Pressure sensor operation state monitoring method based on deviation rate analysis

Abstract

The invention relates to the technical field of pressure sensor operation state monitoring, in particular to a pressure sensor operation state diagnosis method based on deviation rate analysis, which comprises the following steps: firstly, pressure sensor data sets in a plurality of time periods are extracted from historical data, and each data set comprises an actual measured value and a set value; and the running state of the sensor is reflected by calculating the deviation rate of each data point. Secondly, preprocessing the deviation rate data, and removing noise and abnormal values; thirdly, performing time sequence analysis on the deviation rate data by adopting an autoregression (AR) model, and capturing the variation trend of the deviation rate; and finally, dynamically judging whether the sensor needs to be replaced or not by combining a set threshold value based on an analysis result. The operation state of the sensor can be monitored in real time, early signs of performance degradation can be captured in time, and the replacement opportunity can be accurately judged in a data driving mode. According to the method, the production efficiency is guaranteed, and the maintenance cost and unnecessary resource waste of enterprises are effectively reduced.

Description

Technical Field

[0001] The present invention relates to the field of pressure sensor operating status monitoring, and in particular to a pressure sensor operating status monitoring method based on deviation rate analysis. Background Art

[0002] Pressure transducers, devices that sense pressure signals and convert them into usable electrical signals, are widely used in industrial production. Their primary function is to convert physical pressure changes into electrical signals proportional to the pressure value, enabling control systems to monitor and adjust the equipment's operating status in real time. The core components of a pressure sensor consist of a pressure-sensitive element and a signal processing unit. Depending on the pressure measurement requirements, pressure sensors can be divided into several types, such as gauge pressure sensors (which measure pressure relative to atmospheric pressure), differential pressure sensors (which measure the difference between two pressures), and absolute pressure sensors (which measure pressure relative to a vacuum). These different types of pressure sensors are suitable for diverse applications, such as measuring water flow pressure in water conservancy and hydropower projects, monitoring oil well pressure in the petrochemical industry, and even controlling air pressure in railway carriages. In industrial production, pressure sensors are widely used to control equipment operation, ensure the stability of production processes, and guarantee product quality. They not only monitor the pressure of fluid systems but also provide real-time data feedback for fault prediction, performance optimization, and automatic adjustments. For example, in smart buildings, pressure sensors are used to monitor pressure in water supply and air conditioning systems. In the petrochemical industry, they monitor the operating pressure of oil wells and pipelines to prevent equipment damage and leaks caused by excessive pressure. However, with age, pressure sensors inevitably experience issues such as aging, drift, and even failure. Sensor performance degradation typically manifests as reduced measurement accuracy or sluggish response, impacting the operation of the entire system. More seriously, sensor failure often leads to malfunctioning production equipment, resulting in significant economic losses. For example, in certain industrial applications, pressure sensor failure can cause production line downtime, resulting in production losses and even impacting the quality of the final product, leading to customer complaints, compensation, and brand damage.

[0003] Therefore, accurately assessing the health of pressure sensors and rationally scheduling their replacement cycles presents a current technical challenge. Traditionally, production sites often employ a periodic replacement approach, replacing sensors at the end of their expected lifespan. However, this approach is not always optimal. Excessively long replacement cycles can result in sensors failing to be promptly identified and replaced after aging or failure, impacting product quality or leading to production accidents. Excessively short replacement cycles can lead to sensors being replaced prematurely, resulting in unnecessary resource waste and increased costs. Because each sensor's operating environment and workload vary, the traditional periodic replacement approach struggles to precisely control its performance. Therefore, a more scientific, real-time monitoring and maintenance solution is needed. To minimize maintenance costs, improve production efficiency, and ensure proper equipment operation, it is crucial to rationally schedule pressure sensor replacement cycles and maintenance strategies. This not only helps extend sensor lifespan but also avoids wasteful maintenance, ensuring equipment is always in optimal working condition, thus significantly saving manufacturers significant operating costs. Therefore, developing a monitoring method based on real-time sensor data analysis and prediction that can dynamically assess sensor health and provide decision support for optimal replacement timing is an urgent challenge.

[0004] The purpose of the present invention is to effectively monitor the operating status of the sensor and promptly capture early signs of performance degradation by calculating the deviation rate of the sensor and analyzing its time series changes. Specifically, the deviation rate reflects the difference between the actual measurement value of the sensor and the set target value. Through dynamic analysis of its time series, regular changes or abnormal fluctuations that may occur in the sensor during the degradation process can be identified, thereby discovering potential failure risks or performance degradation early. At the same time, dynamic evaluation based on the real-time operating status of the sensor can accurately determine whether it needs to be replaced. By continuously tracking and analyzing the changing trend of the deviation rate, combined with prediction methods such as autoregressive models, the health status of the sensor can be predicted, and whether the time for replacement has been reached can be determined based on the degree of degradation of the sensor. Summary of the Invention

[0005] The purpose of the present invention is to address the above problems and provide a method for monitoring the operating status of a pressure sensor based on deviation rate analysis.

[0006] The object of the present invention is achieved as follows: a method for monitoring the operating status of a pressure sensor based on deviation rate analysis, comprising the following steps: Step 1: selecting a pressure sensor data set for 3-5 time periods from a real operating status data set before the current time of the pressure sensor, wherein the data set for each time period has pressure sensor data of the same duration and the same amount of data, and each data piece includes a set value and a measured value at the same moment, wherein the set value represents the preset or expected target pressure value of the pressure sensor at that moment, and the measured value is the pressure value actually measured by a pressure measurement tool. After preprocessing the pressure sensor data set for each time period, the pressure sensor deviation rate at each moment is calculated using the formula:

[0007]

[0008] Among them, α m is the measured value of the pressure sensor, β m is the pressure sensor set value, α m , β m The unit is KN. The measured value and set value in each data are converted into the pressure sensor deviation rate through formula (1). Finally, the pressure sensor deviation rate data set of each time period is obtained. The pressure sensor deviation rate data set of each time period is split into a training set and a validation set according to the ratio of 7:3. Step 2: In order to establish the AR model of the time series, the data collected in 3-5 time periods are introduced as experimental data to realize the prediction of the deviation rate at the future time t. The formula is expressed as follows:

[0009]

[0010] Among them, X t Indicates the deviation rate of the pressure sensor at time t, in percentage, i.e. %, is the autoregressive coefficient, X t-1 Indicates the deviation rate of the pressure sensor at time t-1, the unit is percentage, that is, %, X t-2 Indicates the deviation rate of the pressure sensor at time t-2, in percentage, i.e., %, X t-p Indicates the deviation rate of the pressure sensor at time tp, in percentage, i.e., μ t is a random term, representing the white noise at time t, and p is the order of the AR model; Step 3: Based on the training set of the same representative time period, the time series prediction model, namely the AR model, is selected to obtain the optimal AR model. The AIC information criterion, namely the Akaike information criterion, is used as the measure of the quality of the AR model. The smaller the AIC value, the better the model fit and the moderate complexity. The parameters obtained by the AR model can be used to calculate the AIC value of each time period. The calculation formula is as follows:

[0011] AIC=2k-2log(L)

[0012] Where k is The number of parameters, L is the maximum value of the likelihood function, that is, the maximum likelihood of the model on the observed data. The AR model with the smallest AIC value is selected as the optimal model, and then the optimal model is applied to the corresponding pressure sensor data set. The deviation rate value of the device at time t is predicted by the time series. The operating status of the corresponding pressure sensor is found by combining the deviation rate value with the pressure sensor life prediction operation, and the life of the pressure sensor is predicted based on the prediction results. The prediction results also need to be tested on the test set.

[0013] In step 1, the data set of each time period is preprocessed to remove noise data and outliers.

[0014] The prediction of the life of the pressure sensor in step 3 is to divide the life of the pressure sensor into three stages, namely the normal period, the degradation period, and the failure period, and calculate the average value of the deviation rate ∈ ave value, set the alarm threshold to ∈ ave +0.05σ, set the failure threshold ∈ ave +3σ, σ is the standard deviation of the deviation rate. When the mean value of the deviation rate in one-minute continuous data is less than or equal to the alarm threshold, it indicates that the life of the pressure sensor is in the normal period; when the mean value of the deviation rate in one-minute continuous data is less than the failure threshold but greater than the alarm threshold, the life of the pressure sensor is in the degradation period; when the mean value of the deviation rate in one-minute continuous data is greater than or equal to the failure threshold, the pressure sensor enters the failure period.

[0015] The beneficial effects of the present invention are as follows: 1. Real-time monitoring and early warning of faults: The present invention can monitor the operating status of pressure sensors in real time and dynamically evaluate changes in the sensor's deviation rate using a predictive model based on historical data analysis (such as an autoregressive model). By continuously tracking and analyzing sensor operating data, early warnings can be issued when the sensor shows signs of degradation or failure, thereby avoiding production line downtime or equipment damage caused by sudden failures. This early warning mechanism can greatly improve the reliability of production equipment and avoid production accidents and potential safety hazards.

[0016] 2. Accurately Determine Replacement Timing: Traditional sensor replacement methods often rely on periodic replacement or empirically defined intervals, ignoring the sensor's actual operating conditions. This invention analyzes the pressure sensor's deviation rate data to determine whether replacement is necessary based on the sensor's actual aging state, enabling more accurate and scientific maintenance decisions. Specifically, when a sensor enters its degradation stage, the system accurately determines whether it needs replacement, avoiding premature or late replacement.

[0017] 3. Cost Savings: This invention not only helps extend the lifespan of pressure sensors but also significantly reduces unnecessary replacement costs. By optimizing sensor replacement cycles, it avoids the waste of resources and increased costs associated with premature replacement. Furthermore, timely identification of potential sensor failures can prevent downtime and production losses caused by sensor failure during production, thereby saving companies significant maintenance and production costs.

[0018] 4. Ensure production continuity and product quality: This invention effectively avoids production interruptions due to sensor failures by monitoring the operating status of pressure sensors in real time. This not only ensures production line continuity but also ensures stable final product quality. In key industries such as aerospace, petrochemicals, and intelligent buildings, product quality and production safety are crucial. Promptly replacing failed sensors ensures that all parameters in the production process remain within the appropriate range, thus avoiding quality fluctuations caused by equipment problems. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] The present invention will be further described below with reference to the accompanying drawings.

[0020] Figure 1 This is an image of the equipment's operating data (deviation rate) from 06:00 to 08:00 (representative).

[0021] Figure 2 This is a pressure sensor life prediction chart. DETAILED DESCRIPTION

[0022] Pressure sensors are widely used in industrial production processes to monitor pressure and ensure equipment stability and product quality. However, over time, pressure sensors may age or malfunction. If not detected promptly, this can lead to production downtime or product quality issues. Therefore, the present invention provides an effective monitoring method that promptly diagnoses the operating status of sensors and accurately determines whether replacement is necessary. The technical solution of the present invention includes the following steps: First, pressure sensor datasets covering multiple time periods are extracted from historical data. Each time period contains 72,000 raw data points, and each dataset includes both actual measured values ​​and set values. The deviation rate of each data point is calculated to reflect the sensor's operating status. The deviation rate data is then preprocessed to remove noise and outliers. Next, an autoregressive (AR) model is used to perform time series analysis on the deviation rate data to capture the deviation rate's changing trends. Finally, based on the analysis results and a set threshold, a dynamic determination is made as to whether the sensor needs replacement, thus avoiding premature or delayed replacement. The present invention enables real-time monitoring of the sensor's operating status, promptly detecting early signs of performance degradation, and accurately determining replacement timing through a data-driven approach. This method not only ensures production efficiency but also effectively reduces maintenance costs and unnecessary resource waste for enterprises.

[0023] The technical solution adopted by the present invention is: a method for diagnosing the operating status of a pressure sensor, using the pressure sensor to control the rolling process of a rolling mill to achieve industrial production, and is carried out according to the following steps.

[0024] Step 1: Select pressure sensor data sets for multiple time periods from the real operating status data set before the current time of the pressure sensor. The data set for each time period has multiple pressure sensor data with the same duration and the same amount of data. Each time period contains 72,000 pieces of raw data, and each piece of data includes the set value and measured value at the same moment. Among them, the set value represents the target pressure value preset or expected by the pressure sensor at that moment, and the measured value is the pressure value actually measured by the pressure measurement tool. After preprocessing the pressure sensor data set for each time period (removing noise data and outliers), calculate the pressure sensor deviation rate at each moment. The formula is:

[0025]

[0026] Among them, α m is the measured value, β m The measured and set values ​​in each data point are converted into pressure sensor deviation rates, forming a pressure sensor deviation rate dataset for multiple time periods. To better evaluate the performance and robustness of different models on the same dataset, the pressure sensor deviation rate dataset for each time period is split into a training set and a validation set in a ratio of 7:3.

[0027] Step 2: The AR (autoregressive) model is a commonly used time series analysis method, which is suitable for predicting future values ​​based on observation data from the past. In the prediction of the deviation rate of the pressure sensor, the AR model can capture the autocorrelation of the time series in the data, that is, the change in the deviation rate often has a certain dependence on the deviation rate of the previous moment. Since the deviation rate of the pressure sensor usually has time correlation, the AR model predicts the deviation rate at the current moment by introducing data from the previous moments, and has good prediction ability and low computational complexity. Selecting the AR model can effectively use historical data to model the behavior of the pressure sensor, especially when the time series data shows a stable trend and periodic changes, the AR model can provide more accurate prediction results. Establish a time series prediction model AR model, expressed using the following formula:

[0028]

[0029] Among them, X t represents the deviation rate of the pressure sensor at time t, is the autoregressive coefficient, X t-1 represents the deviation rate of the pressure sensor at time t-1, X t-2 represents the deviation rate of the pressure sensor at time t-2, X t-p Indicates the deviation rate of the pressure sensor at time tp, μ t is a random term, representing the white noise at time t, and p is the order of the AR model.

[0030] Step 3: Based on a training set of representative time periods (reflecting both normal and faulty conditions of the pressure sensor), provided by experts in the relevant field, the training set includes data from both normal equipment operation and the corresponding time periods when equipment failures occur. This is an existing dataset. The time series prediction model (AR) is trained to obtain the optimal AR model. When selecting a representative time period, it is important to ensure that it fully reflects both normal equipment operation and potential fault conditions, providing sufficient information for model training. To evaluate the performance of the AR model across different time periods, the Akaike Information Criterion (AIC) is used as a metric. A smaller AIC value indicates a better model fit and moderate complexity, avoiding overfitting or underfitting. By calculating the AIC value for each time period, the AR model with the smallest AIC value is selected as the optimal model. This model more accurately fits the deviation rate trend of the pressure sensor and, therefore, predicts the sensor lifespan. This optimal model is then applied to the corresponding pressure sensor dataset. The time series prediction results are used to infer the sensor's future operating state and, based on these predictions, the sensor lifespan is predicted. To verify the model's generalization and accuracy, the prediction results are tested on a test set. By comparing the predicted values ​​on the test set with the actual values, we can further evaluate the AR model's performance in real-world applications, ensuring its ability to stably and effectively predict pressure sensor lifespan across different time periods and datasets. This process not only helps optimize the prediction model but also provides reliable decision support for equipment maintenance and fault warning.

[0031] As a preferred approach, in step 1, during preprocessing of the dataset for each time period, noise and outliers are removed to ensure data quality. Simultaneously, the measured and set values ​​in each data item are converted into pressure sensor deviation rates, ultimately forming a pressure sensor deviation rate dataset for multiple time periods.

[0032] As a preferred method: in step 3, the life prediction of the pressure sensor is to divide the life cycle of the pressure sensor into three stages, namely the normal period, the degradation period, and the failure period, and calculate the average value of the deviation rate ∈ ave value, set the alarm threshold to ∈ ave +0.05σ, set the failure threshold ∈ ave+3σ, where σ is the standard deviation of the deviation rate. When the average deviation rate over a continuous one-minute period is less than or equal to the alarm threshold, the pressure sensor's lifespan is normal. When the average deviation rate over a continuous one-minute period is less than the failure threshold but greater than the alarm threshold, the pressure sensor's lifespan is degraded. When the average deviation rate over a continuous one-minute period is greater than or equal to the failure threshold, the pressure sensor has entered its failure phase.

[0033] Example

[0034] A method for diagnosing the operating status of a pressure sensor, which uses the pressure sensor to control the operation of equipment and realize industrial production, is carried out in the following steps.

[0035] Step 1: Select multiple representative time periods (the selection of these multiple time periods needs to be for pressure sensors of the same model and the same purpose) from the real operating status data set of the pressure sensor before the current time, where each data includes the set value and the measured value at the same moment. The set value represents the target pressure value preset or expected by the pressure sensor at that moment, and the measured value is the pressure value actually measured by the pressure measurement tool. After preprocessing the pressure sensor pressure data set for each time period, the deviation rate of the pressure sensor at time m is used to calculate the pressure sensor pressure. α m is the measured value, β m The measured values ​​and predicted values ​​in each data are converted into pressure sensor deviation rates to form pressure sensor deviation rate data sets for multiple time periods. In order to better evaluate the performance and robustness of different models on the same data set, the pressure sensor deviation rate data set for each time period is split into a training set and a validation set in a ratio of 7:3. The preprocessing of the pressure sensor pressure data set for each time period means that when preprocessing the data set for each time period, noise data and outliers need to be removed to ensure data quality. At the same time, the measured values ​​and set values ​​in each data are converted into pressure sensor deviation rates, and finally pressure sensor deviation rate data sets for multiple time periods are formed.

[0036] In one embodiment, a dataset was studied for the actual operation of a pressure sensor during the rolling process. Time periods with both normal operation and failures were selected and divided into five sub-time periods. Each dataset contained 72,000 data points. Given the characteristic of a dataset containing 10 data points per second, to further improve prediction accuracy and cycle time, the data was preprocessed based on the operating characteristics of the pressure sensor. The average value of the measured values ​​within each unit time was calculated in seconds, reducing the number of observations in each dataset from 72,000 to 7,200. The resulting dataset is shown in Table 1.

[0037] Table 1 Dataset Description

[0038]

[0039]

[0040] Step 2: Establish a time series prediction model AR model, using the following formula

[0041]

[0042] Among them, X t represents the deviation rate of the pressure sensor at time t, is the autoregressive coefficient, X t-1 represents the deviation rate of the pressure sensor at time t-1, X t-2 represents the deviation rate of the pressure sensor at time t-2, X t-p Indicates the deviation rate of the pressure sensor at time tp, μ t is a random term, representing the white noise at time t, and p is the order of the AR model.

[0043] To better understand the data, take the period of 06:00-08:00 as an example and draw a data graph, as shown in the following figure: Figure 1 .pass Figure 1 Direct observation shows that the data has no periodic changes and no obvious trend, indicating that the series is stable and meets the requirements of the ARMA model. The data is tested for autocorrelation coefficient and partial autocorrelation coefficient, and the autocorrelation coefficient and partial autocorrelation coefficient results for each period are obtained. It is found that the autocorrelation coefficient graph is tailing, and the partial autocorrelation coefficient graph gradually approaches 0 as the number of periods increases. Therefore, it can be modeled and predicted according to the AR model.

[0044] Step 3: Based on the data of the same time period, we divide it into a training set and a validation set, and use this data to train the AR (autoregressive) time series prediction model. During the training process, we set different model parameters and generate multiple AR models for different time periods. Then, we use AIC (Akaike Information Criterion) to evaluate the performance of each AR model, and finally select the model with the smallest AIC value as the optimal model. Then, we use the pressure sensor data set corresponding to the optimal AR model for life prediction. In the specific case, we use the data of the five time periods shown in Table 1 as the training set and validation set, and use AR models with different parameters based on the same time period for training. Based on the autocorrelation coefficient and partial autocorrelation coefficient, we preliminarily set five AR models with different parameter values, namely: AR(3,0,0), AR(5,0,0), AR(7,0,0), AR(9,0,0) and AR(11,0,0). Then, using the AIC criterion, we calculate the frequency of selecting the optimal model in different time periods, that is, the number of times these models are selected as the optimal model in each time period. The results are shown in Table 2 below:

[0045] Table 2 Frequency of AR models with different parameters being the optimal model in each time period

[0046]

[0047] Table 2 shows the frequency of each model being the optimal model in different time periods. Based on this frequency, the AR(7, 0, 0) model was finally determined to predict the life of the pressure sensor.

[0048] Time Series Model Forecasting and Accuracy Analysis: To verify the accuracy and effectiveness of the model, we selected data not used in training as a test set and evaluated it according to the following criteria: If both the test set and the predicted data were normal points, the prediction was considered correct; if both the test set and the predicted data were outliers, the prediction was considered correct; if the test set contained normal points and the predicted data were outliers, the prediction was considered incorrect; if the test set contained outliers and the predicted data were normal points, the prediction was considered incorrect. The AR(7,0,0) model's prediction results for the test set included 1256 normal points, 99 outliers, and 85 outliers; the predicted data included 1257 normal points, 97 outliers, and 69 outliers. Based on these data, the model achieved an accuracy of 98.75%. To further verify the effectiveness of the model, we used the AR(7,0,0) model to predict 10 segments of the untrained test set, achieving an average accuracy of 96.14%. This result shows that the model has high prediction accuracy and can accurately predict the operating status of the pressure sensor. At the same time, the mean square error of the data corresponding to the (7,0,0) model is calculated as the judgment standard of the model's accuracy index. The mean square error values ​​for the training set and test set are shown in Table 3.

[0049] Table 3 Mean square error of the model in the training set and test set

[0050]

[0051] From the calculation results in the above table, it can be seen that the mean square error value of the model is small, which means that the model has high accuracy and can accurately predict the operating status of the pressure sensor.

[0052] Predicting the life of a pressure sensor means that the life of the pressure sensor can be divided into three stages: normal period, degradation period, and failure period. Figure 2 As shown. The average value of the deviation rate ∈ is calculated through the pressure sensor pressure data set ave value, set the alarm threshold to ∈ ave +0.05σ, set the failure threshold ∈ ave +3σ, where σ is the standard deviation of the deviation rate. When the average deviation rate over a continuous one-minute period is less than or equal to the alarm threshold, the pressure sensor's lifespan is normal. When the average deviation rate over a continuous one-minute period is less than the failure threshold but greater than the alarm threshold, the pressure sensor's lifespan is degraded. When the average deviation rate over a continuous one-minute period is greater than or equal to the failure threshold, the pressure sensor has entered its failure phase.

[0053] In one embodiment, the life cycle of the pressure sensor can be divided into three stages. By extracting the data of the pressure sensor and using the deviation rate as the research variable, Figure 2 It can be seen that the equipment operation time in the normal period is 240h; according to the set data category criteria, using ∈ ave +0.05σ sets the alarm threshold, and defines that if there are abnormal points for 5 consecutive hours, the next time is the degradation starting point. When the running time is about 264 hours, the deviation rate exceeds the alarm threshold for the first time, indicating that the equipment has entered the degradation period, the bearing begins to degrade, and gradually increases. As the degradation coefficient gradually increases, when the deviation rate exceeds the failure threshold (∈ ave +3σ), the bearing can no longer work, and the working time is 412h. The life of the pressure sensor is as follows Figure 2 .

[0054] The above description is only a specific embodiment of the present invention, but the structural features protected by the present invention are not limited thereto. Any changes or modifications made by any technician in this field within the scope of the present invention are included in the patent scope of the present invention.

Claims

1. A method for monitoring the operating status of a pressure sensor based on deviation rate analysis, characterized in that: The following steps are involved: Step 1: Select 3-5 time periods of pressure sensor data from the actual operating status data set before the current time. Each time period has the same duration and amount of pressure sensor data. Each data entry includes the set value and measured value at the same moment. The set value represents the preset or expected target pressure value of the pressure sensor at that moment, and the measured value is the pressure value actually measured by the pressure measurement tool. After preprocessing the pressure sensor data set for each time period, calculate the pressure sensor deviation rate at each moment. The formula is: Among them, α m is the measured value of the pressure sensor, β m is the pressure sensor set value, α m , β m The unit is KN. The measured value and set value in each data are converted into the pressure sensor deviation rate through formula (1). Finally, the pressure sensor deviation rate data set of each time period is obtained. The pressure sensor deviation rate data set of each time period is split into a training set and a validation set in a ratio of 7:

3. Step 2: To establish the AR model of time series, the data collected over 3-5 time periods are introduced as experimental data to predict the deviation rate at the next moment t. The formula is as follows: Among them, X t Indicates the deviation rate of the pressure sensor at time t, in percentage, i.e. %, is the autoregressive coefficient, X t-1 Indicates the deviation rate of the pressure sensor at time t-1, the unit is percentage, that is, %, X t-2 Indicates the deviation rate of the pressure sensor at time t-2, in percentage, i.e., %, X t-p Indicates the deviation rate of the pressure sensor at time tp, in percentage, i.e., μ t is a random term, representing the white noise at time t, and p is the order of the AR model; Step 3: Based on the training set of the same representative time period, the time series prediction model, namely the AR model, is selected to obtain the optimal AR model. The AIC information criterion, namely the Akaike information criterion, is used as the measure of the quality of the AR model. The smaller the AIC value, the better the model fit and the moderate complexity. The parameters obtained by the AR model can be used to calculate the AIC value of each time period. The calculation formula is as follows: AIC=2k-2log(L) Where k is The number of parameters, L is the maximum value of the likelihood function, that is, the maximum likelihood of the model on the observed data. The AR model with the smallest AIC value is selected as the optimal model. Then, the optimal model is applied to the corresponding pressure sensor data set. The deviation rate value of the device at time t is predicted by the time series. The operating status of the corresponding pressure sensor is found by combining the deviation rate value with the pressure sensor life prediction operation, and the life of the pressure sensor is predicted based on the prediction results. The prediction results also need to be tested on the test set.

2. The method for monitoring the operating status of a pressure sensor based on deviation rate analysis according to claim 1, characterized in that: In step 1, the data set of each time period is preprocessed to remove noise data and outliers.

3. The method for monitoring the operating status of a pressure sensor based on deviation rate analysis according to claim 1, characterized in that: The prediction of the life of the pressure sensor in step 3 is to divide the life of the pressure sensor into three stages, namely the normal period, the degradation period, and the failure period, and calculate the average value of the deviation rate ∈ ave value, set the alarm threshold to ∈ ave +0.05σ, set the failure threshold ∈ ave +3σ, σ is the standard deviation of the deviation rate. When the mean value of the deviation rate in one-minute continuous data is less than or equal to the alarm threshold, it indicates that the life of the pressure sensor is in the normal period; when the mean value of the deviation rate in one-minute continuous data is less than the failure threshold but greater than the alarm threshold, the life of the pressure sensor is in the degradation period; when the mean value of the deviation rate in one-minute continuous data is greater than or equal to the failure threshold, the pressure sensor enters the failure period.

Images