Early warning platform based on pressure sensor

Abstract

The invention discloses an early warning platform based on a pressure sensor, and the platform comprises a signal collection module which is used for collecting a current signal on the surface of equipment in real time through the pressure sensor; the signal processing module is used for converting the current signal into a digital signal through an analog-to-digital converter to obtain an electric quantity resistance value information data set; the data processing module is used for carrying out data integration on the electric quantity and resistance value information data of the electric quantity and resistance value information data set and then carrying out data processing and analysis to obtain a target data set; the model construction module is used for constructing and obtaining a target multi-modal neural network model; and the monitoring prediction module is used for predicting the change trend of the equipment surface based on the target multi-mode network model and triggering corresponding early warning measures according to the prediction result. Through high-precision monitoring of the matrix sensor, the monitoring precision and the system reliability are improved. And trend prediction and abnormity early warning are carried out through the multi-modal neural network model, so that the monitoring convenience is improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of equipment monitoring and early warning, and in particular relates to an early warning platform based on a pressure sensor. Background Art

[0002] With the acceleration of industrialization, various pressure vessels, pipelines and environmental monitoring equipment play an important role in production and life. Traditional pressure monitoring methods are mostly regular inspections or the use of simple mechanical pressure gauges. These methods have problems such as untimely detection, low accuracy, and inability to remotely monitor. In recent years, although some monitoring systems based on electronic sensors have emerged, in existing equipment monitoring systems, a single pressure sensor is usually used for monitoring. These sensors have limitations in accuracy, response speed, and real-time data acquisition. For example: the accuracy of traditional pressure sensors is not enough to capture subtle changes in electrical resistance, and the accuracy is limited; there is a time delay in data acquisition and processing, and real-time monitoring cannot be achieved. The stability is poor. In harsh environments, the existing system often affects the performance of the sensor due to environmental factors, resulting in unstable monitoring data.

[0003] In addition, traditional systems often lack efficient data analysis algorithms after data collection, making it difficult to extract useful information from large amounts of data for effective real-time analysis and early warning mechanisms, making it impossible to predict potential risks. At the same time, existing systems often issue alarms only after a problem occurs, lacking preventive early warning functions. This results in an inability to respond to abnormal changes in equipment status in a timely manner. Most existing systems are unable to implement remote monitoring functions, resulting in the inability to understand the pressure status in real time when far away from the monitoring point, poor user interactivity, and an unfriendly user interface, making it difficult for non-professionals to understand and operate. Summary of the invention

[0004] In order to solve the above technical problems, the present invention provides an early warning platform based on a pressure sensor, comprising:

[0005] A signal acquisition module is used to collect the current signal on the surface of the device in real time through a pressure sensor;

[0006] A signal processing module, connected to the signal acquisition module, for converting the current signal into a digital signal through an analog-to-digital converter to obtain a data set of electrical quantity resistance information;

[0007] A data processing module, connected to the signal processing module, for first integrating the electric quantity resistance value information data of the electric quantity resistance value information data set, and then performing data processing and analysis to obtain a target data set;

[0008] A model building module, connected to the data processing module, for building an initial multimodal neural network model, and training the initial multimodal neural network model based on the target data set to obtain a target multimodal neural network model;

[0009] The monitoring prediction module is connected to the model building module and is used to predict the change trend of the equipment surface based on the target multimodal network model and trigger corresponding early warning measures according to the prediction results.

[0010] Preferably, the pressure sensor is a matrix pressure sensor sheet, which is used to monitor the current signal of 4 to 20 mA on the surface of the sensing device in real time.

[0011] Preferably, the signal processing module includes a signal preprocessing unit, a frequency sampling unit, a quantization unit, and an encoding unit;

[0012] The signal preprocessing unit is used to preprocess the current signal through an amplifier or a filter to ensure that the signal is within the input range of the analog-to-digital converter and reduce noise and interference;

[0013] The frequency sampling unit is used to sample the current signal according to the set sampling frequency, and measure the instantaneous value of the signal within a fixed time interval;

[0014] The quantization unit is used to quantize the sampled current signal value into a digital signal, set the resolution of the analog-to-digital converter according to the target requirements, and map the continuous current signal to a target number of discrete values;

[0015] The encoding unit is used to encode the quantized discrete value into a digital form in binary format to obtain a digital signal.

[0016] Preferably, the preprocessing unit includes a first preprocessing unit and a second preprocessing unit;

[0017] The first preprocessing unit is used to sequentially remove the mean, remove the linear trend, perform waveform pinch-out and perform band-pass filtering on the current signal to obtain a first current signal;

[0018] The second preprocessing unit is used to extract time-frequency domain features of the first current signal through wavelet transform.

[0019] Preferably, the first preprocessing unit includes a mean removal unit, a linear trend removal unit, a waveform pinch-out unit, and a bandpass filtering unit;

[0020] The de-averaging unit is used to remove the average value of the waveform data, eliminate the DC component of the data, and make the average amplitude of the processed data zero;

[0021] The delinear trend unit is used to eliminate the linear deflection in the data, fit the data into a straight line, and then subtract the linear trend represented by the straight line from the data;

[0022] The waveform pinch-out unit is used to gradually reduce the beginning and end of the waveform data from the original value to zero;

[0023] The bandpass filtering unit is used to retain the waveform within a specific frequency range, shield the signals in other frequency bands, obtain the effective signal frequency band and suppress noise.

[0024] Preferably, the data processing module includes a data preprocessing unit and a feature extraction unit;

[0025] The data preprocessing unit is used to perform multi-source data fusion on the electrical quantity resistance value information data through a data fusion algorithm to obtain fused data; and then perform noise filtering and data cleaning on the fused data;

[0026] The feature extraction unit is used to extract features from the data processed by the data preprocessing unit.

[0027] Preferably, the data preprocessing unit includes a data synchronization processing subunit, a data fusion subunit, a data cleaning subunit, a data standardization subunit, and a data filtering subunit;

[0028] The data synchronization processing subunit is used to perform synchronization and alignment processing on the pressure sensor data with different sampling rates and viewing angles in time and space to obtain time series data;

[0029] The data fusion subunit is used to synthesize the associated probability distribution of each sensor through a multi-Bayesian estimation method, provide a final fusion value, and perform multi-source data fusion according to the final fusion value to obtain fused data;

[0030] The data cleaning subunit is used to remove sensor noise through mean filtering, and for missing values ​​in the pressure sensor data, replace the missing values ​​with specific values ​​through the nan_to_num function by interpolation, and then identify and remove outliers through the Z-score method, or identify and correct outliers;

[0031] The data standardization subunit is used to perform data normalization by using StandardScaler of the Scikit-learn library to convert the data into a form with a mean of 0 and a standard deviation of 1; and then perform data standardization to convert the data into a form with zero mean and unit variance;

[0032] The data filtering subunit is used to perform filtering based on a low-pass filter combined with an exponential smoothing method to remove high-frequency noise.

[0033] Preferably, the data fusion subunit includes a priori probability allocation subunit, a likelihood function determination subunit, a posterior probability calculation subunit, a fused posterior probability distribution subunit, and a fusion value calculation subunit;

[0034] The prior probability allocation subunit is used to allocate a prior probability distribution to each sensor, and the prior probability distribution is used to reflect the cognition of the target variable in the absence of observation data;

[0035] The likelihood function determination subunit is used to determine the likelihood function of each sensor, and the likelihood function is used to describe the probability distribution of the target variable under the condition of given observation data;

[0036] The posterior probability calculation subunit is used to calculate the posterior probability distribution by combining the prior probability distribution and the likelihood function using the Bayesian formula; the posterior probability distribution is used to reflect the cognition of each sensor on the target variable when the observation data is available;

[0037] The fusion posterior probability distribution subunit is used to synthesize the posterior probability distributions of all sensors into a joint posterior probability distribution, and the joint posterior probability distribution is the final data fusion result;

[0038] The fusion value calculation subunit is used to calculate the final fusion value according to the fused joint posterior probability distribution.

[0039] Preferably, the model construction module includes a model construction unit, a model training unit, and a model fusion unit;

[0040] The model building unit is used to respectively build an improved temporal convolutional neural network model and an improved long short-term memory network model;

[0041] The model training unit is used to train the improved temporal convolutional neural network model based on the target data set, and then use the output of the trained temporal convolutional neural network model as the input of the improved long short-term memory network model to train the improved long short-term memory network model;

[0042] The model fusion unit is used to fuse the trained temporal convolutional neural network model and the long short-term memory network model to obtain a target multimodal neural network model.

[0043] Preferably, the improved temporal convolutional neural network model includes two identical temporal convolutional neural network models;

[0044] The temporal convolutional neural network model includes a TemporalBlock module and a TemporalConvNet module;

[0045] The TemporalBlock module is used to construct a convolutional block with residual connection. Each TemporalBlock module includes two convolutional layers with ReLU activation function and Dropout regularization, and performs residual connection through 1x1 convolution;

[0046] The TemporalConvNet module is used to combine convolution blocks with residual connections to obtain a complete temporal convolutional neural network;

[0047] Wherein, the TemporalBlock module includes an input layer, a convolutional layer, a residual connection layer, a weight normalization layer and an output layer;

[0048] The input layer is used to receive input of time series data;

[0049] The convolutional layer is used to include two one-dimensional convolutional layers, each of which is followed by a ReLU activation function and a Dropout layer for regularization;

[0050] The residual connection layer is used for each TemporalBlock module to be connected to the previous TemporalBlock module through a residual connection;

[0051] The output layer is used to output a feature map having the same shape as the input layer and the same length of the time series after convolution and nonlinear transformation;

[0052] The one-dimensional convolution layer includes a causal convolution layer and a dilated convolution layer;

[0053] The causal convolution layer is used to ensure that future time step data is not used when predicting the current time step, meeting the causal requirements in time series analysis;

[0054] The dilated convolution layer is used to increase the receptive field by increasing the interval of the convolution kernel, allowing the network to capture longer-term temporal dependencies without increasing the number of parameters;

[0055] The improved long short-term memory network model constructs an attention mechanism module and a residual module, and then introduces the attention mechanism module into the first three layers of the hidden layer of the short-term memory network to obtain an attention classification model; then the residual module is introduced into the attention classification model to generate an improved long short-term memory network model.

[0056] Compared with the prior art, the present invention has the following advantages and technical effects:

[0057] The present invention can capture more subtle changes through high-precision monitoring of matrix sensors, thereby improving monitoring accuracy and system reliability. The digital signal conversion technology ensures the real-time nature of data collection.

[0058] The present invention uses a multimodal neural network model to perform trend prediction and abnormal warning, allowing users to understand the pressure status in real time at any location, improving the convenience of monitoring, providing early warning before problems occur, reducing unexpected downtime, reducing costs, and extending equipment life. Based on the early warning information, it helps users make more scientific operation and maintenance and emergency repair decisions.

[0059] The early warning platform of the present invention can provide users with an efficient, accurate and easy-to-use equipment monitoring and early warning solution. The early warning platform is easy to operate and understand, which greatly improves the user experience and effectively prevents and reduces safety accidents caused by abnormal pressure. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] The drawings constituting a part of the present application are used to provide a further understanding of the present application. The illustrative embodiments and descriptions of the present application are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0061] Figure 1 Schematic diagram of the system structure of an embodiment of the present invention. DETAILED DESCRIPTION

[0062] It should be noted that, in the absence of conflict, the embodiments and features in the embodiments of the present application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0063] It should be noted that the steps shown in the flowcharts of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and that, although a logical order is shown in the flowcharts, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0064] like Figure 1 As shown, in this embodiment, a warning platform based on a pressure sensor is provided, including:

[0065] A signal acquisition module is used to collect the current signal on the surface of the device in real time through a pressure sensor;

[0066] The signal processing module is connected to the signal acquisition module and is used to convert the current signal into a digital signal through an analog-to-digital converter to obtain a data set of electrical quantity and resistance value information;

[0067] The data processing module is connected to the signal processing module and is used to first integrate the electric quantity resistance value information data of the electric quantity resistance value information data set, and then perform data processing and analysis to obtain a target data set;

[0068] A model building module, connected to the data processing module, is used to build an initial multimodal neural network model, and train the initial multimodal neural network model based on a target data set to obtain a target multimodal neural network model;

[0069] The monitoring prediction module is connected to the model building module and is used to predict the change trend of the equipment surface based on the target multimodal network model and trigger corresponding early warning measures according to the prediction results.

[0070] Furthermore, the pressure sensor is a matrix pressure sensor sheet, each sensor is responsible for monitoring the electrical resistance of a small area, and real-time monitoring of the current signal of 4 to 20ma on the surface of the sensing device.

[0071] Further, the signal processing module includes a signal preprocessing unit, a frequency sampling unit, a quantization unit, and an encoding unit;

[0072] The signal preprocessing unit is used to preprocess the current signal through an amplifier or a filter to ensure that the signal is within the input range of the analog-to-digital converter and reduce noise and interference;

[0073] The frequency sampling unit is used to sample the current signal according to the set sampling frequency and measure the instantaneous value of the signal within a fixed time interval; the sampling frequency of the analog-to-digital converter should be high enough to ensure that the changes in the current signal can be accurately captured to avoid aliasing effects.

[0074] The quantization unit is used to quantize the sampled current signal value into a digital signal, set the resolution of the analog-to-digital converter according to the target requirements, and map the continuous current signal to the target number of discrete values; the resolution of the analog-to-digital converter determines the accuracy of the signal conversion. A high-resolution analog-to-digital converter can provide more quantization levels, thereby more accurately reflecting the original signal.

[0075] The encoding unit is used to encode the quantized discrete value into a digital form in binary format to obtain a digital signal.

[0076] In addition, since temperature changes may affect the performance of the analog-to-digital converter, this embodiment uses temperature compensation technology to ensure the stability and accuracy of signal conversion, and regularly calibrates the analog-to-digital converter to eliminate long-term drift and ensure the accuracy of the measurement results.

[0077] The impact of electromagnetic interference on the signal conversion process is reduced through proper shielding and grounding techniques, and filtering algorithms such as moving average or Kalman filtering are applied in the digital domain to further improve the stability and reliability of the signal.

[0078] Through the above-mentioned analog-to-digital conversion process, the early warning platform of the present invention can ensure that the current signal collected from the sensor is accurately and efficiently converted into a digital signal, providing a solid foundation for subsequent data analysis and early warning.

[0079] Further, the preprocessing unit includes a first preprocessing unit and a second preprocessing unit;

[0080] The first preprocessing unit is used to sequentially remove the mean, remove the linear trend, perform waveform pinch-out and perform band-pass filtering on the current signal to obtain a first current signal;

[0081] The second preprocessing unit is used to extract time-frequency domain features of the first current signal through wavelet transform.

[0082] Further, the first preprocessing unit includes a mean removal unit, a linear trend removal unit, a waveform pinch-out unit, and a bandpass filtering unit;

[0083] The de-averaging unit is used to remove the average value of the waveform data, eliminate the DC component of the data, and make the average amplitude of the processed data zero;

[0084] The delinear trend unit is used to eliminate linear skewness in the data by fitting the data to a straight line and then subtracting the linear trend represented by the straight line from the data;

[0085] The waveform pinch-out unit is used to gradually reduce the beginning and end of the waveform data from the original value to zero;

[0086] The bandpass filter unit is used to retain the waveform within a specific frequency range, shield the signals in other frequency bands, obtain the effective signal frequency band and suppress noise.

[0087] Further, the data processing module includes a data preprocessing unit and a feature extraction unit;

[0088] The data preprocessing unit is used to perform multi-source data fusion on the electrical quantity resistance information data through a data fusion algorithm to obtain fused data; and then perform noise filtering and data cleaning on the fused data;

[0089] The feature extraction unit is used to extract features from the data processed by the data preprocessing unit.

[0090] Further, the data preprocessing unit includes a data synchronization processing subunit, a data fusion subunit, a data cleaning subunit, a data standardization subunit, and a data filtering subunit;

[0091] The data synchronization processing subunit is used to synchronize and align the pressure sensor data with different sampling rates and viewing angles in time and space to obtain time series data;

[0092] The data fusion subunit is used to synthesize the associated probability distribution of each sensor through the multi-Bayesian estimation method, provide the final fusion value, and perform multi-source data fusion according to the final fusion value to obtain fused data;

[0093] The data cleaning subunit is used to remove sensor noise through mean filtering. For missing values ​​in the pressure sensor data, the missing values ​​are replaced with specific values ​​through the nan_to_num function using interpolation, and then outliers are identified and eliminated or identified and corrected through the Z-score method.

[0094] The data standardization subunit is used to normalize the data by using the StandardScaler of the Scikit-learn library to convert the data into a form with a mean of 0 and a standard deviation of 1; then data standardization is performed to convert the data into a form with zero mean and unit variance;

[0095] The data filtering subunit is used to perform filtering based on a low-pass filter combined with an exponential smoothing method to remove high-frequency noise.

[0096] Further, the data fusion subunit includes a priori probability allocation subunit, a likelihood function determination subunit, a posterior probability calculation subunit, a fused posterior probability distribution subunit, and a fused value calculation subunit;

[0097] The prior probability assignment subunit is used to assign a prior probability distribution to each sensor, and the prior probability distribution is used to reflect the cognition of the target variable in the absence of observed data;

[0098] The likelihood function determination subunit is used to determine the likelihood function of each sensor, and the likelihood function is used to describe the probability distribution of the target variable under the condition of given observation data;

[0099] The posterior probability calculation subunit is used to calculate the posterior probability distribution by combining the prior probability distribution and the likelihood function using the Bayesian formula; the posterior probability distribution is used to reflect each sensor's knowledge of the target variable when the observation data is available;

[0100] The fusion posterior probability distribution subunit is used to synthesize the posterior probability distributions of all sensors into a joint posterior probability distribution, which is the final data fusion result;

[0101] The fusion value calculation subunit is used to calculate the final fusion value according to the fused joint posterior probability distribution.

[0102] Furthermore, the model construction module includes a model construction unit, a model training unit, and a model fusion unit;

[0103] The model building unit is used to respectively build an improved temporal convolutional neural network model and an improved long short-term memory network model;

[0104] The model training unit is used to train the improved time series convolutional neural network model based on the target data set, and then use the output of the trained time series convolutional neural network model as the input of the improved long short-term memory network model to train the improved long short-term memory network model;

[0105] The model fusion unit is used to fuse the trained temporal convolutional neural network model and the long short-term memory network model to obtain the target multimodal neural network model.

[0106] Furthermore, the improved temporal convolutional neural network model includes two identical temporal convolutional neural network models;

[0107] The temporal convolutional neural network model includes TemporalBlock module and TemporalConvNet module;

[0108] The TemporalBlock module is used to construct a convolutional block with residual connections. Each TemporalBlock module includes two convolutional layers with ReLU activation function and Dropout regularization, and performs residual connection through 1x1 convolution;

[0109] The TemporalConvNet module is used to combine convolutional blocks with residual connections to obtain a complete temporal convolutional neural network;

[0110] Among them, the TemporalBlock module includes input layer, convolution layer, residual connection layer, weight normalization layer and output layer;

[0111] The input layer is used to receive the input of time series data;

[0112] The convolutional layer is used to include two one-dimensional convolutional layers, each of which is followed by a ReLU activation function and a Dropout layer for regularization;

[0113] The residual connection layer is used to connect each TemporalBlock module to the previous TemporalBlock module through a residual connection;

[0114] The output layer is used to output the feature map with the same shape as the input layer and the same length of the time series after convolution and nonlinear transformation;

[0115] One-dimensional convolutional layers include causal convolutional layers and dilated convolutional layers;

[0116] The causal convolutional layer is used to ensure that future time step data is not used when predicting the current time step, meeting the causal requirements in time series analysis;

[0117] Dilated convolutional layers are used to increase the receptive field by increasing the spacing of the convolution kernels, allowing the network to capture longer-term temporal dependencies without increasing the number of parameters while maintaining temporal resolution;

[0118] The TemporalBlock module of this embodiment extracts local features of time series data through convolutional layers, captures short-term and long-term dependencies in time series through causal convolution and dilated convolution, and helps to effectively propagate gradients in deep networks through residual connections, so that the network can be trained deeper. Dropout and weight normalization prevent overfitting and improve the generalization ability of the model.

[0119] The improved temporal convolutional neural network model of this embodiment can effectively process time series data with long-term dependencies, allowing long sequences to be processed as a whole, and the size of the receptive field can be flexibly adjusted in a variety of ways. The gradient is stabilized to avoid the gradient explosion or disappearance problem in the RNN. The memory requirement is low when training long sequence data, so it can process inputs of variable length.

[0120] After constructing the attention mechanism module and the residual module, the improved long short-term memory network model introduces the attention mechanism module into the first three layers of the hidden layer of the short-term memory network to obtain the attention classification model; then the residual module is introduced into the attention classification model to generate the improved long short-term memory network model.

[0121] This embodiment uses a multimodal neural network model to analyze and judge the safety status of equipment operation in real time, promptly discover anomalies and potential safety hazards, and provide corresponding early warning information and decision support.

[0122] Early warning includes triggering corresponding early warning mechanisms based on risk prediction results, such as sound and light alarms, notifying relevant staff via phone or text messages for processing, automatically stopping equipment operation, etc., in order to prevent and control safety risks.

[0123] More specifically, the early warning mechanism triggering process includes:

[0124] Threshold setting: According to the safety requirements of equipment operation and historical operation status data, set the early warning threshold of the risk level. When the risk level predicted by the model exceeds this threshold, the early warning mechanism will be triggered.

[0125] Real-time monitoring: The system monitors the risk level of the model output in real time. Once it detects that the risk level exceeds the warning threshold, the warning mechanism is immediately activated. Real-time monitoring can be achieved through scheduled tasks or event triggering to ensure the timeliness and accuracy of the warning mechanism.

[0126] Multi-level warning: Design a multi-level warning mechanism to trigger different levels of warnings according to different risk levels. For example, when the risk level is medium risk, a warning prompt can be triggered to remind the operator to pay attention; when the risk level is high risk, an emergency stop operation can be triggered to ensure the safety of personnel and equipment.

[0127] Warning information transmission: Warning information needs to be delivered to relevant personnel in a timely and accurate manner. This embodiment ensures that information can be received and understood in a timely manner through various means such as sound and light alarms, SMS notifications, and system pop-up windows. Warning information should include risk levels, possible impacts, and recommended response measures to help operators make correct decisions.

[0128] Emergency response: After the early warning mechanism is triggered, corresponding emergency response measures are required, including stopping equipment operation, self-checking equipment status, etc., to ensure the safety of personnel and equipment. Emergency response measures should formulate detailed operation procedures and division of responsibilities based on risk assessment results and early warning information.

[0129] Feedback mechanism: After the early warning mechanism is triggered, feedback information is collected to evaluate the accuracy of the early warning and the effectiveness of the emergency response. By using this feedback information to optimize the model and early warning mechanism, the early warning effect in the future can be improved. The feedback mechanism can be implemented through log records, user surveys or expert reviews to ensure its comprehensiveness and objectivity. At the same time, in order to improve the effectiveness of the risk assessment and early warning module, the system needs to be continuously optimized. This includes model updates, parameter adjustments, user interactions and other aspects. The model update of this embodiment can be achieved through retraining or incremental learning to adapt to new data distribution and changes in the equipment operating environment. Parameter adjustment optimizes the accuracy and timeliness of risk assessment based on the early warning effect and actual needs. User interaction allows operators to adjust the early warning parameters according to actual conditions, or provide feedback to help the system better adapt to different operating conditions by providing a user interface.

[0130] The above are only preferred specific implementations of the present application, but the protection scope of the present application is not limited thereto. Any changes or substitutions that can be easily thought of by a person skilled in the art within the technical scope disclosed in the present application should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. An early warning platform based on pressure sensors, characterized in that: include: A signal acquisition module is used to collect the current signal on the surface of the device in real time through a pressure sensor; A signal processing module, connected to the signal acquisition module, for converting the current signal into a digital signal through an analog-to-digital converter to obtain a data set of electrical quantity resistance information; A data processing module, connected to the signal processing module, for first integrating the electric quantity resistance value information data of the electric quantity resistance value information data set, and then performing data processing and analysis to obtain a target data set; A model building module, connected to the data processing module, for building an initial multimodal neural network model, and training the initial multimodal neural network model based on the target data set to obtain a target multimodal neural network model; The monitoring prediction module is connected to the model building module and is used to predict the change trend of the equipment surface based on the target multimodal network model and trigger corresponding early warning measures according to the prediction results.

2. The early warning platform based on pressure sensor according to claim 1 is characterized in that: The pressure sensor is a matrix pressure sensor sheet, which is used to monitor the current signal of 4 to 20 mA on the surface of the sensing device in real time.

3. The early warning platform based on pressure sensor according to claim 1 is characterized in that: The signal processing module includes a signal preprocessing unit, a frequency sampling unit, a quantization unit, and an encoding unit; The signal preprocessing unit is used to preprocess the current signal through an amplifier or a filter to ensure that the signal is within the input range of the analog-to-digital converter and reduce noise and interference; The frequency sampling unit is used to sample the current signal according to the set sampling frequency, and measure the instantaneous value of the signal within a fixed time interval; The quantization unit is used to quantize the sampled current signal value into a digital signal, set the resolution of the analog-to-digital converter according to the target requirements, and map the continuous current signal to a target number of discrete values; The encoding unit is used to encode the quantized discrete value into a digital form in binary format to obtain a digital signal.

4. The early warning platform based on pressure sensor according to claim 3 is characterized in that: The preprocessing unit includes a first preprocessing unit and a second preprocessing unit; The first preprocessing unit is used to sequentially remove the mean, remove the linear trend, perform waveform pinch-out and perform band-pass filtering on the current signal to obtain a first current signal; The second preprocessing unit is used to extract time-frequency domain features of the first current signal through wavelet transform.

5. The early warning platform based on pressure sensor according to claim 4 is characterized in that: The first preprocessing unit includes a mean removal unit, a linear trend removal unit, a waveform pinch-out unit, and a bandpass filtering unit; The de-averaging unit is used to remove the average value of the waveform data, eliminate the DC component of the data, and make the average amplitude of the processed data zero; The delinear trend unit is used to eliminate the linear skewness in the data, fit the data into a straight line, and then subtract the linear trend represented by the straight line from the data; The waveform pinch-out unit is used to gradually reduce the beginning and end of the waveform data from the original value to zero; The bandpass filtering unit is used to retain the waveform within a specific frequency range, shield the signals in other frequency bands, obtain the effective signal frequency band and suppress noise.

6. The early warning platform based on pressure sensor according to claim 1, characterized in that: The data processing module includes a data preprocessing unit and a feature extraction unit; The data preprocessing unit is used to perform multi-source data fusion on the electrical quantity resistance value information data through a data fusion algorithm to obtain fused data; and then perform noise filtering and data cleaning on the fused data; The feature extraction unit is used to extract features from the data processed by the data preprocessing unit.

7. The early warning platform based on pressure sensor according to claim 6, characterized in that: The data preprocessing unit includes a data synchronization processing subunit, a data fusion subunit, a data cleaning subunit, a data standardization subunit, and a data filtering subunit; The data synchronization processing subunit is used to perform synchronization and alignment processing on the pressure sensor data with different sampling rates and viewing angles in time and space to obtain time series data; The data fusion subunit is used to synthesize the associated probability distribution of each sensor through a multi-Bayesian estimation method, provide a final fusion value, and perform multi-source data fusion according to the final fusion value to obtain fused data; The data cleaning subunit is used to remove sensor noise through mean filtering, and for missing values ​​in the pressure sensor data, replace the missing values ​​with specific values ​​through the nan_to_num function by interpolation, and then identify and remove outliers through the Z-score method, or identify and correct outliers; The data standardization subunit is used to perform data normalization by using StandardScaler of the Scikit-learn library to convert the data into a form with a mean of 0 and a standard deviation of 1; and then perform data standardization to convert the data into a form with zero mean and unit variance; The data filtering subunit is used to perform filtering based on a low-pass filter combined with an exponential smoothing method to remove high-frequency noise.

8. The early warning platform based on pressure sensor according to claim 7, characterized in that: The data fusion subunit includes a priori probability allocation subunit, a likelihood function determination subunit, a posterior probability calculation subunit, a fused posterior probability distribution subunit, and a fusion value calculation subunit; The prior probability allocation subunit is used to allocate a prior probability distribution to each sensor, and the prior probability distribution is used to reflect the cognition of the target variable in the absence of observation data; The likelihood function determination subunit is used to determine the likelihood function of each sensor, and the likelihood function is used to describe the probability distribution of the target variable under the condition of given observation data; The posterior probability calculation subunit is used to calculate the posterior probability distribution by combining the prior probability distribution and the likelihood function using the Bayesian formula; the posterior probability distribution is used to reflect the cognition of each sensor on the target variable when the observation data is available; The fusion posterior probability distribution subunit is used to synthesize the posterior probability distributions of all sensors into a joint posterior probability distribution, and the joint posterior probability distribution is the final data fusion result; The fusion value calculation subunit is used to calculate the final fusion value according to the fused joint posterior probability distribution.

9. The early warning platform based on pressure sensor according to claim 1, characterized in that: The model construction module includes a model construction unit, a model training unit, and a model fusion unit; The model building unit is used to respectively build an improved temporal convolutional neural network model and an improved long short-term memory network model; The model training unit is used to train the improved temporal convolutional neural network model based on the target data set, and then use the output of the trained temporal convolutional neural network model as the input of the improved long short-term memory network model to train the improved long short-term memory network model; The model fusion unit is used to fuse the trained temporal convolutional neural network model and the long short-term memory network model to obtain a target multimodal neural network model.

10. The early warning platform based on pressure sensor according to claim 9, characterized in that: The improved temporal convolutional neural network model includes two identical temporal convolutional neural network models; The temporal convolutional neural network model includes a TemporalBlock module and a TemporalConvNet module; The TemporalBlock module is used to construct a convolutional block with residual connection. Each TemporalBlock module includes two convolutional layers with ReLU activation function and Dropout regularization, and performs residual connection through 1x1 convolution; The TemporalConvNet module is used to combine convolution blocks with residual connections to obtain a complete temporal convolutional neural network; Wherein, the TemporalBlock module includes an input layer, a convolutional layer, a residual connection layer, a weight normalization layer and an output layer; The input layer is used to receive input of time series data; The convolutional layer is used to include two one-dimensional convolutional layers, each of which is followed by a ReLU activation function and a Dropout layer for regularization; The residual connection layer is used for each TemporalBlock module to be connected to the previous TemporalBlock module through a residual connection; The output layer is used to output a feature map having the same shape as the input layer and the same length of the time series after convolution and nonlinear transformation; The one-dimensional convolution layer includes a causal convolution layer and a dilated convolution layer; The causal convolution layer is used to ensure that future time step data is not used when predicting the current time step, meeting the causal requirements in time series analysis; The dilated convolution layer is used to increase the receptive field by increasing the interval of the convolution kernel, allowing the network to capture longer-term temporal dependencies without increasing the number of parameters; The improved long short-term memory network model constructs an attention mechanism module and a residual module, and then introduces the attention mechanism module into the first three layers of the hidden layer of the short-term memory network to obtain an attention classification model; then the residual module is introduced into the attention classification model to generate an improved long short-term memory network model.

Images