Real-time alarming method and system for thickness abnormity of PVC (Polyvinyl Chloride) pipeline

Abstract

The invention discloses a real-time alarming method for thickness abnormity of a PVC (polyvinyl chloride) pipeline, which comprises the following steps: acquiring thickness data, temperature parameters and pressure parameters of the PVC pipeline, and constructing multi-dimensional monitoring data; processing the multi-dimensional monitoring data to form purified monitoring data; executing multivariate nonlinear regression analysis according to the purification monitoring data, and establishing a coupling relation expression; based on the coupling relation expression, calculating a deviation sequence between a thickness predicted value and a measured value by adopting a sliding window technology, and extracting a deviation trend index reflecting an abnormal change mode; performing differential analysis according to the deviation trend index, and extracting duration, strength grade and service attribute set; calculating a credibility value of the alarm event according to the strength grade, the duration time and the service attribute set; based on the credibility value, a preset alarm threshold value is adjusted, parameter configuration in the coupling relation expression is iteratively corrected, and the method is used for continuous detection and self-adaptive adjustment of the thickness abnormity of the PVC pipeline. According to the method, accurate identification of early-stage thickness abnormity and high-reliability alarm generation can be realized.

Description

Technical Field

[0001] The present invention relates to the technical field of pipeline thickness monitoring, and in particular to a method, system, electronic equipment and storage medium for real-time alarm of abnormal thickness of a PVC pipeline. Background Art

[0002] At present, the operating status of PVC pipes is monitored in real time by deploying mechanical sensors, speed sensors, electrical sensors, tactile sensors, image sensors and temperature sensors, in order to achieve automatic identification and early warning of abnormal pipe thickness.

[0003] One existing technology, when fusing multi-source sensor data, struggles to accurately identify early, subtle anomalies in thickness trends, leading to unreliable alarms. Because the data collected by various sensors is subject to environmental noise, and pipeline thickness changes are affected by factors such as temperature, load, and corrosion, traditional fixed threshold or single-factor judgment mechanisms often lack dynamic adaptability, leading to false alarms or missed alarms, limiting the system's effectiveness.

[0004] The existing technology has the problem of difficulty in accurately identifying early abnormalities in PVC pipe thickness based on multi-type sensor data and generating high-confidence alarms. Summary of the Invention

[0005] The present invention provides a real-time alarm method, system, electronic device and storage medium for abnormal thickness of PVC pipes, so as to realize accurate identification of thickness abnormalities in the early stage and generate high-reliability alarms.

[0006] In a first aspect, in order to solve the above technical problems, the present invention provides a real-time alarm method for abnormal thickness of PVC pipes, comprising:

[0007] Obtain thickness data, temperature parameters, and pressure parameters of PVC pipes and construct multi-dimensional monitoring data including timestamps;

[0008] The multi-dimensional monitoring data is processed by a wavelet transform noise reduction algorithm to obtain purified thickness data, purified temperature data and purified pressure data to form purified monitoring data;

[0009] Performing a multivariate nonlinear regression analysis based on the purification monitoring data to establish a coupling relationship expression between the temperature parameter, the pressure parameter and the thickness change;

[0010] Based on the coupling relationship expression, a sliding window technique is used to calculate the deviation sequence between the thickness prediction value and the measured value, and a deviation trend index reflecting the abnormal change pattern is extracted;

[0011] Perform differential analysis based on the deviation trend indicator to extract the duration and intensity level of the abnormal pattern, and extract a set of business attributes in combination with the purification monitoring data;

[0012] Calculating a credibility value of the alarm event using a Bayesian inference algorithm according to the intensity level, the duration, and the set of service attributes;

[0013] Based on the credibility value, the alarm threshold is dynamically adjusted, and based on the updated alarm threshold, the parameter configuration in the coupling relationship expression is iteratively corrected for continuous detection and adaptive adjustment of PVC pipe thickness anomalies.

[0014] Preferably, the multi-dimensional monitoring data is processed using a wavelet transform noise reduction algorithm to obtain purified thickness data, purified temperature data and purified pressure data to form purified monitoring data, including:

[0015] extracting time series of purification thickness data, purification temperature data, and purification pressure data respectively according to the multi-dimensional monitoring data;

[0016] For the time series, wavelet decomposition is performed using Daubechies4 as the wavelet basis function to obtain high-frequency detail components and low-frequency approximate components;

[0017] Performing energy analysis on the high-frequency detail component to determine whether the energy of the high-frequency detail component exceeds a preset noise energy threshold;

[0018] If the energy of the high-frequency detail component exceeds the noise energy threshold, performing soft threshold denoising on the high-frequency detail component to obtain a denoised high-frequency component;

[0019] Reconstructing the low-frequency approximate component and the noise-reduced high-frequency component to obtain purified thickness data, purified temperature data, and purified pressure data;

[0020] According to the time series of the purification thickness data, the purification temperature data and the purification pressure data, data structure splicing is performed to form purification monitoring data.

[0021] Preferably, performing a multivariate nonlinear regression analysis based on the purification monitoring data to establish a coupling relationship expression between temperature parameters, pressure parameters and thickness change includes:

[0022] extracting a multidimensional time series of thickness, temperature, and pressure from the purification monitoring data, and normalizing the multidimensional time series to obtain standardized data;

[0023] Calculating the mean, variance, and interaction characteristics of the standardized data to construct a feature matrix containing high-order combination terms of temperature and pressure;

[0024] Based on the characteristic matrix, the purification thickness data is used as the target vector, and the least squares fitting regression method is adopted to calculate the regression coefficient;

[0025] Based on the regression coefficient, a mathematical expression is established to describe the coupling relationship between the temperature parameter, the pressure parameter and the thickness change.

[0026] Preferably, the method of calculating the deviation sequence between the thickness prediction value and the measured value based on the coupling relationship expression and adopting the sliding window technology to extract the deviation trend index reflecting the abnormal change pattern includes:

[0027] Based on the coupling relationship expression, the thickness prediction value is calculated in combination with the purification temperature data and the purification pressure data, and compared with the purification thickness data to generate a thickness deviation sequence;

[0028] The thickness deviation sequence is segmented using a sliding window technique, with the window length set to 24 hours, to obtain local deviation sequences of multiple time segments;

[0029] Calculating statistics such as mean, variance and skewness of the local deviation sequence, and extracting a set of trend indicators reflecting the deviation change trend;

[0030] If the mean in the trend indicator set deviates from a preset mean range, or the variance exceeds a preset variance threshold, or the skewness deviates from a normal distribution interval, the trend indicator is determined to be a deviation trend indicator reflecting an abnormal change pattern.

[0031] Preferably, performing differential analysis based on the deviation trend indicator to extract the duration and intensity level of the abnormal pattern, and extracting a set of business attributes in combination with the purification monitoring data, includes:

[0032] The deviation trend indicator is segmented into continuous time windows, and the first-order difference and the second-order difference of the deviation sequence in each window are calculated respectively;

[0033] If the absolute value of the first-order difference exceeds 1.5 times the preset dynamic reference value, and the second-order difference maintains the same direction as the first-order difference, the corresponding window is determined to be in abnormal mode;

[0034] Tracing back to the time point when the differential condition was first met to determine the duration of the abnormal pattern;

[0035] Based on the time period corresponding to the abnormal pattern, the peak characteristics of the first-order difference are extracted, and the intensity level is obtained by using a cluster analysis method;

[0036] Based on the operation time, load level, ambient temperature and pipeline material parameters in the purification monitoring data, combined with the duration and the intensity level, a business attribute set is constructed.

[0037] Preferably, the calculating the credibility value of the alarm event using a Bayesian inference algorithm according to the intensity level, the duration and the service attribute set includes:

[0038] Based on preset historical abnormal event data, set the prior probability of abnormal events;

[0039] Calculating based on the intensity level, the duration, and the service attribute set to obtain the conditional probability of each observation parameter under the condition of anomaly occurrence;

[0040] The prior probability and the conditional probability are substituted into the Bayesian inference formula to calculate the posterior probability of the current alarm event as the alarm credibility value.

[0041] Preferably, the method of dynamically adjusting the alarm threshold based on the credibility value and iteratively correcting the parameter configuration in the coupling relationship expression based on the updated alarm threshold for continuous detection and adaptive adjustment of PVC pipe thickness anomalies includes:

[0042] Based on the credibility value, dynamically adjust the alarm threshold to form an updated alarm judgment standard;

[0043] Acquiring actual monitoring data corresponding to the alarm judgment criteria;

[0044] Substituting the actual monitoring data into the coupling relationship expression to generate an adjusted thickness prediction value;

[0045] Comparing the adjusted thickness prediction value with a preset thickness threshold to obtain a deviation value;

[0046] Calculating a feedback adjustment factor according to the deviation value as an iterative correction of the parameter configuration in the coupling relationship expression;

[0047] The corrected parameter configuration is used for the continuous execution of subsequent PVC pipe thickness anomaly detection, supporting adaptive adjustment of alarm judgment.

[0048] In a second aspect, the present invention provides a real-time alarm system for abnormal thickness of PVC pipes, comprising:

[0049] The data acquisition module is used to obtain the thickness data, temperature parameters and pressure parameters of the PVC pipe and construct multi-dimensional monitoring data including time stamps;

[0050] A data purification module is used to process the multi-dimensional monitoring data using a wavelet transform noise reduction algorithm to obtain purified thickness data, purified temperature data and purified pressure data to form purified monitoring data;

[0051] A coupling relationship module, configured to perform a multivariate nonlinear regression analysis based on the purification monitoring data to establish a coupling relationship expression between temperature parameters, pressure parameters and thickness change;

[0052] A deviation trend module is used to calculate the deviation sequence between the thickness prediction value and the measured value based on the coupling relationship expression using a sliding window technology, and extract a deviation trend indicator reflecting the abnormal change pattern;

[0053] an anomaly extraction module, configured to perform differential analysis based on the deviation trend indicator, extract the duration and intensity level of the anomaly pattern, and extract a set of business attributes in combination with the purification monitoring data;

[0054] an alarm confirmation module, configured to calculate a credibility value of an alarm event using a Bayesian inference algorithm based on the intensity level, the duration, and the set of service attributes;

[0055] The adaptive adjustment module is used to dynamically adjust the alarm threshold based on the credibility value, and iteratively correct the parameter configuration in the coupling relationship expression based on the updated alarm threshold, for continuous detection and adaptive adjustment of PVC pipe thickness anomalies.

[0056] In a third aspect, the present invention also provides an electronic device comprising a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein when the processor executes the computer program, the method for real-time alarm of abnormal PVC pipe thickness as described above is implemented.

[0057] In a fourth aspect, the present invention also provides a computer-readable storage medium, which includes a stored computer program, wherein when the computer program is running, the device where the computer-readable storage medium is located is controlled to execute any one of the above-mentioned real-time alarm methods for abnormal PVC pipe thickness.

[0058] Compared with the prior art, the present invention has the following beneficial effects:

[0059] (1) This invention introduces a wavelet transform algorithm to perform component decomposition and energy analysis on thickness, temperature, and pressure data, effectively filtering out high-frequency noise while preserving thickness variation trends. Multiple nonlinear regression analysis is then used to establish an expression for the coupled relationship between temperature and pressure and thickness variation. This expression not only preserves the nonlinear correlation characteristics between the various physical quantities but also enhances the fitting capability through standardization and high-order interaction terms, providing a reliable basis for subsequent deviation trend determination.

[0060] (2) After using a sliding window to extract deviation trend indicators, the system performs preliminary anomaly identification based on first-order and second-order differential directionality, and traces back to the time point when the anomaly conditions were first met to accurately locate the onset of the anomaly. The first-order differential peak within this time period is then extracted and the intensity level is divided using an unsupervised clustering method. This not only avoids the problem of setting a fixed threshold for anomaly amplitude, but also enhances the ability to distinguish between slowly changing and sudden anomalies.

[0061] (3) The present invention sets a priori probability based on historical abnormal events, and substitutes the intensity level, duration, and business attributes into the Bayesian formula to calculate the credibility, thereby obtaining the posterior probability of the alarm. This credibility value is further used as the basis for dynamically adjusting the alarm threshold. In the feedback phase, the error calculation between the monitoring data and the predicted value is introduced, and the coupling expression parameter configuration is corrected by the feedback adjustment factor. Ultimately, a closed-loop thickness anomaly monitoring system is constructed to achieve dynamic adaptive updating of the alarm standard, thereby improving the accuracy of early alarms and the self-learning ability of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] Figure 1 This is a flow chart of a method for real-time alarming of abnormal PVC pipe thickness provided by the first embodiment of the present invention;

[0063] Figure 2 It is a structural diagram of a real-time alarm system for abnormal PVC pipe thickness provided by the second embodiment of the present invention. DETAILED DESCRIPTION

[0064] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0065] Reference Figure 1 The first embodiment of the present invention provides a real-time alarm method for abnormal thickness of PVC pipes, comprising the following steps:

[0066] S11, obtaining thickness data, temperature parameters, and pressure parameters of the polyvinyl chloride pipe, and constructing multidimensional monitoring data including a timestamp;

[0067] S12, processing the multidimensional monitoring data using a wavelet transform noise reduction algorithm to obtain purified thickness data, purified temperature data, and purified pressure data to form purified monitoring data;

[0068] S13, performing a multivariate nonlinear regression analysis based on the purification monitoring data to establish a coupling relationship expression between the temperature parameter, the pressure parameter and the thickness change;

[0069] S14, based on the coupling relationship expression, using a sliding window technique to calculate a deviation sequence between the thickness prediction value and the measured value, and extracting a deviation trend indicator reflecting an abnormal change pattern;

[0070] S15, performing differential analysis based on the deviation trend indicator to extract the duration and intensity level of the abnormal pattern, and extracting a business attribute set in combination with the purification monitoring data;

[0071] S16, calculating a credibility value of the alarm event using a Bayesian inference algorithm based on the intensity level, the duration, and the service attribute set;

[0072] S17, dynamically adjusting the alarm threshold based on the credibility value, and iteratively correcting the parameter configuration in the coupling relationship expression based on the updated alarm threshold, for continuous detection and adaptive adjustment of PVC pipe thickness anomalies.

[0073] In step S11, it is necessary to obtain the thickness data, temperature parameters, and pressure parameters of the polyvinyl chloride pipe and construct multi-dimensional monitoring data including a timestamp, including:

[0074] First, thickness data is collected. Ultrasonic thickness sensors deployed on or within the pipe surface provide non-contact, periodic measurements of pipe wall thickness. For example, on an industrial pipeline, the ultrasonic sensors are set to measure once an hour. In a 24 / 7 operation scenario, 24 pieces of thickness data can be stably collected daily. The sensors output measurement values ​​in millimeters and include measurement timestamps, forming a preliminary thickness time series.

[0075] In a specific embodiment, temperature and pressure parameters are obtained. In the pipeline operation system, temperature sensors (such as thermocouples or platinum resistance thermometers) and pressure sensors (such as capacitive or strain gauge pressure transmitters) are equipped to record temperature changes and fluid pressure fluctuations inside or outside the pipeline, respectively. Taking a typical PVC delivery pipeline as an example, the temperature range is set to 0-80°C and the pressure range is set to 0-1.6MPa. The system collects relevant data every 10 minutes and records the corresponding timestamp to ensure the timeliness and continuity of temperature and pressure information.

[0076] In a specific embodiment, the system first synchronizes the data collected by multiple types of sensors (including thickness, temperature, and pressure). Because the sampling period for thickness data is one hour, while the sampling period for temperature and pressure data is 10 minutes, there is an inconsistency in the time granularity between the two. To solve this problem, the system uses the thickness sampling time point as the reference node of the unified time axis. Within 30 minutes before and after each thickness sampling time point (such as 08:00, 09:00, etc.), the system extracts the temperature and pressure data within the corresponding time window (containing a total of 6 data points) and calculates the average temperature and pressure values ​​of this time period by arithmetic averaging.

[0077] For example, if the system records a thickness of 8.32 mm at 8:00 AM on June 11, 2025, the temperature data (e.g., 7:30, 7:40, ..., 8:00) and pressure data within half an hour before and after that time point are averaged to obtain an aggregated temperature value of 64.5°C and a pressure value of 0.85 MPa. Ultimately, a set of structured monitoring data is generated at that time point: [8:00 AM on June 11, 2025, 8.32 mm, 64.5°C, 0.85 MPa]. This method retains the trend characteristics of high-frequency temperature and pressure data through window aggregation based on the main parameter (thickness).

[0078] In a specific embodiment, a multi-dimensional monitoring data structure with a unified format is constructed. The system organizes the aligned data into a four-tuple structure containing a timestamp, thickness value, temperature value, and pressure value, i.e., a data matrix of the form [timestamp, thickness, temperature, pressure]. For example, the continuous records of a certain section of pipeline on a certain day include:

[0079] [2025-06-1108:00,8.32mm,64.5℃,0.85MPa]、

[0080] [2025-06-1109:00,8.30mm,65.1℃,0.87MPa]、……

[0081] And so on to build a complete data set.

[0082] In step S12, the multi-dimensional monitoring data needs to be processed using a wavelet transform noise reduction algorithm to obtain purified thickness data, purified temperature data, and purified pressure data to form purified monitoring data, including:

[0083] extracting time series of purification thickness data, purification temperature data, and purification pressure data respectively according to the multi-dimensional monitoring data;

[0084] For the time series, wavelet decomposition is performed using Daubechies4 as the wavelet basis function to obtain high-frequency detail components and low-frequency approximate components;

[0085] Performing energy analysis on the high-frequency detail component to determine whether the energy of the high-frequency detail component exceeds a preset noise energy threshold;

[0086] If the energy of the high-frequency detail component exceeds the noise energy threshold, performing soft threshold denoising on the high-frequency detail component to obtain a denoised high-frequency component;

[0087] Reconstructing the low-frequency approximate component and the noise-reduced high-frequency component to obtain purified thickness data, purified temperature data, and purified pressure data;

[0088] According to the time series of the purification thickness data, the purification temperature data and the purification pressure data, data structure splicing is performed to form purification monitoring data.

[0089] First, extract the time series of each physical quantity. Based on the unified format data structure generated in step S11, the system can extract the original time series data of three types of physical quantities, namely the thickness time series {T1, T2, ..., Tn}, the temperature time series {θ1, θ2, ..., θn}, and the pressure time series {P1, P2, ..., Pn}, where each sampling point is accompanied by a corresponding timestamp. Assuming that sampling is performed at hourly intervals and a total of 7 days of data are collected, the length of each type of sequence is 168. The extraction process is completed using the DataFrame['column_name'] operation of the pandas library in Python, and the time index is uniformly aligned to ensure consistency in subsequent processing.

[0090] Secondly, wavelet decomposition is performed. For each time series of physical quantities, one-dimensional discrete wavelet decomposition is performed using the Daubechies4 wavelet basis function. Due to its good tight support characteristics and smoothness, the db4 wavelet is suitable for processing engineering monitoring data with mutation characteristics. In this embodiment, a three-layer wavelet decomposition is performed on the sequence, that is, the original signal is decomposed into three layers of high-frequency detail components D1, D2, D3 and one layer of low-frequency approximate component A3. This operation is implemented using the pywt.wavedec() function of the PyWavelets library in Python.

[0091] Next, the system performs energy analysis and soft threshold noise reduction on each layer of high-frequency components obtained by decomposition. For each layer of high-frequency detail components, the system calculates its energy value by summing the squares of each coefficient. The specific calculation formula is:

[0092]

[0093] Among them, (D i,j ) represents the jth wavelet coefficient in the detail component of the i-th layer, and m is the number of coefficients in that layer. To determine whether there is significant noise interference, the system further calculates the overall energy upper limit (noise energy threshold), which is set to 10% of the sum of the total energy of the high-frequency component and the energy of the low-frequency approximate component. This threshold is calculated using the following formula:

[0094] Where α = 0.1, EA3 represents the energy of the third-layer low-frequency approximation component A3, which is the square of the second norm of the signal vector and is mathematically defined as:

[0095]

[0096] Where A3={A 3,1 ,A 3,2 ,…,A 3,n} is the low-frequency approximation coefficient of the third layer in wavelet decomposition; A 3,i is the i-th value in the coefficient;

[0097] If the energy E of a high-frequency component i If the value exceeds the corresponding threshold, it is considered to contain noise interference and should be denoised. The denoising method uses a soft threshold algorithm, and its specific transformation formula is:

[0098]

[0099] Here, λ is a soft threshold, set to 1.5 times the standard deviation of the component in that layer. In actual implementation, the threshold is dynamically generated by calculating the standard deviation using NumPy's np.std() function.

[0100] In one specific embodiment, after denoising the high-frequency components, the system also implements a screening mechanism for the low-frequency approximate components to ensure the authenticity of the signal's trend components. First, the system extracts the fifth-level low-frequency approximate component based on the wavelet decomposition results. This component primarily reflects the long-term trend of thickness, temperature, or pressure signals. To determine whether this low-frequency component is contaminated with broadband low-frequency noise, the system evaluates its overall energy level and stability of variation.

[0101] Specifically, the system first detects the energy ratio of the low-frequency component in the overall signal. If the ratio is significantly higher, it may indicate the presence of abnormal low-frequency drift components. Subsequently, the system further analyzes the stability of the change of the component, such as determining whether the amplitude of change between adjacent sampling points fluctuates violently. If the degree of fluctuation exceeds the preset empirical threshold (such as the temperature change rate exceeds 0.05°C / h, or the thickness change rate exceeds 0.01mm / h), the system will smooth the low-frequency component using methods such as three-point sliding average or weighted average to eliminate pseudo-changes or background disturbances in the trend. If the low-frequency component changes smoothly, the system directly retains its original value.

[0102] In a specific embodiment, signal reconstruction is performed by combining the retained low-frequency approximation component with all the high-frequency components that have been denoised. Together, the inverse wavelet transform pywt.waverec() is used to reconstruct the signal, thereby obtaining the purified thickness, temperature, and pressure time series. The reconstructed signal is smoother than the original sequence, eliminating the sudden spikes and random jitter during the acquisition process, while retaining the low-frequency trend component to ensure the physical authenticity of the signal. Finally, the three types of purified data are spliced ​​to construct the purification monitoring data structure. The system merges the above three groups of purified time series into a unified structure data frame with timestamp as index, in the following format:

[0103] [timestamp,thickness_clean,temperature_clean,pressure_clean]

[0104] For example, a sampling point is recorded as [2025-06-12 14:00, 8.28 mm, 66.1°C, 0.91 MPa]. This structure is saved in a standardized CSV file or database table format and can be directly called by subsequent modeling and analysis modules.

[0105] In step S13, it is necessary to perform a multivariate nonlinear regression analysis based on the purification monitoring data to establish a coupling relationship expression between the temperature parameter, the pressure parameter and the thickness change, including:

[0106] extracting a multidimensional time series of thickness, temperature, and pressure from the purification monitoring data, and normalizing the multidimensional time series to obtain standardized data;

[0107] Calculating the mean, variance, and interaction characteristics of the standardized data to construct a feature matrix containing high-order combination terms of temperature and pressure;

[0108] Based on the characteristic matrix, the purification thickness data is used as the target vector, and the least squares fitting regression method is adopted to calculate the regression coefficient;

[0109] Based on the regression coefficient, a mathematical expression is established to describe the coupling relationship between the temperature parameter, the pressure parameter and the thickness change.

[0110] First, the system extracts multidimensional time series of thickness, temperature, and pressure from the purification monitoring data. Assuming that the monitoring time range is 60 consecutive days and sampling is performed in units of 1 hour per day, each type of data contains a total of N = 60 × 24 = 1440 sampling points. The system constructs the thickness time series H = {h1, h2, ..., hN}, the temperature time series T = {t1, t2, ..., tN}, and the pressure time series P = {p1, p2, ..., pN}, and normalizes them. Normalization uses the min-max method to convert each dimensional signal x i Mapping to the interval [0,1]:

[0111] where x min and x max are the minimum and maximum values ​​of the parameter sequence, respectively. This step ensures that data of different dimensions can be uniformly entered into the regression framework to avoid the influence of scale differences on the fitting results.

[0112] In a specific embodiment, statistical feature extraction is performed on the standardized time series. The system calculates the mean μ for temperature and pressure respectively. T ,μ P ,variance And construct interaction characteristic terms to express nonlinear coupling effects, for example:

[0113] First-order interaction term t i ·p i , used to express the linear coupling of temperature and pressure;

[0114] Second-order terms Used to describe the respective nonlinear effects;

[0115] These features form the input matrix for the regression analysis Each row corresponds to the data at a time point, each column represents a characteristic variable (such as temperature, pressure, interaction term, etc.), and the value of M is between 5 and 10, which is adjusted according to the final model selection.

[0116] In a specific embodiment, the clean thickness data is used as the target vector Construct the regression problem Y = Xβ + ε, where is the regression coefficient to be determined, and ε is the residual vector. The system uses the least squares method for fitting, and the solution process is:

[0117]

[0118] Specifically, the coefficient vector reflects the sensitivity of each characteristic item to thickness changes. For example, if β3 corresponds to t i ·p i , and its value is negative, it can be assumed that a simultaneous increase in temperature and pressure will weaken the thickness. This process can be performed using the regress function in MATLAB or numpy.linalg.lstsq in Python, which is computationally stable and convergent.

[0119] Specifically, the system is based on the obtained regression coefficients A mathematical expression between thickness change and environmental parameters was established. The final coupling relationship selected was:

[0120]

[0121] Among them, β0 is the intercept term, which represents the thickness reference value when both temperature and pressure are normalized to zero; β1 and β2 are first-order coefficients, representing the individual effects of temperature and pressure; β3 and β4 are second-order coefficients, describing their nonlinear modulation; β5 is the interaction term, reflecting the coupling relationship between the two; ΔT = t i -T ref Indicates the current temperature relative to the reference temperature (using T ref =298K or 25℃); ΔP = p i -P ref Indicates the difference between the current pressure and the reference pressure (0.1MPa); is the normalized thickness prediction value (unit is dimensionless).

[0122] In step S14, based on the coupling relationship expression, a sliding window technique is used to calculate the deviation sequence between the thickness prediction value and the measured value, and to extract the deviation trend index reflecting the abnormal change pattern, including:

[0123] Based on the coupling relationship expression, the thickness prediction value is calculated in combination with the purification temperature data and the purification pressure data, and compared with the purification thickness data to generate a thickness deviation sequence;

[0124] The thickness deviation sequence is segmented using a sliding window technique, with the window length set to 24 hours, to obtain local deviation sequences of multiple time segments;

[0125] Calculating statistics such as mean, variance and skewness of the local deviation sequence, and extracting a set of trend indicators reflecting the deviation change trend;

[0126] If the mean in the trend indicator set deviates from a preset mean range, or the variance exceeds a preset variance threshold, or the skewness deviates from a normal distribution interval, the trend indicator is determined to be a deviation trend indicator reflecting an abnormal change pattern.

[0127] First, the system calculates the thickness prediction value based on the coupling relationship expression.

[0128] where t i With p i are the purification temperature and purification pressure data at the i-th time point, The system traverses the purified temperature and pressure data point by point and calculates the thickness prediction value at each time point by substituting the coupling expression into the equation. And compare it with the measured purification thickness value Make the difference to form a thickness deviation sequence:

[0129]

[0130] in, Indicates that the actual thickness will be measured Normalized value.

[0131] Get the deviation sequence Δ={δ1,δ2,...,δ N}, used for subsequent statistical analysis.

[0132] In a specific embodiment, the system uses a sliding window technique to segment the deviation sequence. The window length is set to 24 hours. If the data sampling period is 1 hour, each window contains 24 data points. The sliding step can be set to 1 hour or 6 hours to balance the detection sensitivity and computational cost. For example, if the deviation sequence length is 1440, the system will divide it into local deviation subsequences, each subsequence is denoted as Δ j ={δ j,1 ,...,δ j,24}.

[0133] Specifically, in each local subsequence, the system calculates the following three types of statistical indicators:

[0134] mean Reflects the overall deviation of the forecast deviation within the window;

[0135] variance Characterizes the intensity of the fluctuation of the deviation;

[0136] Skewness Used to describe the asymmetry of a skewed distribution.

[0137] The system records the above statistical indicator set as trend indicator vector Used to determine whether there is an abnormal trend in the time segment.

[0138] In a specific embodiment, the system determines whether an abnormal change pattern exists based on preset criteria.

[0139] The specific standards are:

[0140] If μ j The absolute value exceeds the mean threshold of the historical normal operating range (e.g. set to ±0.02mm);

[0141] or σ j Exceeding the global variance threshold (e.g. 0.005mm 2 );

[0142] or γ j The absolute value of is greater than the set upper limit (e.g., a deviation from 0.5 indicates severe skewness);

[0143] Then the trend indicator S of this window j It is determined to be an “abnormal deviation trend” and serves as a candidate for subsequent differential analysis and strength determination.

[0144] During the implementation of the present invention, in order to ensure that the determination of deviation trend indicators has engineering practicality and robustness, the system sets clear judgment thresholds for statistical quantities such as mean, variance and skewness, and makes reasonable settings and dynamic corrections through historical data analysis and field verification.

[0145] Specifically, during the sliding window statistical processing of the thickness deviation series, the mean threshold was set to ±0.02mm. Based on the empirical statistics of thickness prediction deviations during historical normal operation, the deviation between the predicted and measured thickness should be kept within ±0.01mm. Taking into account sensor errors and operating condition disturbances, the system selected a deviation value of 0.02mm as the critical point. When the mean deviation within a certain window exceeds this range, it is determined that a systematic offset exists, reflecting a persistent abnormality in thickness variation.

[0146] For the variance threshold, the system is set to 0.004mm 2 , derived from statistical analysis of no less than 10,000 historical monitoring data. In the absence of abnormalities, the variance of the deviation sequence is stable at 0.001 to 0.003 mm 2 Between, select 0.004mm 2 As an upper limit, this increases sensitivity to sudden or fluctuating thickness changes without introducing excessive false alarms. The system also supports an adaptive learning mechanism that dynamically fine-tunes the variance threshold based on the stability of deviations over the past 30 days, further enhancing its adaptability to different operating scenarios.

[0147] For the skewness threshold, the system is set to ±0.5. Skewness is used to describe the asymmetry of the deviation distribution. Under normal circumstances, the deviation should obey an approximately symmetrical distribution, and the skewness fluctuates within ±0.3. When the skewness of a certain window is greater than +0.5 or less than -0.5, the system believes that there is an asymmetric distribution trend, reflecting a certain type of hidden trend anomaly (such as long-term slow corrosion or local desorption). In engineering implementation, the system simulates the thickness response under different temperature and pressure fluctuations and finds that the deviation sequence shows skewness amplification under abnormal conditions. Therefore, using ±0.5 as the standard threshold has better resolution.

[0148] For example, in the actual deployment of a chemical plant, the pipeline is in a constant temperature and pressure state at the initial stage of operation, with the mean deviation kept within 0.008mm and the variance within 0.002mm. 2 , with a skewness of approximately 0.12. As pressure fluctuations intensify, the mean value in some time windows increases to 0.028 mm and the variance increases to 0.006 mm. 2 , the skewness reaches 0.66. Based on this, the system determines that a thickness change trend deviation has occurred and accurately triggers the subsequent abnormal analysis process.

[0149] For example, during a certain monitoring period, the system found that the mean of the deviation sequence in the 132nd sliding window (corresponding to the time period from the 11th to the 12th day) was 0.03mm and the variance was 0.006mm. 2 , and the skewness is 0.68, both exceeding the above preset threshold range. Based on this, the system marks the trend indicator as an abnormal deviation trend and records the duration of the window for use in the next stage.

[0150] In step S15, it is necessary to perform differential analysis based on the deviation trend indicator to extract the duration and intensity level of the abnormal pattern, and extract a set of business attributes in combination with the purification monitoring data, including:

[0151] The deviation trend indicator is segmented into continuous time windows, and the first-order difference and the second-order difference of the deviation sequence in each window are calculated respectively;

[0152] If the absolute value of the first-order difference exceeds 1.5 times the preset dynamic reference value, and the second-order difference maintains the same direction as the first-order difference, the corresponding window is determined to be in abnormal mode;

[0153] Tracing back to the time point when the differential condition was first met to determine the duration of the abnormal pattern;

[0154] Based on the time period corresponding to the abnormal pattern, the peak characteristics of the first-order difference are extracted, and the intensity level is obtained by using a cluster analysis method;

[0155] Based on the operation time, load level, ambient temperature and pipeline material parameters in the purification monitoring data, combined with the duration and the intensity level, a business attribute set is constructed.

[0156] First, the system divides the deviation trend indicators obtained in the previous step into multiple continuous time windows in chronological order. The length of each window is recommended to be set between 12 hours and 24 hours, and can be flexibly adjusted according to the system monitoring cycle and the historical anomaly evolution speed. In each window, the system calculates the first-order difference (Δ1 = x t -x t-1 ) and second-order difference (Δ2=Δ 1t -Δ 1t-1 ), used to identify the slope and curvature characteristics of data changes.

[0157] The system presets a dynamic baseline value, taking the mean and standard deviation of the thickness deviation series over the past seven days (e.g., a mean deviation of 0.012mm and a standard deviation of 0.004mm), and calculates the dynamic baseline value as μ + σ ≈ 0.016mm. If the absolute value of the first-order difference, |Δ1|, exceeds 1.5 times this baseline value (i.e., > 0.024mm), and the corresponding second-order difference, Δ2, has the same sign as Δ1 (i.e., continuously increasing or decreasing), then the time window is considered to have experienced an accelerating deviation evolution, and is marked as an abnormal mode window.

[0158] In one specific embodiment, the system backtracks all time windows that meet the aforementioned differential determination criteria, extracts the duration point where the criteria were first met, and marks it as the duration of the abnormal trend. For example, if the aforementioned acceleration deviation feature appears in three consecutive windows starting at 4:00 PM on August 1, 2024, the system will record the starting point as "2024-08-01 16:00," which serves as an important reference for alarm assessment and subsequent policy adjustments.

[0159] In a specific embodiment, within the identified abnormal time period, the system further extracts the peak change of the first-order difference sequence in the period (such as the maximum Δ1 is 0.065mm) to reflect the abnormal intensity. The system uses the K-means clustering algorithm to classify the peak features of multiple abnormal segments. For example, setting K = 3 corresponds to three intensity levels of "weak abnormality", "medium abnormality" and "strong abnormality". This classification sets the initial center point of the cluster based on the distribution of historical data, such as weak abnormality <0.03mm, moderate abnormality between 0.03 and 0.06mm, and strong abnormality >0.06mm, to enhance the quantification of the severity of the abnormality and the actual response guidance.

[0160] In a specific embodiment, the system extracts business parameters related to the operating status and environmental conditions from the purification monitoring data corresponding to the abnormal section, including: ① operating time (e.g., continuous operation for more than 600 hours), ② load level (e.g., pump load for more than 90%), ③ ambient temperature (e.g., more than 50°C), and ④ pipe material parameters (e.g., PVC-U pipe, thermal expansion coefficient of 1.4×10 -4 These parameters are automatically extracted through the sensor system and the process database, and are synchronized and matched based on the timestamps of the abnormal segments.

[0161] Finally, the system combines the duration, intensity level, and multi-dimensional service parameters into a standardized service attribute set, which can be expressed in the form of a structure as follows:

[0162] {"start_time":"2024-08-0116:00",

[0163] "severity_level":"strong",

[0164] "runtime_hours":612,

[0165] "load_level":92.3,

[0166] "ambient_temp":51.2,

[0167] "material_type":"PVC-U"}

[0168] This set serves as one of the inputs for subsequent alarm credibility assessment and dynamic threshold adjustment algorithms, ensuring that the system has the depth and breadth of anomaly recognition under multi-parameter coupling conditions, thereby improving the overall intelligent early warning level.

[0169] In step S16, a Bayesian inference algorithm is used to calculate the credibility value of the alarm event based on the intensity level, the duration, and the service attribute set, including:

[0170] Based on preset historical abnormal event data, set the prior probability of abnormal events;

[0171] Calculating based on the intensity level, the duration, and the service attribute set to obtain the conditional probability of each observation parameter under the condition of anomaly occurrence;

[0172] The prior probability and the conditional probability are substituted into the Bayesian inference formula to calculate the posterior probability of the current alarm event as the alarm credibility value.

[0173] In step S16, the system uses a Bayesian inference algorithm to calculate the credibility of the alarm event based on multi-source observation data and historical anomaly samples, which serves as the decision-making basis for subsequent alarm judgment and adaptive adjustment. Specifically, the prior probability P(A) of the abnormal event must first be set. This probability is obtained by counting the proportion of confirmed real abnormal events in the total alarm samples. For example, when there are 5,000 alarm events recorded in history, of which 1,250 are real anomalies, the prior probability is P(A) = 1,250 / 5,000 = 0.25. The system then extracts a set of business attributes consisting of intensity level, abnormal duration, operating time, load level, ambient temperature, and pipeline material type from the current observation sample to form a feature vector X = {x1, x2, ..., xn}. In the construction of conditional probabilities, the system adopts the naive Bayesian hypothesis, which assumes that each observation feature is independent of each other under abnormal conditions. By analyzing the frequency of occurrence of each feature value in historical real abnormal samples, each conditional probability P(xi|A) is calculated, such as: the probability of medium-intensity alarms occurring in abnormal events, the proportion of long-term anomalies in real events, etc.

[0174] In a specific embodiment, after completing the conditional probability construction, the system substitutes the prior probability and conditional probability into the Bayesian formula to derive the posterior probability P(A|X) that the current alarm event is a true anomaly. To facilitate numerical calculations and logical judgments, the present invention defines the alarm credibility value C as the normalized form of the posterior probability, that is:

[0175]

[0176] in, represents the prior probability of non-anomalous events, The probability distribution of feature vectors under non-abnormal conditions is modeled using historical false alarm samples. For a specific sample, if the joint probability of the feature combination of the current alarm event is 0.246 under abnormal conditions and 0.0301 under non-abnormal conditions, the confidence value calculated based on the prior probability is approximately 0.731, indicating that the event is a true anomaly.

[0177] In one specific embodiment, the system sets a credibility threshold, θ, such as 0.6. If the calculated C ≥ θ, the alarm event is considered high-credibility, triggering subsequent feedback control and model parameter updates. Otherwise, it is recorded as a low-credibility event, and no control response is executed. Through these steps, the system achieves quantitative identification of abnormal alarms, effectively improving the real-time and accuracy of PVC pipe thickness anomaly detection.

[0178] It should be noted that to ensure sufficient engineering applicability and stability in alarm credibility assessments, the credibility threshold (denoted as θ) used in the system must be appropriately set through backtesting of actual operational data and analysis of false positive rates. Specifically, the credibility threshold determines whether the currently calculated posterior probability C = P(A|X) is sufficient to support a high-confidence alarm assessment.

[0179] In a practical implementation, system developers used a three-month sample of on-site pipe network monitoring data (approximately 8,000 alarm records) to label true anomalies and false alarms, and based on this, constructed a distribution curve for alarm credibility values. Statistical analysis showed that when the posterior probability C ≥ 0.6, the true hit rate of alarm events reached 85%, and the false alarm rate was controlled below 9%. In the range of C < 0.6, the false alarm rate exceeded 40%. Therefore, the system initially set θ = 0.6 as the credibility judgment threshold for subsequent online alarm triggering.

[0180] Furthermore, the threshold supports dynamic configuration to account for differences in tolerance for false positives across different industries or operating conditions. For example, in a chemical plant scenario, where personal safety is paramount, the system can raise the threshold to θ = 0.7 to reduce false positives. In a low-risk environment, to detect potential anomalies earlier, the threshold can be lowered to θ = 0.5, maintaining a balance between sensitivity and stability.

[0181] In step S17, it is necessary to dynamically adjust the alarm threshold based on the credibility value, and iteratively correct the parameter configuration in the coupling relationship expression based on the updated alarm threshold for continuous detection and adaptive adjustment of PVC pipe thickness anomalies, including:

[0182] Based on the credibility value, dynamically adjust the alarm threshold to form an updated alarm judgment standard;

[0183] Acquiring actual monitoring data corresponding to the alarm judgment criteria;

[0184] Substituting the actual monitoring data into the coupling relationship expression to generate an adjusted thickness prediction value;

[0185] Comparing the adjusted thickness prediction value with a preset thickness threshold to obtain a deviation value;

[0186] Calculating a feedback adjustment factor according to the deviation value as an iterative correction of the parameter configuration in the coupling relationship expression;

[0187] The corrected parameter configuration is used for the continuous execution of subsequent PVC pipe thickness anomaly detection, supporting adaptive adjustment of alarm judgment.

[0188] First, the system dynamically adjusts the threshold based on the current alarm credibility value C∈[0,1]. When the credibility value is close to the previously set alarm threshold boundary (such as θ=0.6) and the historical false positive rate is high, the system performs adaptive updates in the following form:

[0189] θ′=θ+η(C-θ′)

[0190] Here, η∈(0,1) is the learning rate, which controls the threshold update speed and is set to 0.1-0.3. For example, when θ=0.6, C=0.7, and η=0.2, the updated threshold is θ'=0.62. This process avoids threshold rigidity and improves adaptability to fluctuations in alarm credibility.

[0191] In a specific embodiment, the system extracts the actual monitoring data within the corresponding time period according to the updated alarm judgment standard θ' Where T i ,P i The purified temperature and pressure data are substituted into the existing coupling expression to calculate the thickness prediction value under the current parameter configuration. The predicted value is compared with the actual thickness measurement value to calculate the deviation value:

[0192]

[0193] The deviation values ​​are accumulated in multiple time windows and the mean square error (MSE) is calculated as the global deviation indicator:

[0194]

[0195] In a specific embodiment, the system constructs a feedback adjustment factor to correct the parameter configuration in the coupling expression. The adjustment factor is constructed based on the gradient descent method:

[0196]

[0197] in, The value of the current j-th parameter (or weight) in the t+1-th iteration; is the value of the jth parameter after the tth iteration update; α is the iteration step size (set to 0.01~0.1), is the partial derivative of the loss function with respect to each parameter. This mechanism ensures that the system can quickly fine-tune parameters based on prediction errors.

[0198] In one specific implementation, the system applies the optimized parameter configuration to the next round of PVC pipe thickness anomaly detection. With the continuous input of new data and the updating of credibility, the entire monitoring process forms a closed-loop mechanism of "credibility → threshold adjustment → parameter optimization → prediction correction." This ensures that the alarm judgment criteria align with actual operating conditions, enabling adaptive, dynamic tracking, and response to thickness anomaly trends. This mechanism is particularly suitable for industrial sites with frequent environmental changes or complex operating conditions, enhancing the intelligence level and engineering reliability of the alarm system.

[0199] In summary, the present invention provides a real-time alarm method, system, electronic device and storage medium for PVC pipe thickness anomaly, so as to achieve accurate identification of early thickness anomalies and generate high-reliability alarms.

[0200] Reference Figure 2 The second embodiment of the present invention provides a real-time alarm system for abnormal thickness of PVC pipes, comprising:

[0201] The data acquisition module is used to obtain the thickness data, temperature parameters and pressure parameters of the PVC pipe and construct multi-dimensional monitoring data including time stamps;

[0202] A data purification module is used to process the multi-dimensional monitoring data using a wavelet transform noise reduction algorithm to obtain purified thickness data, purified temperature data and purified pressure data to form purified monitoring data;

[0203] A coupling relationship module, configured to perform a multivariate nonlinear regression analysis based on the purification monitoring data to establish a coupling relationship expression between temperature parameters, pressure parameters and thickness change;

[0204] A deviation trend module is used to calculate the deviation sequence between the thickness prediction value and the measured value based on the coupling relationship expression using a sliding window technology, and extract a deviation trend indicator reflecting the abnormal change pattern;

[0205] an anomaly extraction module, configured to perform differential analysis based on the deviation trend indicator, extract the duration and intensity level of the anomaly pattern, and extract a set of business attributes in combination with the purification monitoring data;

[0206] an alarm confirmation module, configured to calculate a credibility value of an alarm event using a Bayesian inference algorithm based on the intensity level, the duration, and the set of service attributes;

[0207] The adaptive adjustment module is used to dynamically adjust the alarm threshold based on the credibility value, and iteratively correct the parameter configuration in the coupling relationship expression based on the updated alarm threshold, for continuous detection and adaptive adjustment of PVC pipe thickness anomalies.

[0208] It should be noted that the PVC pipe thickness abnormality real-time alarm system provided in an embodiment of the present invention is used to execute all the process steps of the PVC pipe thickness abnormality real-time alarm method in the above embodiment. The working principles and beneficial effects of the two correspond one to one, so they will not be repeated here.

[0209] An embodiment of the present invention further provides an electronic device. The electronic device includes: a processor, a memory, and a computer program stored in the memory and executable on the processor, such as a data purification program. When the processor executes the computer program, the steps in the above-mentioned embodiments of the method for real-time alarm of abnormal PVC pipe thickness are implemented, such as Figure 1 Alternatively, when the processor executes the computer program, the functions of the modules / units in the above-mentioned device embodiments are realized, such as the data purification module.

[0210] Exemplarily, the computer program may be divided into one or more modules / units, which are stored in the memory and executed by the processor to implement the present invention. The one or more modules / units may be a series of computer program instruction segments capable of implementing specific functions, and the instruction segments are used to describe the execution process of the computer program in the electronic device.

[0211] The electronic device may be a computing device such as a desktop computer, notebook, PDA, or smart tablet. The electronic device may include, but is not limited to, a processor and memory. Those skilled in the art will appreciate that the aforementioned components are merely examples of electronic devices and do not constitute a limitation of the electronic device. The electronic device may include more or fewer components than those described above, or a combination of certain components, or different components. For example, the electronic device may also include input / output devices, network access devices, buses, and the like.

[0212] The processor may be a central processing unit (CPU), other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor, etc. The processor is the control center of the electronic device, connecting various parts of the entire electronic device using various interfaces and lines.

[0213] The memory can be used to store the computer programs and / or modules, and the processor realizes various functions of the electronic device by running or executing the computer programs and / or modules stored in the memory and calling the data stored in the memory. The memory can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system, at least one application required for a function (such as a sound playback function, an image playback function, etc.); the data storage area can store data created based on the use of the mobile phone (such as audio data, a phone book, etc.). In addition, the memory can include a high-speed random access memory and can also include a non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), at least one disk storage device, a flash memory device, or other volatile solid-state storage device.

[0214] Wherein, if the module / unit integrated in the electronic device is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the present invention implements all or part of the process in the above-mentioned embodiment method, and can also be completed by a computer program to instruct the relevant hardware. The computer program can be stored in a computer-readable storage medium. When the computer program is executed by the processor, it can implement the steps of each of the above-mentioned method embodiments. Wherein, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording medium, USB flash drive, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electric carrier signal, telecommunication signal and software distribution medium. It should be noted that the content contained in the computer-readable medium can be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, computer-readable media do not include electric carrier signals and telecommunication signals.

[0215] It should be noted that the device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed across multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the present embodiment. In addition, in the drawings of the device embodiments provided by the present invention, the connection relationship between the modules indicates that there is a communication connection between them, which may be specifically implemented as one or more communication buses or signal lines. A person of ordinary skill in the art can understand and implement the present invention without inventive effort.

[0216] The specific embodiments described above further illustrate the objectives, technical solutions, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the scope of protection of the present invention. In particular, it should be noted that any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included within the scope of protection of the present invention for those skilled in the art.

Claims

1. A real-time alarm method for abnormal thickness of PVC pipes, characterized in that include: Obtain thickness data, temperature parameters, and pressure parameters of PVC pipes and construct multi-dimensional monitoring data including timestamps; The multi-dimensional monitoring data is processed by a wavelet transform noise reduction algorithm to obtain purified thickness data, purified temperature data and purified pressure data to form purified monitoring data; Performing a multivariate nonlinear regression analysis based on the purification monitoring data to establish a coupling relationship expression between the temperature parameter, the pressure parameter and the thickness change; Based on the coupling relationship expression, a sliding window technique is used to calculate the deviation sequence between the thickness prediction value and the measured value, and a deviation trend index reflecting the abnormal change pattern is extracted; Perform differential analysis based on the deviation trend indicator to extract the duration and intensity level of the abnormal pattern, and extract a set of business attributes in combination with the purification monitoring data; Calculating a credibility value of the alarm event using a Bayesian inference algorithm according to the intensity level, the duration, and the set of service attributes; Based on the credibility value, the preset alarm threshold is dynamically adjusted, and based on the updated alarm threshold, the parameter configuration in the coupling relationship expression is iteratively corrected for continuous detection and adaptive adjustment of PVC pipe thickness anomalies.

2. The PVC pipe thickness abnormality real-time alarm method according to claim 1 is characterized in that: The multi-dimensional monitoring data is processed using a wavelet transform noise reduction algorithm to obtain purified thickness data, purified temperature data, and purified pressure data to form purified monitoring data, including: extracting time series of thickness data, temperature data and pressure data respectively according to the multi-dimensional monitoring data; For the time series, wavelet decomposition is performed using Daubechies4 as the wavelet basis function to obtain high-frequency detail components and low-frequency approximate components; Performing energy analysis on the high-frequency detail component to determine whether the energy of the high-frequency detail component exceeds a preset noise energy threshold; If the energy of the high-frequency detail component exceeds the noise energy threshold, performing soft threshold denoising on the high-frequency detail component to obtain a denoised high-frequency component; Reconstructing the low-frequency approximate component and the noise-reduced high-frequency component to obtain purified thickness data, purified temperature data, and purified pressure data; According to the time series of the purification thickness data, the purification temperature data and the purification pressure data, data structure splicing is performed to form purification monitoring data.

3. The method for real-time alarm of abnormal PVC pipe thickness according to claim 1 is characterized in that: The method of performing a multivariate nonlinear regression analysis based on the purification monitoring data to establish a coupling relationship expression between the temperature parameter, the pressure parameter and the thickness change includes: extracting a multidimensional time series of thickness, temperature, and pressure from the purification monitoring data, and normalizing the multidimensional time series to obtain standardized data; Calculating the mean, variance, and interaction characteristics of the standardized data to construct a feature matrix containing high-order combination terms of temperature and pressure; Based on the characteristic matrix, the purification thickness data is used as the target vector, and the least squares fitting regression method is adopted to calculate the regression coefficient; Based on the regression coefficient, a mathematical expression is established to describe the coupling relationship between the temperature parameter, the pressure parameter and the thickness change.

4. The method for real-time alarm of abnormal PVC pipe thickness according to claim 1, characterized in that: Based on the coupling relationship expression, the sliding window technology is used to calculate the deviation sequence between the thickness prediction value and the measured value, and the deviation trend index reflecting the abnormal change pattern is extracted, including: Based on the coupling relationship expression, the thickness prediction value is calculated in combination with the purification temperature data and the purification pressure data, and compared with the purification thickness data to generate a thickness deviation sequence; The thickness deviation sequence is segmented using a sliding window technique, with the window length set to 24 hours, to obtain local deviation sequences of multiple time segments; Calculating statistics such as mean, variance and skewness of the local deviation sequence, and extracting a set of trend indicators reflecting the deviation change trend; If the mean in the trend indicator set deviates from a preset mean range, or the variance exceeds a preset variance threshold, or the skewness deviates from a normal distribution interval, the trend indicator is determined to be a deviation trend indicator reflecting an abnormal change pattern.

5. The method for real-time alarm of abnormal PVC pipe thickness according to claim 1 is characterized in that: The differential analysis is performed based on the deviation trend indicator to extract the duration and intensity level of the abnormal pattern, and the business attribute set is extracted in combination with the purification monitoring data, including: The deviation trend indicator is segmented into continuous time windows, and the first-order difference and the second-order difference of the deviation sequence in each window are calculated respectively; If the absolute value of the first-order difference exceeds 1.5 times the preset dynamic reference value, and the second-order difference maintains the same direction as the first-order difference, the corresponding window is determined to be in abnormal mode; Tracing back to the time point when the differential condition was first met to determine the duration of the abnormal pattern; Based on the time period corresponding to the abnormal pattern, the peak characteristics of the first-order difference are extracted, and the intensity level is obtained by using a cluster analysis method; Based on the operation time, load level, ambient temperature and pipeline material parameters in the purification monitoring data, combined with the duration and the intensity level, a business attribute set is constructed.

6. The method for real-time alarm of abnormal PVC pipe thickness according to claim 1, characterized in that: The calculating the credibility value of the alarm event using a Bayesian inference algorithm according to the intensity level, the duration, and the service attribute set includes: Based on preset historical abnormal event data, set the prior probability of abnormal events; Calculating based on the intensity level, the duration, and the service attribute set to obtain the conditional probability of each observation parameter under the condition of anomaly occurrence; The prior probability and the conditional probability are substituted into the Bayesian inference formula to calculate the posterior probability of the current alarm event as the alarm credibility value.

7. The method for real-time alarm of abnormal PVC pipe thickness according to claim 1, characterized in that: The method of dynamically adjusting the alarm threshold based on the credibility value and iteratively correcting the parameter configuration in the coupling relationship expression based on the updated alarm threshold for continuous detection and adaptive adjustment of PVC pipe thickness anomalies includes: Based on the credibility value, dynamically adjust the alarm threshold to form an updated alarm judgment standard; Acquiring actual monitoring data corresponding to the alarm judgment criteria; Substituting the actual monitoring data into the coupling relationship expression to generate an adjusted thickness prediction value; Comparing the adjusted thickness prediction value with a preset thickness threshold to obtain a deviation value; Calculating a feedback adjustment factor according to the deviation value as an iterative correction of the parameter configuration in the coupling relationship expression; The corrected parameter configuration is used for the continuous execution of subsequent PVC pipe thickness anomaly detection, supporting adaptive adjustment of alarm judgment.

8. A real-time alarm system for abnormal thickness of PVC pipes, characterized by: include: The data acquisition module is used to obtain the thickness data, temperature parameters and pressure parameters of the PVC pipe and construct multi-dimensional monitoring data including time stamps; A data purification module is used to process the multi-dimensional monitoring data using a wavelet transform noise reduction algorithm to obtain purified thickness data, purified temperature data and purified pressure data to form purified monitoring data; A coupling relationship module, configured to perform a multivariate nonlinear regression analysis based on the purification monitoring data to establish a coupling relationship expression between temperature parameters, pressure parameters and thickness change; A deviation trend module is used to calculate the deviation sequence between the thickness prediction value and the measured value based on the coupling relationship expression using a sliding window technology, and extract a deviation trend indicator reflecting the abnormal change pattern; an anomaly extraction module, configured to perform differential analysis based on the deviation trend indicator, extract the duration and intensity level of the anomaly pattern, and extract a set of business attributes in combination with the purification monitoring data; an alarm confirmation module, configured to calculate a credibility value of an alarm event using a Bayesian inference algorithm based on the intensity level, the duration, and the set of service attributes; The adaptive adjustment module is used to dynamically adjust the alarm threshold based on the credibility value, and iteratively correct the parameter configuration in the coupling relationship expression based on the updated alarm threshold, for continuous detection and adaptive adjustment of PVC pipe thickness anomalies.

9. An electronic device, characterized in that: The method comprises a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, the method for real-time alarm of abnormal thickness of a PVC pipe according to any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium, characterized in that The computer-readable storage medium includes a stored computer program, wherein when the computer program is executed, the device where the computer-readable storage medium is located is controlled to execute the real-time alarm method for abnormal thickness of a PVC pipe according to any one of claims 1 to 7.

Images