A pressure transmitter output linearization correction method based on range ratio self-adaption

Abstract

The application provides a pressure transmitter output linearization correction method based on a range ratio self-adaption, and relates to the technical field of signal processing.The method comprises the following steps: guiding a linearization correction model to perform self-adaptive correction on a current pressure signal according to a current range ratio; obtaining a plurality of monitoring reinforcement vectors, performing multi-mode failure probability prediction on the pressure transmitter, performing causal modeling and tracing correction on a first corrected pressure signal according to a significant failure probability, and obtaining a second corrected pressure signal; and performing over-correction backtracking compensation on the second corrected pressure signal, and obtaining a third corrected pressure signal.The technical problems of insufficient correction accuracy of the pressure transmitter under different range ratios, output distortion caused by multi-working condition interference, failure factors that cannot be quantitatively corrected in real time, and over-correction deviation are solved.The technical effect of adaptively improving the sensitivity of the pressure transmitter under variable range conditions and realizing high-precision and stable output of the pressure transmitter is achieved.

Description

Technical Field

[0001] This application relates to the field of signal processing technology, and in particular to a method for linearizing the output of a pressure transmitter based on adaptive range ratio. Background Technology

[0002] In the field of industrial automation process control, pressure transmitters are core measurement and control equipment, and their output accuracy directly affects production stability. Due to factors such as the process tolerances of sensitive elements and drift in signal conditioning circuits, pressure transmitter outputs inherently exhibit nonlinear errors compared to actual pressures, with the deviation becoming more significant in low-range segments or after range switching. Existing calibration methods have significant drawbacks: full-range single linear fitting and fixed-segment fitting cannot adapt to dynamic changes in range ratios; hardware compensation methods are costly and difficult to maintain, and both lack adaptive range ratio capabilities. Manual recalibration is required after range adjustments, resulting in low efficiency, unstable accuracy, and failure to consider the impact of environmental interference and parameter drift.

[0003] Therefore, current technologies suffer from several technical problems, including insufficient calibration accuracy of pressure transmitters at different range ratios, easy output distortion due to interference from multiple operating conditions, inability to quantify and correct failure factors in real time, and the tendency for overcalibration deviations. Summary of the Invention

[0004] This application provides a linearization correction method for pressure transmitter output based on adaptive range ratio. This method addresses the technical problems in existing technologies, such as insufficient correction accuracy of pressure transmitters under different range ratios, output distortion due to interference from multiple operating conditions, inability to quantify and correct failure factors in real time, and susceptibility to overcorrection. It achieves the technical effects of adaptively improving the sensitivity of the pressure transmitter under variable range conditions, real-time identification and suppression of multi-mode failure errors, and high-precision stable output of the pressure transmitter through causal tracing and compensation of signals at the same node.

[0005] This application provides a method for linearizing the output of a pressure transmitter based on adaptive range ratio. The method includes: performing full-range multi-point calibration on the factory test dataset of the pressure transmitter to establish a linearization correction model; acquiring the current pressure signal, current range ratio, and multimodal monitoring sequence of the pressure transmitter, and guiding the linearization correction model to adaptively correct the current pressure signal based on the current range ratio to obtain a first corrected pressure signal; performing multi-mode failure contribution attention enhancement on the multimodal monitoring sequence to obtain multiple monitoring enhancement vectors, and predicting the multimodal failure probability of the pressure transmitter based on the multiple monitoring enhancement vectors to determine a significant failure probability; performing causal modeling and retrospective correction on the first corrected pressure signal based on the significant failure probability to obtain a second corrected pressure signal; performing overcorrection and backtracking compensation on the second corrected pressure signal based on the pressure signal set of the same node of the pressure transmitter to obtain a third corrected pressure signal, and converting and displaying the third corrected pressure signal.

[0006] This application proposes a linearization correction method for pressure transmitter output based on adaptive range ratio. The method establishes a linearization correction model by performing full-range multi-point calibration on the factory test dataset of the pressure transmitter. The linearization correction model is then used to adaptively correct the current pressure signal based on the current range ratio, obtaining a first corrected pressure signal. Multiple monitoring enhancement vectors are obtained by performing multi-mode failure contribution attention enhancement on the multi-modal monitoring sequence, and multi-mode failure probability prediction is performed on the pressure transmitter based on these vectors to determine the significant failure probability. The first corrected pressure signal is then corrected using causal modeling and tracing based on the significant failure probability, resulting in a second corrected pressure signal. Finally, the second corrected pressure signal is overcorrected and back-compensated based on the pressure signal set of the same node of the pressure transmitter, resulting in a third corrected pressure signal, which is then converted and displayed. This method solves the technical problems in existing technologies, such as insufficient correction accuracy of pressure transmitters at different range ratios, output distortion due to interference under multiple operating conditions, inability to quantify and correct failure factors in real time, and susceptibility to overcorrection deviations. It achieves the technical effect of adaptively improving the sensitivity of the pressure transmitter under variable process conditions, identifying and suppressing multi-mode failure errors in real time, and realizing high-precision and stable output of the pressure transmitter through causal tracing and same-node signal compensation. Attached Figure Description

[0007] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings of the embodiments of the present invention will be briefly described below. Flowcharts are used in this application to illustrate the operations performed by the system according to the embodiments of the present application. It should be understood that the preceding or following operations are not necessarily performed precisely in sequence. Instead, various steps can be processed in reverse order or simultaneously as needed. Furthermore, other operations can be added to these processes, or one or more steps can be removed from these processes.

[0008] Figure 1 This application provides a flowchart illustrating a method for linearizing the output of a pressure transmitter based on adaptive range ratio.

[0009] Figure 2 This is a flowchart illustrating the process of establishing a linearization correction model in a pressure transmitter output linearization correction method based on range ratio adaptation provided in this application. Detailed Implementation

[0010] The above description is only an overview of the technical solution of this application. In order to better understand the technical means of this application and to implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of this application more obvious and understandable, the following are specific embodiments of this application.

[0011] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description of this application will be provided in conjunction with the accompanying drawings. The described embodiments should not be considered as limitations on this application. All other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0012] In the following description, references to "some embodiments" describe a subset of all possible embodiments. However, it is understood that "some embodiments" can be the same or different subsets of all possible embodiments and can be combined with each other without conflict. The terms "first" and "second" are used merely to distinguish similar objects and do not represent a specific ordering of objects. The terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or server that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or modules not explicitly listed or inherent to such processes, methods, products, or devices. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains. The terminology used herein is for the purpose of describing this application only.

[0013] In this embodiment, the present application provides a method for linearizing the output of a pressure transmitter based on adaptive range ratio, such as... Figure 1 As shown, the method includes: Step A100: Perform full-range multi-point calibration on the factory test dataset of the pressure transmitter to establish a linearized correction model. Among other things, such as... Figure 2 As shown, step A100 further includes steps A110 to A140. Step A110: Clean and extract the factory test dataset to obtain multiple pressure excitation test sequences corresponding to multiple range ratios. The factory test dataset is the set of raw test data collected by a standard pressure calibration device during the factory testing phase of the pressure transmitter, based on its full range of 0-40MPa and a 4:1 range ratio (M1 / M2 / M4, L1 / L2 range ratio). The factory test dataset includes the transmitter's original output signal, test environment temperature, static pressure, and other relevant data corresponding to different range ratios and different pressure excitation values. Full-range multi-point calibration refers to the process of selecting multiple pressure excitation points for testing within the full range of 0-40MPa and multiple range ratios of the pressure transmitter, and constructing a linearized correction model adapted to different range ratios through data processing and model fitting. The specific implementation process of full-range multi-point calibration includes steps A110 to A140. First, abnormal data in the factory test dataset is initially screened to remove missing values, abrupt changes, and data that significantly deviates from the standard range due to test equipment malfunctions or operational errors. Then, the cleaned factory test dataset is categorized and extracted based on multiple range ratios, with each range ratio corresponding to a pressure excitation test sequence. Each pressure excitation test sequence includes multiple pressure excitation points from the lower to the upper limit of the range (at least 10 evenly distributed pressure excitation points are selected for each range ratio, covering different intervals from the small range of 40 kPa to the full range of 40 MPa), as well as the transmitter's original output signal and test environment parameters (temperature, static pressure) corresponding to each pressure excitation point. The pressure excitation point represents the pressure excitation value, which is the standard pressure value (the actual pressure value).

[0014] Step A120: Perform anomaly identification and correction on the multiple pressure excitation test sequences to obtain multiple reliable pressure test sequences. Specifically, an anomaly identification algorithm based on the 3σ criterion is used to detect data points in each pressure excitation test sequence point by point, identifying abnormal data points that exceed the mean ± 3σ range. The specific implementation process of the anomaly identification algorithm based on the 3σ criterion is as follows: First, for the same set of data in each pressure excitation test sequence (the original output signals of the transmitters corresponding to all pressure excitation points under the same range ratio), calculate the arithmetic mean μ and standard deviation σ of the set of data. Then, construct a normal data range centered on the arithmetic mean μ, which includes the interval from μ-3σ to μ+3σ. Data within the normal data range is determined to be normal data, while data outside the normal data range (i.e., less than μ-3σ or greater than μ+3σ) is identified as abnormal data points. By using the anomaly identification algorithm based on the 3σ criterion to identify anomalies in multiple pressure excitation test sequences, abnormal data points in multiple pressure excitation test sequences can be effectively identified, improving the accuracy of the linearization correction model. Furthermore, for identified outlier data points, instead of directly discarding them, a linear interpolation method is used for correction. Specifically, based on the two adjacent normal data points before and after the outlier data point, a corrected value for that outlier point is calculated through linear fitting and used to replace the original outlier data. After correction, a consistency check is performed on each pressure excitation test sequence to ensure that the output signal deviation at the same pressure excitation point in multiple repeated tests does not exceed ±0.075%. Once the check passes, a reliable pressure test sequence corresponding to each range ratio is obtained, ensuring the accuracy of subsequent model fitting.

[0015] Step A130: Perform least squares fitting on the multiple reliable pressure test sequences to obtain multiple signal correction models. By performing least squares fitting on multiple reliable pressure test sequences, a precise linear correspondence between the "pressure excitation value" and the "transmitter's original output signal" is established, thereby correcting the nonlinear error inherent in the pressure transmitter and ensuring the accuracy of subsequent corrections. Specifically, an independent least squares fitting operation is performed on each reliable pressure test sequence to avoid cross-contamination of data with different range ratios. During fitting, the independent and dependent variables are clearly defined: the pressure excitation point in each reliable pressure test sequence is taken as the independent variable x, and the original output signal of the transmitter corresponding to that pressure excitation point is taken as the dependent variable y. Subsequently, based on the core principle of least squares, the coefficients of the linear regression equation y=kx+b are solved by calculating the minimum sum of squares of the deviations of all data points from the fitted line. k is the slope of the regression line, and b is the intercept of the regression line. The specific calculation process is as follows: First, calculate the average value of all x values ​​and the average value of all y values ​​in each reliable pressure test sequence, and then calculate the slope k and intercept b to ensure that the fitted line can fit all test data points to the greatest extent.

[0016] In one possible implementation, the slope k is calculated using the following formula: ; Where k is the slope, and n is the total number of pressure excitation points in each reliable pressure test sequence. Let i be the i-th pressure excitation point in the reliable pressure test sequence, where i is a positive integer. This is the original output signal of the transmitter corresponding to the i-th pressure excitation point.

[0017] In one possible implementation, the intercept b is calculated using the following formula: ; Where b is the intercept. The y-value is the average of all y-values ​​in the reliable stress test sequence, and k is the slope. This is the average of all x values ​​in the reliable stress test sequence.

[0018] Each signal correction model is a linear regression equation (y=kx+b) obtained by fitting each reliable pressure test sequence using the least squares method. The functional expression of the signal correction model is y=kx+b. The core function of multiple signal correction models is to establish a precise correspondence between the "transmitter's original output signal (y)" and the "true standard pressure value (x)," thereby correcting the nonlinear deviation between the original output signal and the true pressure value caused by inherent biases and other factors of the pressure transmitter itself, into a linear correspondence. Since the signal characteristics of the reliable pressure test sequences corresponding to each range ratio are different (the signal response law of the pressure transmitter differs under different range ratios), an independent signal correction model is obtained after fitting each reliable pressure test sequence. Each signal correction model is only adapted to its corresponding range ratio, laying the foundation for subsequent adaptive calling of the signal correction model based on the range ratio and achieving accurate correction under different range ratios.

[0019] Step A140: Using the multiple range ratios as multiple calibration drive pointers, and driving and encapsulating the multiple signal calibration models according to the multiple calibration drive pointers, the linearized calibration model is generated. Each range ratio is treated as an independent calibration drive pointer, and each calibration drive pointer corresponds to a unique signal calibration model, establishing a one-to-one correspondence between calibration drive pointers and signal calibration models. Subsequently, a modular encapsulation method is used to integrate multiple signal calibration models and their corresponding multiple calibration drive pointers into a unified linearized calibration model. During the encapsulation process, a range ratio recognition module is embedded to ensure that after obtaining the current range ratio of the pressure transmitter, the matching signal calibration model can be quickly called through the corresponding calibration drive pointer. At the same time, the adjustable interface of the model parameters is retained during encapsulation, compatible with the three-button local operation of the pressure transmitter (handheld device / configuration software / mobile APP remote operation mode), which facilitates subsequent fine-tuning of the model parameters according to actual working conditions, ultimately generating a linearized calibration model that can adaptively switch according to the range ratio. The calibration drive pointer is a unique identifier for each range ratio, used to establish a one-to-one correspondence between the range ratio and the signal calibration model. The core function of the calibration drive pointer is as an "index." When the current range ratio of the pressure transmitter is subsequently obtained, the corresponding calibration drive pointer can be used to quickly call the signal calibration model adapted to the current range ratio, achieving adaptive switching of the calibration model. The linearized calibration model is a unified calibration model that integrates multiple calibration drive pointers and corresponding signal calibration models through modular encapsulation. The linearized calibration model has a built-in range ratio identification module and an adjustable model parameter interface. It can adaptively call the matching signal calibration model according to the current range ratio, correct the transmitter's nonlinear error, and achieve accurate linear calibration of the pressure signal under different range ratios. It is also compatible with the transmitter's local and remote operation modes, providing basic support for subsequent pressure signal calibration.

[0020] Step A200: Acquire the current pressure signal, current range ratio, and multimodal monitoring sequence of the pressure transmitter, and guide the linearization correction model to adaptively correct the current pressure signal based on the current range ratio to obtain the first corrected pressure signal. It should be noted that the current pressure signal is the raw pressure signal acquired by the pressure transmitter in real time, without any correction processing. The current range ratio is the range ratio corresponding to the current pressure signal. The multimodal monitoring sequence includes real-time temperature data, sensor power supply voltage data, noise data, transmitter housing vibration data, ambient temperature data, ambient humidity data, electromagnetic interference data, and other data from the pressure transmitter. The multimodal monitoring sequence is acquired synchronously with the current pressure signal by the auxiliary monitoring elements integrated into the pressure transmitter (such as temperature sensors, voltage monitoring modules, humidity sensors, electromagnetic interference monitoring modules, etc.). The multimodal monitoring sequence covers the pressure transmitter's own state data and the field environmental state information, providing comprehensive data support for subsequent failure probability prediction of the pressure transmitter.

[0021] Furthermore, the current range ratio is input into the linearization correction model. The range ratio identification module within the linearization correction model quickly identifies the current range ratio and, through the corresponding correction drive pointer, calls the signal correction model uniquely corresponding to the current range ratio. Then, the current pressure signal is input into the called signal correction model, and the corresponding independent variable x (i.e., the true standard pressure value corresponding to the current pressure signal) is solved in reverse. Through the signal correction model, the nonlinear error caused by the inherent deviation of the pressure transmitter in the current pressure signal is corrected, achieving a precise conversion from "original output signal → true pressure value." The entire correction process is automatically completed by the linearization correction model based on the current range ratio, effectively improving the intelligence and reliability of the pressure transmitter's output correction. That is, under different current range ratios, the linearization correction model automatically calls the corresponding suitable signal correction model, achieving adaptive correction under different range ratios and avoiding the problem of decreased correction accuracy caused by a fixed model. The first corrected pressure signal is the signal after the current pressure signal has been initially corrected by the linearization correction model.

[0022] Step A300: Perform multi-mode failure contribution attention enhancement on the multi-modal monitoring sequence to obtain multiple monitoring enhancement vectors, and predict the multi-mode failure probability of the pressure transmitter based on the multiple monitoring enhancement vectors to determine the significant failure probability. Step A300 further includes steps A310 to A340. Step A310: Evaluate the parameter drift failure contribution of the multi-modal monitoring sequence to obtain the drift failure contribution degree of each parameter. Parameter drift failure refers to a common failure mode in which various monitoring parameters in the multi-modal monitoring sequence deviate from the factory calibration normal range during long-term operation, leading to a decrease in the measurement accuracy of the pressure transmitter and distortion of the output signal. The parameter drift failure contribution evaluation is a quantitative calculation to assess the degree of influence of each type of monitoring parameter on parameter drift failure, ultimately obtaining the drift failure contribution degree of each parameter (values ​​range from 0 to 1, with values ​​closer to 1 indicating a greater impact). Specifically, set the factory calibration normal parameter range for each type of monitoring parameter in the multi-modal monitoring sequence, and then calculate the deviation rate between the real-time value of each type of monitoring parameter and the corresponding normal range center value. The deviation rate is calculated using the formula: Deviation Rate = |Real-time Parameter Value - Center Value of Normal Range| / (Upper Limit of Normal Range - Lower Limit of Normal Range). Then, a weight is assigned to each type of monitoring parameter using the Analytic Hierarchy Process (AHP), with a total weight of 1. That is, the weights are preset according to the degree of influence of each type of monitoring parameter on the pressure transmitter failure, such as a weight of 0.25 for sensor power supply voltage, 0.2 for transmitter operating temperature, and 0.15 for field electromagnetic interference intensity, with other parameters assigned corresponding weights. Finally, the deviation rate of each type of monitoring parameter is multiplied by its corresponding weight to obtain the contribution of each parameter to drift failure.

[0023] Step A320: Based on the drift failure contribution of each parameter, the multimodal monitoring sequence is screened and attention is allocated to obtain a first monitoring enhancement vector. It is determined whether the drift failure contribution of each parameter is less than a contribution threshold. Monitoring parameters in the multimodal monitoring sequence with drift failure contributions less than the contribution threshold are removed to obtain a candidate monitoring parameter set. The candidate monitoring parameter set includes multiple candidate monitoring parameters. Candidate monitoring parameters are monitoring parameters in the multimodal monitoring sequence with drift failure contributions greater than or equal to the contribution threshold. The contribution threshold can be adaptively set. Attention allocation is based on the drift failure contribution of candidate monitoring parameters, assigning different attention weights to them so that subsequent drift failure prediction focuses on parameters with high contributions. That is, the drift failure contribution of candidate monitoring parameters is normalized, and the sum of the attention weights of each parameter after normalization is 1. The normalization formula is: Attention weight of a parameter = Drift failure contribution of that parameter / Sum of drift failure contributions of multiple candidate monitoring parameters. Multiple candidate monitoring parameters are weighted and fused with their corresponding parameter attention weights, and arranged in a preset parameter order (e.g., device parameters first, then environmental parameters) to obtain the first monitoring enhancement vector.

[0024] Step A330: Perform short-circuit failure contribution attention enhancement on the multimodal monitoring sequence to obtain a second monitoring enhancement vector. Short-circuit failure refers to a short-circuit fault in the internal circuitry of the pressure transmitter (such as the sensor element interface, signal processing unit circuit, etc.), leading to abnormal current, signal transmission interruption or distortion, and ultimately causing pressure transmitter failure. Short-circuit failure contribution attention enhancement targets relevant monitoring parameters that may lead to short-circuit failure, strengthening their feature weights so that these parameters receive priority attention in subsequent short-circuit failure monitoring. Essentially, it extracts and enhances features related to short-circuit failure from the multimodal monitoring sequence. Specifically, based on the working principle of the pressure transmitter, several key monitoring parameters directly related to short-circuit failure are selected, such as sensor power supply voltage data, transmitter housing vibration data, and field electromagnetic interference intensity data. Then, a short-circuit failure contribution weight is assigned to each key monitoring parameter, with a total weight of 1. The weight can be preset according to the degree of influence of each key monitoring parameter on short-circuit failure, such as a sensor power supply voltage weight of 0.4, a field electromagnetic interference intensity weight of 0.35, and a transmitter housing vibration data weight of 0.25. Next, the key monitoring parameters and their corresponding short-circuit failure contribution weights are weighted and fused together, and arranged in a fixed order of "sensor power supply voltage → field electromagnetic interference intensity → transmitter housing vibration" to form a three-dimensional vector, which is the second monitoring enhancement vector.

[0025] Step A340: Perform open-circuit failure contribution attention enhancement on the multimodal monitoring sequence to obtain a third monitoring enhancement vector. Combine the first and second monitoring enhancement vectors to generate the multiple monitoring enhancement vectors. Open-circuit failure refers to a critical failure type where an open circuit occurs in the internal circuitry of the pressure transmitter (such as sensor element connection lines or signal transmission lines), causing the signal to fail to transmit normally and the transmitter to fail to acquire and output pressure signals. Open-circuit failure contribution attention enhancement targets relevant monitoring parameters that may lead to open-circuit failure, strengthens their feature weights, extracts and enhances features related to open-circuit failure in the multimodal monitoring sequence, and finally obtains the third monitoring enhancement vector. Specifically, based on the working principle of the pressure transmitter, several core monitoring parameters directly related to open-circuit failure are selected, such as sensor power supply voltage data, noise data, and transmitter temperature data. Then, an open-circuit failure contribution weight is assigned to each core monitoring parameter, with a total weight of 1. The impact of each core monitoring parameter on open-circuit failure can be preset, such as a weight of 0.45 for sensor power supply voltage, 0.3 for noise data, and 0.25 for transmitter temperature data. Then, each core monitoring parameter is weighted and fused with its corresponding open-circuit failure contribution weight, arranged in a fixed order of "sensor power supply voltage → noise → transmitter temperature" to form a three-dimensional vector, which is the third monitoring enhancement vector. Multiple monitoring enhancement vectors include the first, second, and third monitoring enhancement vectors. These vectors focus on the characteristics of parameter drift, short circuit, and open circuit failures, respectively, achieving comprehensive enhancement of various failure-related features in the multi-modal monitoring sequence. This ensures that subsequent multi-mode failure probability prediction can comprehensively and accurately identify various potential failure probabilities.

[0026] Step A300 further includes steps A350 to A390. Step A350: Based on the first monitoring reinforcement vector, predict the parameter drift failure probability of the pressure transmitter to determine a first failure probability. Step A350 further includes steps A351 to A355. Step A351: Iteratively train a Bayesian network based on the drift failure event set of the pressure transmitter to obtain a first drift failure prediction model.

[0027] The drift failure event set refers to the complete set of all parameter drift failure-related events that have occurred during the entire life cycle of the pressure transmitter. Each event contains complete information such as historical monitoring parameter data at the time of the parameter drift failure, the degree of historical parameter drift, the duration of historical failure, the scope of historical failure impact, and whether it ultimately led to transmitter failure.

[0028] Bayesian networks are inference tools based on probabilistic graphical models, consisting of nodes and edges. Nodes represent variables related to drift failures (such as monitoring parameters and whether a drift failure occurs), while edges represent causal relationships between variables. The core of a Bayesian network is to predict the probability of drift failures by calculating prior and posterior probabilities. The specific training process is as follows: First, the drift failure event set is preprocessed, removing invalid and missing event data. The remaining valid event data is divided into a training set and a validation set in a 7:3 ratio. The training set is used for model training, and the validation set is used to verify the model's accuracy. Then, the topology of the Bayesian network is constructed, using the monitoring parameters contained in the first monitoring reinforcement vector as input nodes and the drift failure occurrence probability as the output node. Based on the causal relationships in the drift failure event set, the nodes are connected, and an initial conditional probability table is set between the nodes (the initial conditional probabilities are obtained based on the event frequency statistics in the drift failure event set). Next, the training set is input into the Bayesian network, the network's conditional probability table is updated, and the causal relationship weights between nodes are adjusted, completing one iteration of training. When the number of iterations reaches a preset upper limit (e.g., 100 times), the iteration stops. The resulting Bayesian network is the first model for drift failure prediction. The first model for drift failure prediction is a Bayesian network structure based on a probabilistic graph. It consists of "nodes" and "directed edges". Nodes represent variables related to drift failure, and directed edges represent causal relationships between variables.

[0029] Step A352: Iteratively train the temporal neural network based on the drift failure event set to obtain the second drift failure prediction model. A temporal neural network is a neural network model capable of processing time-series data and capturing the temporal dependencies of data. Preferably, this application uses a Long Short-Term Memory (LSTM) network. The specific training process is as follows: First, the drift failure event set is processed into a temporal sequence. The historical monitoring parameter data corresponding to each failure event is organized into time-series data segments, with each segment's length preset to 60 seconds (adjustable). Each time-series data segment is labeled (1 indicates a drift failure occurred, 0 indicates no drift failure occurred). Then, the processed time-series data is divided into a training set and a validation set in a 7:3 ratio. Next, an LSTM temporal neural network structure is constructed, where the input layer dimension is consistent with the dimension of the first monitoring reinforcement vector, 2-3 hidden layers are set, each layer has a preset number of neurons of 64, and the output layer is a single output node used to output the probability of drift failure. The activation function is the Sigmoid function. The time-series data from the training set is then input into an LSTM temporal neural network structure. The Adam optimizer, using the cross-entropy loss function, is employed to adjust the network weights and biases, completing one iteration of training. When the number of iterations reaches a preset upper limit, the iteration stops, yielding the second drift failure prediction model. The input layer dimension of the second drift failure prediction model is consistent with the dimension of the first monitoring reinforcement vector. It has 2-3 hidden layers, with a preset number of neurons per layer of 64, and a single output node.

[0030] Step A353: Based on the first drift failure prediction model and the second drift failure prediction model, predict the failure characteristics of the first monitoring reinforcement vector to obtain a first probability of drift failure and a second probability of drift failure. Specifically, input the first monitoring reinforcement vector into the first drift failure prediction model to obtain the first probability of drift failure. The first probability of drift failure ranges from 0 to 1; the closer the value is to 1, the greater the probability of drift failure occurring. Simultaneously, input the first monitoring reinforcement vector into the second drift failure prediction model. The second drift failure prediction model calculates and outputs the second probability of drift failure by capturing the time-varying characteristics of the first monitoring reinforcement vector. The second probability of drift failure also ranges from 0 to 1, consistent with the meaning of the first probability of drift failure, but the prediction basis is different (the former is based on probabilistic inference, the latter on temporal characteristics).

[0031] Step A354: Construct a drift failure fusion node based on the first accuracy of drift failure prediction and the second accuracy of drift failure prediction.

[0032] The first precision of drift failure prediction refers to the prediction accuracy of the first drift failure prediction model on the validation set, that is, the degree of agreement between the model's prediction results and the actual labels on the validation set. It is calculated by dividing the number of correctly predicted events on the validation set by the total number of events in the validation set, and its value ranges from 0 to 1. The higher the first precision of drift failure prediction, the more reliable the drift failure prediction model's predictions. The second precision of drift failure prediction refers to the prediction accuracy of the second drift failure prediction model on the same validation set. The calculation method for the second precision of drift failure prediction is the same as that for the first precision.

[0033] The drift failure fusion node is a functional node used to fuse the predicted probabilities of two models. Essentially, it assigns different fusion weights based on the prediction accuracy of the two models, achieving a weighted fusion of the two probability values. Its core purpose is to combine the advantages of both models (the probabilistic inference advantage of Bayesian networks and the temporal capture advantage of temporal neural networks) to improve the prediction accuracy of drift failures. The specific construction process of the drift failure fusion node is as follows: Let R1 be the first accuracy of drift failure prediction and R2 be the second accuracy. Then, normalize R1 and R2 to obtain the first model fusion weight W1 and the second model fusion weight W2. Specifically, W1 = R1 / (R1+R2), W2 = R2 / (R1+R2). Finally, the first model fusion weight W1, the second model fusion weight W2, and the corresponding probability fusion logic formula are encapsulated into an independent functional node, which is the drift failure fusion node.

[0034] Step A355: Input the first probability of drift failure and the second probability of drift failure into the drift failure fusion node, and output the first failure probability. Specifically, according to the probability fusion logic formula within the drift failure fusion node, the first probability of drift failure and the second probability of drift failure are weighted and fused to obtain the first failure probability. The drift failure fusion node not only compensates for the limitations of single-model prediction but also improves the prediction accuracy of parameter drift failure probability. The output first failure probability can serve as the core basis for subsequent first correction pressure signal failure correction, effectively reducing the impact of parameter drift failure on the measurement accuracy of the pressure transmitter and ensuring the stable and accurate operation of the pressure transmitter in the industrial field.

[0035] It should be noted that the probability fusion logic formula is: First failure probability = W1 × P_drift1 + W2 × P_drift2. Where P_drift1 is the first probability of drift failure, and P_drift2 is the second probability of drift failure.

[0036] Step A360: Predict the short-circuit failure probability of the pressure transmitter based on the second monitoring reinforcement vector to determine the second failure probability. The second failure probability is a quantified probability value of the pressure transmitter experiencing a short-circuit failure, ranging from 0 to 1. The closer the value is to 1, the higher the probability of the pressure transmitter experiencing a short-circuit failure. Specifically, firstly, a short-circuit failure prediction model is constructed, using a Bayesian network structure consistent with the first drift failure prediction model. The input nodes of the short-circuit failure prediction model correspond to multiple parameters within the second monitoring reinforcement vector, and the output node is the "short-circuit failure probability." The directed edges between nodes are constructed based on the causal relationship of short-circuit failure. The initial conditional probability is obtained based on the event frequency statistics of the historical event set of pressure transmitter short-circuit failures (including various parameter data and failure conditions at the time of short-circuit failure). Subsequently, the historical event set is divided into a training set and a validation set in a 7:3 ratio, and the maximum likelihood estimation method is used to iteratively train the model until the predetermined number of iterations is reached, completing the construction of the short-circuit failure prediction model. Then, the second monitoring reinforcement vector is input into the short-circuit failure prediction model to obtain the second failure probability.

[0037] Step A370: Based on the third monitoring reinforcement vector, predict the open-circuit failure probability of the pressure transmitter to determine the third failure probability. The third failure probability is a quantified probability value of the pressure transmitter experiencing an open-circuit failure, ranging from 0 to 1. The closer the value is to 1, the higher the probability of the pressure transmitter experiencing an open-circuit failure. Specifically, an open-circuit failure prediction model is constructed, which uses a Bayesian network with the same structure as the first drift failure prediction model. The input nodes of the open-circuit failure prediction model correspond to the parameters of the third monitoring reinforcement vector, and the output node is the "open-circuit failure occurrence probability". The directed edges between nodes are constructed based on the causal relationship of open-circuit failure (such as excessive temperature leading to line breakage, abnormal output noise reflecting poor line contact, etc.). The initial conditional probability is obtained based on the event frequency statistics of the historical event set of open-circuit failure of the pressure transmitter (including complete information such as parameter data when the open-circuit failure occurred, failure duration, and fault cause). Then, the historical event set is divided into a training set and a validation set in a 7:3 ratio, and the maximum likelihood estimation method is used to iteratively train the model to complete the construction of the open-circuit failure prediction model. The third monitoring enhancement vector is then input into the open-circuit failure prediction model to obtain the third failure probability.

[0038] Step A380: Perform significance detection on the first failure probability, the second failure probability, and the third failure probability based on the predetermined failure probability to obtain multiple failure significance levels. These multiple failure significance levels include a first failure significance level, a second failure significance level, and a third failure significance level. The core logic of the significance detection is to calculate the difference between each failure probability and the predetermined failure probability. Specifically, the first failure significance level = first failure probability - predetermined failure probability. The calculation methods for the second and third failure significance levels are the same as for the first failure significance level and will not be repeated here. The larger the failure significance level, the greater the risk of the corresponding failure mode. By performing significance detection on the first, second, and third failure probabilities, the targeting and effectiveness of failure correction can be improved, thereby enhancing the accuracy of pressure transmitter output correction.

[0039] Step A390: Select the first failure probability, the second failure probability, and the third failure probability based on the multiple failure significance values ​​to obtain the significant failure probability. Determine whether each failure significance value is greater than or equal to a failure significance value threshold. Output the failure probabilities corresponding to failure significance values ​​greater than or equal to the failure significance value threshold as significant failure probabilities. This provides a clear and targeted basis for subsequent failure correction, ensuring the targeted nature of failure correction and improving the measurement accuracy and operational stability of the pressure transmitter. The failure significance value threshold can be adaptively set and determined. Furthermore, if multiple failure significance values ​​are all less than the failure significance value threshold, it is determined that there is currently no significant failure probability, and a "no significant failure probability" prompt is output. In this case, no additional failure correction is required for the first calibration pressure signal.

[0040] Step A400: Perform causal modeling and retrospective correction on the first corrected pressure signal based on the significant failure probability to obtain the second corrected pressure signal. In one possible implementation, step A400 further includes steps A410 to A440.

[0041] Step A410: Match the multiple monitoring reinforcement vectors according to the significant failure probability to obtain a failure registration monitoring vector. Match the significant failure probability with the multiple monitoring reinforcement vectors, and record the monitoring reinforcement vector corresponding to the significant failure probability as the failure registration monitoring vector. If the significant failure probability only corresponds to parameter drift failure, then the first monitoring reinforcement vector is used as the failure registration monitoring vector. If the significant failure probability only corresponds to short-circuit failure, then the second monitoring reinforcement vector is used as the failure registration monitoring vector; if the significant failure probability only corresponds to open-circuit failure, then the third monitoring reinforcement vector is used as the failure registration monitoring vector. If the significant failure probability corresponds to two or three failure types, then the corresponding two or three monitoring reinforcement vectors are integrated to form a failure registration monitoring vector.

[0042] Step A420: Based on the failure registration monitoring vector, predict disturbance accidents for the pressure transmitter to obtain signal disturbance accident characteristics. First, construct a disturbance accident prediction model using a temporal neural network (LSTM) structure, adapted to the temporal data characteristics of the failure registration monitoring vector. Its input layer dimension is consistent with the dimension of the failure registration monitoring vector, with two hidden layers of 64 neurons each, and a multi-output layer with multiple output nodes corresponding to various characteristic parameters of the signal disturbance accident. The model training dataset uses the historical disturbance accident set of the pressure transmitter. The historical disturbance accident set contains cases of disturbance accidents caused by various failure types, each case containing complete information such as the corresponding historical failure monitoring vector, the historical disturbance accident occurrence time, and historical disturbance characteristic parameters. Divide the historical disturbance accident set into a 6:4 ratio into a disturbance accident training set and a disturbance accident validation set. Iteratively train the model using a cross-entropy loss function until the prediction accuracy of the disturbance accident validation set reaches a preset threshold (e.g., 93%), completing the construction of the disturbance accident prediction model. The failure registration monitoring vector is then input into the disturbance incident prediction model to predict potential disturbance incidents and output the corresponding signal disturbance incident characteristics. These characteristics include: disturbance type (e.g., signal jump, excessive noise, signal interruption), disturbance intensity (quantized as 0-1, with larger values ​​indicating more severe disturbances), predicted disturbance duration (in seconds), and the amplitude of the impact on the first corrected pressure signal (in MPa). By using the failure registration monitoring vector to predict disturbance incidents in the pressure transmitter, a complete set of signal disturbance incident characteristics is formed, ensuring the reliability and comprehensiveness of subsequent causal tracing and reverse correction.

[0043] Step A430: Based on the characteristics of the signal disturbance incident, perform causal modeling and tracing of the failure registration monitoring vector to establish a signal disturbance tracing model. First, extract the core features (disturbance type, disturbance intensity, etc.) from the characteristics of the signal disturbance incident, and combine them with the monitoring parameters in the failure registration monitoring vector to construct a causal graph model. The nodes of the causal graph model are divided into two categories: one category consists of the monitoring parameters in the failure registration monitoring vector (such as sensor power supply voltage, on-site electromagnetic interference intensity, etc.), serving as cause nodes; the other category consists of the core features in the characteristics of the signal disturbance incident, serving as result nodes. The causal graph model is the basic framework of the signal disturbance tracing model, used to intuitively present the correlation between the failure registration monitoring vector and the characteristics of the signal disturbance incident.

[0044] Subsequently, based on the historical disturbance event set of the pressure transmitter, the causal correlation strength between each causal node (monitoring parameter anomaly) and the result node (disturbance event) is statistically analyzed, and the conditional probability between each node is calculated. The conditional probability between nodes quantifies the degree of correlation between causal nodes and result nodes; that is, the probability of a disturbance event occurring when a certain monitoring parameter is abnormal. It is a core quantitative indicator for judging the reliability of causal correlation. Simultaneously, based on the influence of each causal node on the result node, the correlation weight between nodes is determined. The correlation weight ranges from 0 to 1; the larger the weight, the more significant the impact of the causal node (monitoring parameter) anomaly on the disturbance event, used to distinguish the importance priority of each parameter. Based on this, a disturbance cause tracing path is formed. The disturbance cause tracing path is constructed based on the causal graph model, the conditional probability between nodes, and the correlation weight between nodes, tracing the complete link from the result node (disturbance event) back to the causal node (monitoring parameter anomaly). The disturbance cause tracing path includes "disturbance event → core influencing parameter → root cause of parameter anomaly". Finally, the causal graph model, conditional probabilities between nodes, correlation weights between nodes, and the disturbance cause tracing path are integrated and encapsulated to form a signal disturbance tracing model. The signal disturbance tracing model includes the causal graph model, conditional probabilities between nodes, correlation weights between nodes, and the disturbance cause tracing path. Using this model, the causes of disturbances can be quickly and accurately traced, providing a clear causal guide for subsequent targeted reverse corrections and avoiding blind corrections.

[0045] Step A440: The first corrected pressure signal is reverse-corrected according to the signal disturbance tracing model to generate the second corrected pressure signal. Reverse correction refers to eliminating the impact of signal disturbances on the first corrected pressure signal based on the causal causes of the disturbance, ensuring the corrected signal accurately reflects the actual pressure. Specifically, firstly, the core cause parameters and corresponding influence weights of the current signal disturbance are extracted using the signal disturbance tracing model, clarifying the disturbance amplitude of each parameter anomaly on the first corrected pressure signal (i.e., the influence amplitude output in step A420). Then, for each core cause parameter, a corresponding correction coefficient is calculated based on its degree of anomaly and influence weight. The formula for calculating the correction coefficient is: Correction coefficient = 1 - (Parameter anomaly deviation rate × Influence weight), where the parameter anomaly deviation rate is the deviation rate between the core cause parameter and the center value of the factory-calibrated normal range. Finally, the correction coefficients corresponding to each core cause parameter are weighted and summed to obtain the total correction coefficient. Finally, the value of the first corrected pressure signal is multiplied by the total correction coefficient. Simultaneously, compensation correction is performed by incorporating the disturbance amplitude from the signal disturbance accident characteristics (i.e., if the disturbance causes the signal to be too large, the amplitude is subtracted; if the disturbance causes the signal to be too small, the amplitude is added). The corrected pressure signal, i.e., the second corrected pressure signal, is then calculated. Based on this, dual correction of nonlinear errors and failure disturbance errors is achieved, ensuring the output accuracy of the pressure transmitter.

[0046] Step A500: Based on the pressure signal set of the same node from the pressure transmitter, perform overcorrection backtracking compensation on the second corrected pressure signal to obtain a third corrected pressure signal, and convert and display the third corrected pressure signal. Step A500 further includes steps A510 to A550. Step A510: Preprocess the pressure signal set of the same node to obtain a standardized same-node signal set. The same-node pressure signal set includes multiple historical pressure signals from the pressure transmitter within the same monitoring time zone. The same monitoring time zone includes the same continuous monitoring period corresponding to the current timestamp. The specific process of preprocessing the same-node pressure signal set includes: First, data filtering, removing invalid data from the same-node pressure signal set. Invalid data includes missing signal values, jump values ​​(data deviating from the normal range by more than ±10%), and abnormal noise data. Second, data alignment, aligning the filtered same-node pressure signal set according to timestamps to ensure that each timestamp corresponds to one signal value, avoiding subsequent prediction errors caused by time deviations. The third step is to standardize the data. The units of the time-aligned pressure signal sets from the same node are uniformly converted to MPa (consistent with the unit of the second calibration pressure signal of the current pressure transmitter). Simultaneously, the signal values ​​are normalized, with the normalized signal values ​​ranging from 0 to 1, eliminating interference from inherent signal fluctuations. The fourth step is redundancy removal. The normalized signal set is deduplicated. If duplicate records exist at the same timestamp, only one signal value is retained. At the same time, continuous redundant data segments with signal variations less than 0.001 MPa are removed (such data lacks effective characteristics and increases computational complexity), resulting in a standardized signal set from the same node.

[0047] Step A520: Perform time series prediction on the current timestamp based on the standardized same-node signal set to obtain the current reference signal. The current timestamp is the time parameter corresponding to the current pressure signal. Time series prediction refers to calculating the predicted pressure signal of the pressure transmitter at the current timestamp using a Long Short-Term Memory (LSTM) network based on the time series characteristics of the standardized same-node signal set. This predicted pressure signal is the current reference signal. Preferably, this application uses an LSM network to perform time series prediction on the standardized same-node signal set to obtain the current reference signal. Further, step A530 is executed, and the second corrected pressure signal is over-corrected based on the current reference signal to obtain an over-correction determination result. The absolute value of the difference between the current reference signal and the second corrected pressure signal is recorded as the pressure signal deviation. It is determined whether the pressure signal deviation is greater than or equal to a pre-set signal deviation threshold to obtain the over-correction determination result. The over-correction determination result includes the presence of over-correction / absence of over-correction. If the pressure signal deviation is greater than or equal to the signal deviation threshold, the over-correction determination result is that over-correction exists. If the pressure signal deviation is less than the signal deviation threshold, the over-correction determination result is that over-correction does not exist.

[0048] Step A540: If the overcalibration determination result indicates the existence of overcalibration, the generation process of the second calibrated pressure signal is traced back to establish a signal generation process chain. If the overcalibration determination result indicates the existence of overcalibration, step A540 is executed. Generation process tracing back refers to reversing the complete generation process of the second calibrated pressure signal, systematically reviewing all processing steps from the original pressure signal to the second calibrated pressure signal, and locating the specific link leading to overcalibration. The signal generation process chain refers to the complete link formed by connecting the generation steps of the second calibrated pressure signal in chronological order. The signal generation process chain includes the processing parameters, processing results, and parameter adjustment status of all processing steps of the second calibrated pressure signal, used to accurately locate the root cause of overcalibration and provide a clear basis for subsequent overcalibration offset correction.

[0049] The specific process of backtracking the generation of the second corrected pressure signal includes: First, clarifying the complete generation process of the second corrected pressure signal, namely, current pressure signal (step A200) → linearization correction → first corrected pressure signal (step A200) → failure probability prediction and significant failure probability selection (steps A351-A390) → failure registration monitoring vector matching (step A410) → disturbance accident prediction (step A420) → signal disturbance tracing model establishment (step A430) → reverse correction → second corrected pressure signal (step A440). The second step involves backtracking each step according to the above process, extracting the core processing parameters and results for each step: including the linearization correction model parameters (slope k, intercept b) and the first correction pressure signal value in step A200; the significant failure probability types and values ​​in steps A351-A390; the failure registration monitoring vector parameters in step A410; the signal disturbance accident characteristic parameters in step A420; and the correction coefficient, total correction coefficient, disturbance impact amplitude, and compensation correction method in step A440. The third step involves verifying the processing parameters and results of each step, focusing on the reverse correction stage in step A440 (the stage where overcorrection is most likely to occur). This includes checking the calculation process of the correction coefficient, the value of the disturbance impact amplitude, and whether there are any deviations in the direction of compensation correction. Simultaneously, it involves checking for anomalies in the parameters of other steps (such as linearization correction model parameter drift and failure probability prediction deviation). The fourth step involves connecting all the backtracked steps, processing parameters, and processing results in sequence according to their generation order to form a signal generation process chain. The signal generation process chain clearly presents the input, processing, and output of each stage, clearly marks the parameter adjustment records of each stage, and identifies suspicious stages that may lead to overcalibration, ensuring that subsequent corrections can accurately target suspicious stages and avoid blind corrections.

[0050] Step A550: Perform overcorrection offset correction on the second corrected pressure signal according to the signal generation process chain to obtain the third corrected pressure signal.

[0051] Specifically, based on the signal generation process chain established in step A540, the root cause of overcorrection and the reason for the deviation are determined. If the root cause is excessive reverse correction in step A440, the correction coefficient and compensation correction amplitude are adjusted. If the root cause is drift in the linearization correction parameters in step A200, the slope k and intercept b of the linearization correction model are adjusted. If the root cause is the prediction deviation of the disturbance effect amplitude in step A420, the disturbance effect amplitude is recalculated. Furthermore, targeted overcorrection offset correction is performed. Taking the most common "excessive reverse correction in step A440" as an example, the specific correction process is as follows: First, the correction coefficient, total correction coefficient, disturbance effect amplitude, and deviation direction of step A440 in the signal generation process chain are extracted. If it is a positive deviation, the total correction coefficient is reduced. At this time, the correction coefficient adjustment formula is: adjusted total correction coefficient = original total correction coefficient - (absolute deviation / current reference signal), while reducing the compensation correction amplitude (or increasing the decrement). If the deviation is negative, the total correction coefficient is increased. In this case, the correction coefficient adjustment formula is: adjusted total correction coefficient = original total correction coefficient + (absolute deviation / current reference signal). At the same time, the amplitude of the compensation correction is increased (or the amplitude is decreased) to obtain the third correction pressure signal.

[0052] Finally, the third calibration pressure signal is converted and displayed. First, the third calibration pressure signal (MPa unit) is converted into an industrial standard 4-20mA DC electrical signal. The conversion formula is: electrical signal value (mA) = 4 + (third calibration pressure signal / upper limit of full range) × 16, where the upper limit of the full range is 40MPa. The converted electrical signal is then transmitted to the pressure transmitter's display module, simultaneously displaying the pressure value in MPa and the 4-20mA DC electrical signal value, with the display accuracy retaining 4 decimal places. This achieves adaptive enhancement of the pressure transmitter's sensitivity under variable range conditions, real-time identification and suppression of multi-mode failure errors, and high-precision stable output of the pressure transmitter through causal tracing and same-node signal compensation.

[0053] Step A530 further includes step A531. Step A531: If the overcalibration determination result indicates that there is no overcalibration phenomenon, the second calibrated pressure signal is converted and displayed. Specifically, when the overcalibration determination result indicates that there is no overcalibration phenomenon, step A531 is executed. The second calibrated pressure signal is converted from a numerical pressure signal to a standard output form commonly used in industrial settings using the above conversion formula, and then displayed on the display module of the pressure transmitter.

[0054] In summary, the pressure transmitter output linearization correction method based on range ratio adaptation provided in this application has the following technical advantages: A linearized correction model is established by performing full-range multi-point calibration on the factory test dataset of the pressure transmitter. Based on the current range ratio, the linearized correction model adaptively corrects the current pressure signal to obtain the first corrected pressure signal. Multiple monitoring enhancement vectors are obtained by enhancing the multi-mode failure contribution of the multi-modal monitoring sequence, and the multi-mode failure probability of the pressure transmitter is predicted based on these vectors to determine the significant failure probability. The first corrected pressure signal is then corrected using causal modeling and tracing based on the significant failure probability to obtain the second corrected pressure signal. The second corrected pressure signal is then overcorrected and back-compensated based on the pressure signal set of the same node of the pressure transmitter to obtain the third corrected pressure signal, which is then converted and displayed. This solves the technical problems in existing technologies, such as insufficient correction accuracy of pressure transmitters under different range ratios, output distortion caused by interference under multiple operating conditions, inability to quantify and correct failure factors in real time, and the tendency for overcorrection deviation. It achieves the technical effect of adaptively improving the sensitivity of the pressure transmitter under variable range conditions, real-time identification and suppression of multi-mode failure errors, and high-precision stable output of the pressure transmitter through causal tracing and compensation of signals from the same node.

[0055] The specific embodiments described above do not constitute a limitation on the scope of protection of this application. Those skilled in the art should understand that various modifications, combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. A method for linearizing the output of a pressure transmitter based on adaptive range ratio, characterized in that, The method includes: A full-range multi-point calibration was performed on the factory test dataset of the pressure transmitter to establish a linear correction model; The current pressure signal, current range ratio, and multimodal monitoring sequence of the pressure transmitter are acquired, and the linearization correction model is guided to adaptively correct the current pressure signal based on the current range ratio to obtain a first corrected pressure signal. The multi-mode monitoring sequence is subjected to multi-mode failure contribution attention enhancement to obtain multiple monitoring enhancement vectors, and the multi-mode failure probability of the pressure transmitter is predicted based on the multiple monitoring enhancement vectors to determine the significant failure probability. Based on the significant failure probability, the first corrected pressure signal is causally modeled and retrospectively corrected to obtain the second corrected pressure signal; The second corrected pressure signal is overcorrected and backtracked to compensate based on the pressure signal set of the same node of the pressure transmitter to obtain the third corrected pressure signal, and the third corrected pressure signal is converted and output for display.

2. The method as described in claim 1, characterized in that, The factory test dataset of the pressure transmitter was used for full-range multi-point calibration, and a linear correction model was established, including: The factory test dataset is cleaned and extracted to obtain multiple pressure excitation test sequences corresponding to multiple range ratios; Anomaly identification and correction are performed on the multiple stress excitation test sequences to obtain multiple reliable stress test sequences; The multiple reliable pressure test sequences are fitted using the least squares method to obtain multiple signal correction models; The multiple range ratios are used as multiple correction drive pointers, and the multiple signal correction models are driven and encapsulated according to the multiple correction drive pointers to generate the linearized correction model.

3. The method as described in claim 1, characterized in that, The multimodal monitoring sequence is subjected to multimodal failure contribution attention enhancement to obtain multiple monitoring enhancement vectors, including: The parameter drift failure contribution of the multimodal monitoring sequence is evaluated to obtain the failure contribution of each parameter. The multimodal monitoring sequence is filtered and attention is allocated based on the drift failure contribution of each parameter to obtain the first monitoring enhancement vector; The multimodal monitoring sequence is subjected to short-circuit failure contribution attention enhancement to obtain a second monitoring enhancement vector; The multimodal monitoring sequence is subjected to open-circuit failure contribution attention enhancement to obtain a third monitoring enhancement vector, and the first monitoring enhancement vector and the second monitoring enhancement vector are combined to generate the plurality of monitoring enhancement vectors.

4. The method as described in claim 3, characterized in that, Based on the multiple monitoring enhancement vectors, a multi-mode failure probability prediction is performed on the pressure transmitter to determine the significant failure probability, including: Based on the first monitoring enhancement vector, the parameter drift failure probability of the pressure transmitter is predicted, and the first failure probability is determined. Based on the second monitoring enhancement vector, the pressure transmitter is predicted for short-circuit failure probability to determine the second failure probability. Based on the third monitoring enhancement vector, the open-circuit failure probability of the pressure transmitter is predicted, and the third failure probability is determined. Based on a predetermined failure probability, a significance test is performed on the first failure probability, the second failure probability, and the third failure probability to obtain multiple failure significances; The first failure probability, the second failure probability, and the third failure probability are selected based on the multiple failure significance values ​​to obtain the significant failure probability.

5. The method as described in claim 4, characterized in that, Based on the first monitoring enhancement vector, the pressure transmitter is used to predict the parameter drift failure probability and determine the first failure probability, including: Based on the drift failure event set of the pressure transmitter, the Bayesian network is iteratively trained to obtain the first model for drift failure prediction. The temporal neural network is iteratively trained based on the drift failure event set to obtain a second drift failure prediction model. The first monitoring reinforcement vector is predicted for failure characteristics based on the first drift failure prediction model and the second drift failure prediction model, respectively, to obtain the first probability of drift failure and the second probability of drift failure. Based on the first accuracy and second accuracy of drift failure prediction, a drift failure fusion node is constructed; The first probability of drift failure and the second probability of drift failure are input into the drift failure fusion node, and the first failure probability is output.

6. The method as described in claim 1, characterized in that, Based on the significant failure probability, the first corrected pressure signal is causally modeled and retrospectively corrected to obtain the second corrected pressure signal, including: Based on the significant failure probability, match the multiple monitoring enhancement vectors to obtain the failure registration monitoring vector; Based on the failure registration monitoring vector, the pressure transmitter is used to predict disturbance accidents and obtain signal disturbance accident characteristics. Based on the characteristics of the signal disturbance incident, a causal modeling and tracing method is used to trace the failure registration monitoring vector, and a signal disturbance tracing model is established. The first corrected pressure signal is reverse-corrected according to the signal disturbance tracing model to generate the second corrected pressure signal.

7. The method as described in claim 1, characterized in that, Based on the pressure signal set of the same node of the pressure transmitter, the second corrected pressure signal is overcorrected and backtracked to obtain the third corrected pressure signal, including: The pressure signal set at the same node is preprocessed to obtain a standardized signal set at the same node; Based on the standardized same-node signal set, perform time series prediction on the current timestamp to obtain the current reference signal; Based on the current reference signal, the second calibration pressure signal is overcalibrated to determine the overcalibration result. If the overcalibration determination result indicates that overcalibration exists, the generation process of the second calibration pressure signal is traced back to establish a signal generation process chain. The second corrected pressure signal is overcorrected and offset according to the signal generation process chain to obtain the third corrected pressure signal.

8. The method as described in claim 7, characterized in that, Obtain the overcorrection determination result, including: If the overcalibration determination result indicates that there is no overcalibration phenomenon, the second calibration pressure signal is converted and output for display.

Images