibration sensors at rotating component positions. Design a multi-channel data acquisition system to process data from different sensors simultaneously, support the simultaneous access of multiple sensors, and each channel independently collects data to ensure that data from all sensors is collected at the same time point for data fusion and comparison.

[0086] S2: Preprocess the data to remove outliers and missing values, use a filter to remove noise in the sensor data, convert data with different dimensions to a unified range, and extract time-domain features and frequency-domain features from the preprocessed data.

[0087] Preferably, for data points that deviate from other observed values, calculate the IQR based on the quartiles of the data and determine the outlier range:

[0088] IQR = Q3 - Q1

[0089] Lower bound = Q1 - 1.5 × IQR

[0090] Upper bound = Q3 + 1.5 × IQR

[0091] where Q1 and Q3 are the first and third quartiles respectively, and data points outside the upper and lower bounds are regarded as outliers;

[0092] When there are missing observations in the data set, fill the missing values with the mean of this feature:

[0093]

[0094] where X new is the mean used to fill the missing values, n is the number of non-missing values in the feature, and X i is the i-th non-missing value in the feature;

[0095] For noise, perform state estimation of the dynamic system through Kalman filtering:

[0096] Prediction step:

[0097] X k|k-1 = AX k-1|k-1 + Bu k-1

[0098] P k|k-1 = APk-1|k-1 A T +Q

[0099] Update steps:

[0100] K k = P k|k-1 H T (HP k|k-1 H T + R) -1

[0101] X k|k = X k|k-1 + K k (Z k - HX k|k-1 )

[0102] P k|k = (I - K k H)P k|k-1

[0103] Wherein, X is the state estimate, P is the estimation error covariance, A is the state transition matrix, B is the control matrix, u is the control input, Q is the process noise covariance, K is the Kalman gain, H is the observation matrix, R is the observation noise covariance, and Z is the observed value.

[0104] Preferably, the time-domain features are calculated from the time series data. The standard deviation measures the dispersion of the signal amplitude relative to the mean and reflects the volatility of the signal. Its expression is as follows:

[0105]

[0106] Wherein, N is the total number of samples, x i is the i-th sample value, and Mean is the mean of the signal;

[0107] The frequency-domain features are analyzed by converting the time series signal to the frequency domain. Based on the Fourier transform, the time-domain signal is converted to the frequency-domain representation to analyze the amplitude and phase of each frequency component:

[0108]

[0109] Wherein, X(f) is the complex value at frequency f, representing the amplitude and phase of the frequency component, N is the total number of samples, x n is the n-th time-domain sample value, and j is the imaginary unit.

[0110] Furthermore, assuming that the Kalman filter is to be used to remove the noise of the temperature sensor, the following assumptions are made:

[0111] The state transition matrix A = 1 (assuming that the state does not change over time)

[0112] Control matrix B = 0 (assuming no external control input)

[0113] Process noise covariance Q = 0.1

[0114] Observation matrix H = 1 (sensor directly observes the state)

[0115] Observation noise covariance R = 0.5

[0116] Assume that at time step k - 1, the estimated state of the system is Xk - 1 = 30, and the error covariance is Pk - 1 = 2.

[0117] Prediction step:

[0118] X_k^- = A × X_k - 1 + B × u_k = 30

[0119] P_k^- = A × P_k - 1 × A^T + Q = 2 + 0.1 = 2.1

[0120] Update step: Assume the observation value Z_k = 32

[0121] K_k = 2.1 / (2.1 + 0.5) = 0.808

[0122] X_k = 30 + 0.808 × (32 - 30) = 30 + 0.808 × 2 = 31.616

[0123] P_k = (1 - 0.808) × 2.1 = 0.392

[0124] The updated estimated state is X_k = 31.616, and the error covariance is P_k = 0.392;

[0125] Furthermore, assume we have a periodic signal: xn = sin(2πf 0 n / N),

[0126] where, f 0 is the frequency of the signal. Assume the signal frequency is 2Hz and the number of samples N = 100;

[0127] Assume the sampling frequency f s = 10Hz. For the given frequency f 0 = 2Hz, the amplitude of the signal at this frequency will exhibit an obvious peak, while other frequency components will be relatively small;

[0128] By calculating the complex magnitude of the Fourier transform result, the amplitude of each frequency point can be obtained:

[0129]

[0130] The amplitude of the corresponding frequency point represents the intensity of that frequency component, and the calculation is as follows: Assume the discrete form of the signal is:

[0131] x n = sin(2π × 2 × n / 100)

[0132] Then, perform a discrete Fourier transform on it to obtain the frequency-domain representation of the signal. In the frequency domain, the 2 Hz frequency will appear as an obvious peak.

[0133] S3: Fuse the features of different sensors to form a comprehensive feature vector.

[0134] Preferably, extract features such as the mean, standard deviation, and spectral features from the preprocessed data of each sensor, align the features of different sensors in time and space, and perform normalization on the features to eliminate the influence of dimensional differences. Min-max normalization:

[0135]

[0136] where x is the original feature value, x min is the minimum value of the feature, x max is the maximum value of the feature, and x' is the normalized feature value;

[0137] Standardization:

[0138]

[0139] where x is the original feature value, μ is the mean of the feature, σ is the standard deviation of the feature, and x' is the standardized feature value;

[0140] Fuse the aligned and normalized features to form a comprehensive feature vector, and perform weighted averaging on the features of each sensor according to the weight ratio:

[0141]

[0142] where F is the fused feature vector, n is the number of sensors, wi is the weight of the i-th sensor, and F i is the feature vector of the i-th sensor;

[0143] Reduce the dimension of the features, extract the main components, and form a new feature vector:

[0144] F fused = W T F

[0145] where F fused is the feature vector after dimension reduction, W is the projection matrix of PCA, obtained by eigenvalue decomposition, and F is the original feature matrix.

[0146] Furthermore, assume there are two sensors that measure the temperature and vibration data of the device respectively. The data from these two sensors will be preprocessed, feature extracted, normalized, and fused, and finally a comprehensive fault prediction feature vector will be obtained.

[0147] Step 1: Feature Extraction

[0148] Sensor 1 (temperature) data: 23.5, 24.1, 24.3, 24.6, 25.1, 24.9;

[0149] Mean: Mean = 24.4

[0150] Standard deviation: σ ≈ 0.305

[0151] Frequency domain feature (Fourier transform): Analyze based on frequency components

[0152] Sensor 2 (vibration):

[0153] Data: 0.01, 0.02, 0.03, 0.04, 0.05, 0.03

[0154] Mean: Mean = 0.03

[0155] Standard deviation: σ ≈ 0.015

[0156] Frequency domain feature (Fourier transform): Analyze based on frequency components

[0157] Assume the maximum value of temperature is 25.1 and the minimum value is 23.5; the maximum value of vibration is 0.05 and the minimum value is 0.01.

[0158] Normalize the temperature data:

[0159]

[0160] Normalize the vibration data:

[0161]

[0162] Assume the weight of the temperature sensor is 0.7 and the weight of the vibration sensor is 0.3. We will perform weighted averaging of the features of the two sensors to obtain the fused feature vector:

[0163] F = 0.7·F temp + 0.3·F vib

[0164] Assume PCA is used to reduce the features to 1 dimension, and the obtained feature vector after dimensionality reduction is:

[0165] F fused = W T F.

[0166] S4: Train a fault prediction model using a machine learning algorithm based on historical data.

[0167] Preferably, according to historical records, the data is divided into two categories: normal and faulty, and labels are generated. If a specific type of fault is to be predicted, different fault types are classified and labeled.

[0168] The data set is divided into a training set, a validation set, and a test set. A long short-term memory network is selected for model training. The forward propagation process of the model is as follows:

[0169] Input gate:

[0170] i t = σ(W i · [h t-1 , x t + b i )

[0171] where i t is the input gate vector at time t, σ is the activation function, W i is the weight matrix of the input gate, h t-1 is the hidden state vector at the previous time, x t is the input vector at the current time, and b i is the bias vector of the input gate;

[0172] Forget gate:

[0173] f t = σ(W f · [h t-1 , x t + b f )

[0174] where f t is the forget gate vector at time t, W f is the weight matrix of the forget gate, and b f is the bias vector of the forget gate;

[0175] Output gate:

[0176] o t = σ(W o · [h t-1 , x t + b o )

[0177] where o t is the output gate vector at time t, Wo is the weight matrix of the output gate, and bo is the bias vector of the output gate.

[0178] Furthermore, the LSTM model is trained using the training set data, and the network parameters are optimized through the backpropagation algorithm. Usually, the cross-entropy loss function is adopted for binary classification (normal / fault) training.

[0179] During the training process, a validation set is used for validation, and hyperparameters (such as learning rate, batch size, number of LSTM layers, etc.) are adjusted to better fit the training data.

[0180] The trained model is evaluated using the test set, and metrics such as accuracy, recall, F1-score, etc. of the prediction results are calculated to evaluate the performance of the model. If the goal is to predict multiple fault types, a multi-classification loss function is used and the recognition ability of each fault category is evaluated.

[0181] Real-time data input: When the device is running, the sensor data collected in real time will be input into the trained LSTM model. The model will predict the device state at the current moment based on the feature patterns learned from the historical data. If the device is operating within the normal range, the model will output a prediction result of the "normal" class (e.g., 0). If the device has a fault (such as overheating, abnormal vibration, etc.), the model will output the corresponding fault type (e.g., 1 represents overheating, 2 represents abnormal vibration).

[0182] Fault probability calculation: For each input sensor data, the model will calculate the occurrence probability of each fault type. Based on the Softmax activation function (for multi-classification problems), the output value of each fault type can be converted into a probability value, indicating the probability of the occurrence of that fault type.

[0183] S5 inputs the sensor data collected in real time into the trained model, calculates the probability of fault occurrence, sets an early warning threshold, and triggers an early warning when the predicted fault probability exceeds the threshold.

[0184] Preferably, the extracted feature vectors are input into the trained neural network model. According to the historical data and requirements, a fault probability threshold θ is set, and the threshold is optimized through the cross-validation method. When the fault probability P output by the model exceeds the threshold θ, an early warning is triggered.

[0185] Furthermore, appropriate training sets and validation sets are selected, and the model prediction results of each validation set are compared with the actual labels to calculate the prediction accuracy and error.

[0186] By adjusting the fault probability P output by the model and the threshold θ, the classification accuracy of the model is gradually optimized.

[0187] During the optimization process, different θ values are tried to find the most suitable value, and this process ensures that the model can predict faults as accurately as possible.

[0188] Input the test data into the trained neural network, and the neural network outputs a fault probability P according to the currently input sensor data.

[0189] For example, when the inputs are features such as temperature and vibration, the network outputs a probability value regarding whether a fault has occurred. If the output fault probability P exceeds the set threshold θ, the warning mechanism is triggered.

[0190] When the fault probability exceeds the threshold, the system immediately alarms and notifies relevant personnel, indicating the risk of possible faults.

[0191] After introducing the method of the exemplary embodiment of the present invention, next, refer to Figure 2 A fault prediction and warning system based on multi-sensor monitoring according to an exemplary embodiment of the present invention will be described. The system includes:

[0192] A data acquisition module for deploying various types of sensors to collect device operation data in real time and form a multi-dimensional original data set;

[0193] A feature extraction module for preprocessing the data to remove outliers and missing values, using a filter to remove noise in the sensor data, converting data with different dimensions to a unified range, and extracting time-domain features and frequency-domain features from the preprocessed data;

[0194] A vector fusion module for fusing the features of different sensors to form a comprehensive feature vector;

[0195] A model training module for training a fault prediction model based on historical data using machine learning algorithms;

[0196] A fault warning module for inputting the sensor data collected in real time into the trained model, calculating the probability of a fault occurring, setting a warning threshold, and triggering a warning when the predicted fault probability exceeds the threshold.

[0197] After introducing the method and device of the exemplary embodiment of the present invention, next, refer to Figure 3 A computer-readable storage medium according to an exemplary embodiment of the present invention will be described. Please refer to Figure 3, which shows that the computer-readable storage medium is an optical disc 30, on which a computer program (i.e., program product) is stored. When the computer program is run by a processor, it will implement the steps recorded in the above method embodiments. For example, deploying multiple types of sensors to collect device operation data in real time to form a multi-dimensional original data set; preprocessing the data to remove outliers and missing values, using a filter to remove noise in the sensor data, converting data with different dimensions to a unified range, and extracting time-domain features and frequency-domain features from the preprocessed data; fusing the features of different sensors to form a comprehensive feature vector; based on historical data, using a machine learning algorithm to train a fault prediction model; inputting the sensor data collected in real time into the trained model, calculating the probability of a fault occurring, setting an early warning threshold, and triggering an early warning when the predicted fault probability exceeds the threshold; the specific implementation methods of each step will not be repeated here.

[0198] It should be noted that examples of the computer-readable storage medium may also include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory, or other optical and magnetic storage media, which will not be elaborated one by one here.

[0199] After introducing the methods, devices, and media of the exemplary embodiments of the present invention, next, refer to Figure 4 a computing device for fault prediction and early warning based on multi-sensor monitoring of the exemplary embodiments of the present invention.

[0200] Figure 4 FIG. shows a block diagram of an exemplary computing device 40 suitable for implementing the embodiments of the present invention. The computing device 40 may be a computer system or a server. Figure 4 The shown computing device 40 is merely an example and should not impose any limitation on the functions and usage scope of the embodiments of the present invention.

[0201] As Figure 4 shown, the components of the computing device 40 may include, but are not limited to: one or more processors or processing units 401, a system memory 402, and a bus 403 connecting different system components (including the system memory 402 and the processing unit 401).

[0202] The computing device 40 typically includes a variety of computer system-readable media. These media can be any available media accessible by the computing device 40, including volatile and non-volatile media, removable and non-removable media.

[0203] System memory 402 may include computer system readable media in the form of volatile memory, such as random access memory (RAM) 4021 and / or cache memory 4022. Computing device 40 may further include other removable / non-removable, volatile / non-volatile computer system storage media. By way of example only, ROM 4023 may be used to read from and write to a non-removable, non-volatile magnetic medium ( Figure 4 not shown in the figure, commonly referred to as a "hard disk drive"). Although not shown in Figure 4 the figure, a disk drive for reading from and writing to a removable non-volatile disk (such as a "floppy disk"), and an optical disk drive for reading from and writing to a removable non-volatile optical disk (such as a CD-ROM, DVD-ROM or other optical medium) may be provided. In these cases, each drive may be connected to bus 403 through one or more data media interfaces. System memory 402 may include at least one program product having a set (e.g., at least one) of program modules that are configured to perform the functions of the various embodiments of the present invention.

[0204] A program / utilities 4025 having a set (at least one) of program modules 4024 may be stored, for example, in system memory 402, and such program modules 4024 include, but are not limited to: an operating system, one or more application programs, other program modules, and program data, and an implementation of a network environment may be included in each or some combination of these examples. Program modules 4024 generally perform the functions and / or methods in the embodiments described in the present invention.

[0205] Computing device 40 may also communicate with one or more external devices 404 (such as a keyboard, a pointing device, a display, etc.). Such communication may be through an input / output (I / O) interface 405. Also, computing device 40 may further communicate with one or more networks (such as a local area network (LAN), a wide area network (WAN), and / or a public network, such as the Internet) through a network adapter 406. As Figure 4 shown, network adapter 406 communicates with other modules (such as processing unit 401, etc.) of computing device 40 through bus 403. It should be understood that although Figure 4 not shown in the figure, other hardware and / or software modules may be used in conjunction with computing device 40.

[0206] The processing unit 401 executes various functional applications and data processing by running programs stored in the system memory 402. For example, it deploys various types of sensors to collect device operation data in real time, forming a multi-dimensional raw data set; preprocesses the data to remove outliers and missing values, uses filters to remove noise in the sensor data, converts data with different dimensions to a unified range, and extracts time-domain features and frequency-domain features from the preprocessed data; fuses the features of different sensors to form a comprehensive feature vector; trains a fault prediction model using machine learning algorithms based on historical data; inputs the sensor data collected in real time into the trained model, calculates the probability of a fault occurring, sets a warning threshold, and triggers a warning when the predicted fault probability exceeds the threshold.

[0207] The specific implementation methods of each step will not be repeated here. It should be noted that although several units / modules or sub-units / sub-modules of the synchronous escape wiring device based on multi-commodity flow are mentioned in the above detailed description, this division is only exemplary and not mandatory. In fact, according to the embodiments of the present invention, the features and functions of two or more units / modules described above can be embodied in one unit / module. Conversely, the features and functions of one unit / module described above can be further divided and embodied by multiple units / modules.

[0208] In the description of the present invention, it should be noted that the terms "first", "second", and "third" are only used for descriptive purposes and cannot be understood as indicating or implying relative importance.

[0209] Those skilled in the art can clearly understand that for the convenience and brevity of description, the specific working processes of the above-described system, device, and unit can refer to the corresponding processes in the foregoing method embodiments and will not be elaborated here.

[0210] In several embodiments provided by the present invention, it should be understood that the disclosed system, device, and method can be implemented in other ways. The device embodiments described above are only illustrative. For example, the division of the units is only a logical function division, and there can be other division methods in actual implementation. For another example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the displayed or discussed couplings or direct couplings or communication connections to each other can be through some communication interfaces. The indirect couplings or communication connections of the devices or units can be in electrical, mechanical, or other forms.

[0211] The unit described as a separation component may or may not be physically separated. The component shown as a unit may or may not be a physical unit, that is, it may be located in one place or distributed to multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0212] In addition, in each embodiment of the present invention, each functional unit can be integrated in a processing unit, or each unit can exist physically alone, or two or more units can be integrated in one unit.

[0213] If the described function is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a non-volatile computer-readable storage medium executable by a processor. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or a part of this 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 enable a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in each embodiment of the present invention. The aforementioned storage medium includes: various media such as USB flash drives, mobile hard disks, read-only memories (ROM, Read-Only Memory), random access memories (RAM, Random Access Memory), magnetic disks, or optical discs that can store program codes.

[0214] Finally, it should be noted that: the above-described embodiments are only specific implementation manners of the present invention, used to illustrate the technical solutions of the present invention, and are not intended to limit it. The protection scope of the present invention is not limited thereto. Although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that: any person skilled in the art within the technical scope disclosed by the present invention can still modify the technical solutions recorded in the foregoing embodiments, or can easily think of changes, or perform equivalent replacements on some of the technical features; and these modifications, changes, or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be covered by the protection scope of the present invention. Therefore, the protection scope of the present invention should be subject to the protection scope of the claims.

[0215] In addition, although the operations of the method of the present invention are described in a specific order in the drawings, this does not require or imply that these operations must be performed in this specific order, or that all the operations shown must be performed to achieve the desired result. Additionally or alternatively, some steps can be omitted, multiple steps can be combined into one step for execution, and / or one step can be decomposed into multiple steps for execution.

Claims

1. A fault prediction and early warning method based on multi-sensor monitoring, characterized in that: include: Deploy various types of sensors to collect equipment operation data in real time and form multi-dimensional raw data sets; Preprocess the data to remove outliers and missing values, use filters to remove noise from sensor data, convert data of different dimensions to a unified range, and extract time domain features and frequency domain features from the preprocessed data; Fuse the features of different sensors to form a comprehensive feature vector; Based on historical data, a fault prediction model is trained using machine learning algorithms; The sensor data collected in real time is input into the trained model, the probability of failure is calculated, and the warning threshold is set. When the predicted failure probability exceeds the threshold, the warning is triggered.

2. The fault prediction and early warning method based on multi-sensor monitoring according to claim 1, characterized in that: The deployment of various types of sensors collects equipment operation data in real time to form a multi-dimensional raw data set, including: The temperature sensor is installed at the location where overheating occurs, and the vibration sensor is installed at the location of the rotating part. A multi-channel data acquisition system is designed to process data from different sensors at the same time, support the simultaneous access of multiple sensors and independent data collection for each channel, and ensure that the data of all sensors are collected at the same time point for data fusion and comparison.

3. The fault prediction and early warning method based on multi-sensor monitoring according to claim 1, characterized in that: The data is preprocessed to remove outliers and missing values, a filter is used to remove noise in the sensor data, and data of different dimensions are converted to a uniform range, including: For data points that deviate from other observations, the IQR is calculated based on the quartiles of the data and the outlier range is determined: IQR=Q3-Q1 Lower bound = Q1-1.5×IQR Upper bound = Q3 + 1.5 × IQR Among them, Q1 and Q3 are the first and third quartiles, respectively, and data points beyond the upper and lower bounds are considered outliers; When there are missing observations in the dataset, fill the missing values ​​with the mean of the feature: Among them, X new is the mean used to fill missing values, n is the number of non-missing values ​​in the feature, X i is the i-th non-missing value in the feature; For noisy dynamic systems, the state estimation is performed by Kalman filtering: Prediction steps: X k|k-1 =AX k-1|k-1 +Bu k-1 P k|k-1 =AP k-1|k-1 From T +Q Update steps: K k =P k|k-1 H T (HP k|k-1 H T +R) -1 X k|k =X k|k-1 +K k (Z k -HX k|k-1 ) P k|k =(I-K k H)P k|k-1 Among them, X is the state estimate, P is the estimation error covariance, A is the state transfer matrix, B is the control matrix, u is the control input, Q is the process noise covariance, K is the Kalman gain, H is the observation matrix, R is the observation noise covariance, and Z is the observation value.

4. The fault prediction and early warning method based on multi-sensor monitoring according to claim 1, characterized in that: The step of extracting time domain features and frequency domain features from the preprocessed data includes: The time domain features are calculated from the time series data. The standard deviation measures the dispersion of the signal amplitude relative to the mean and reflects the volatility of the signal. Its expression is as follows: Where N is the total number of samples, x i is the i-th sample value, Mean is the mean of the signal; Frequency domain features are analyzed by converting time series signals into frequency domain. Based on Fourier transform, the time domain signal is converted into frequency domain representation, and the amplitude and phase of each frequency component are analyzed: Where X(f) is the complex value at frequency f, representing the amplitude and phase of the frequency component, N is the total number of samples, and x n is the nth time domain sample value, and j is the imaginary unit.

5. The fault prediction and early warning method based on multi-sensor monitoring according to claim 1, characterized in that: The features of different sensors are integrated to form a comprehensive feature vector, including: Extract features such as mean, standard deviation, and spectral features from the preprocessed data of each sensor, align the features of different sensors in time and space, and normalize the features to eliminate the impact of dimensional differences. Minimum-maximum normalization: Among them, x is the original eigenvalue, x min is the minimum value of the feature, x max is the maximum value of the feature, and x' is the normalized feature value; standardization: Among them, x is the original eigenvalue, μ is the mean of the feature, σ is the standard deviation of the feature, and x' is the standardized eigenvalue; The aligned and normalized features are fused to form a comprehensive feature vector, and the features of each sensor are weighted averaged according to the weight ratio: Among them, F is the fused feature vector, n is the number of sensors, wi is the weight of the i-th sensor, and F i is the feature vector of the i-th sensor; Reduce the dimension of the features, extract the main components, and form a new feature vector: F fused =W T F Among them, F fused is the eigenvector after dimensionality reduction, W is the projection matrix of PCA, obtained by eigenvalue decomposition, and F is the original feature matrix.

6. The fault prediction and early warning method based on multi-sensor monitoring according to claim 1, characterized in that: Based on historical data, the fault prediction model is trained using a machine learning algorithm: According to historical records, the data is divided into normal and faulty categories, and labels are generated. If a specific type of fault is to be predicted, different fault types are classified and labeled; The data set is divided into training set, validation set and test set, and the long short-term memory network is selected for model training. The forward propagation process of the model is as follows: Input Gate: i t =σ(W i ·[h t-1 ,x t ]+b i ) Among them, i t is the input gate vector at time t, σ is the activation function, W i is the weight matrix of the input gate, h t-1 is the hidden state vector at the previous moment, x t is the input vector at the current moment, b i is the bias vector of the input gate; Forget Gate: f t =σ(W f ·[h t-1 ,x t ]+b f ) Among them, f t is the forget gate vector at time t, W f is the weight matrix of the forget gate, b f is the bias vector of the forget gate; Output Gate: the t =σ(W o ·[h t-1 ,x t ]+b o ) Among them, t is the output gate vector at time t, Wo is the weight matrix of the output gate, and bo is the bias vector of the output gate.

7. The fault prediction and early warning method based on multi-sensor monitoring according to claim 1, characterized in that: The sensor data collected in real time is input into the trained model, the probability of failure is calculated, and the warning threshold is set. When the predicted failure probability exceeds the threshold, the warning is triggered, including: The extracted feature vector is input into the trained neural network model. According to historical data and requirements, a fault probability threshold θ is set. The threshold is optimized through cross-validation method. When the fault probability P output by the model exceeds the threshold θ, an early warning is triggered.

8. A fault prediction and early warning system based on multi-sensor monitoring, characterized in that: include: The data acquisition module is used to deploy various types of sensors to collect equipment operation data in real time and form a multi-dimensional raw data set; Feature extraction module, which is used to preprocess the data to remove outliers and missing values, use filters to remove noise in sensor data, convert data of different dimensions to a unified range, and extract time domain features and frequency domain features from the preprocessed data; Vector fusion module, used to fuse the features of different sensors to form a comprehensive feature vector; Model training module, used to train fault prediction models using machine learning algorithms based on historical data; The fault warning module is used to input the real-time collected sensor data into the trained model, calculate the probability of fault occurrence, set the warning threshold, and trigger the warning when the predicted fault probability exceeds the threshold.

9. A computing device, comprising: at least one processor, memory, and input-output unit; The memory is used to store a computer program, and the processor is used to call the computer program stored in the memory to execute the steps of the fault prediction and early warning method based on multi-sensor monitoring as described in any one of claims 1 to 7.

10. A computer-readable storage medium comprising instructions, which, when executed on a computer, enables the computer to execute the steps of the fault prediction and early warning method based on multi-sensor monitoring as claimed in any one of claims 1 to 7.

Images