Cloud platform-based corrugated compensator multi-project management and early warning analysis system

Abstract

The application relates to the technical field of intelligent early warning, and discloses a corrugated compensator multi-project management and early warning analysis system based on a cloud platform, which comprises a sensor module, a wireless communication module and a data processing module; the sensor module comprises a temperature sensor and a displacement sensor; the wireless communication module is connected with the sensor module and is used for transmitting temperature data and / or displacement data; the data processing module collects expansion and contraction data of a corrugated pipe, obtains the fatigue frequency of the corrugated pipe, determines the residual fatigue frequency of the corrugated compensator, collects temperature data, compares the temperature data with a temperature threshold value, judges whether there is a leakage risk, extracts a feature vector data group when it is determined that there is a risk, processes the feature vector data group based on a multi-head time sequence attention mechanism, and determines comprehensive time sequence features; and the early warning module performs risk early warning according to the comprehensive time sequence features. The application improves the identification capability of early fatigue degradation and slow leakage, reduces the false alarm and missed alarm rates, and realizes unified management of multiple projects and cross-regions through centralized deployment of the cloud.

Description

Technical Field

[0001] This invention relates to the field of intelligent early warning technology, and more specifically, to a cloud-based multi-project management and early warning analysis system for corrugated compensators. Background Technology

[0002] With the continuous expansion of urban heating pipe networks, corrugated compensators, as key flexible connection components in heating pipelines, directly affect the safety of the heating network and the lifespan of the pipelines through their operational stability. Corrugated compensators primarily compensate for axial displacement caused by thermal expansion and contraction. However, during long-term operation, due to frequent heat-cold cycles and changes in the soil environment, they are prone to fatigue damage, structural deformation, or sealing failure, leading to media leakage, energy loss, equipment corrosion, and even major safety accidents.

[0003] To ensure the stable operation of heating systems, some pipe networks have deployed online monitoring methods for compensators based on temperature or displacement sensors, attempting to identify abnormal states through thresholds. However, existing technologies have several shortcomings. Traditional solutions often use static alarm thresholds set by rules, lacking the ability to mine potential risk trends and multi-dimensional interaction information in sensor time series, thus failing to accurately identify early signs of leakage or fatigue degradation. Most systems rely on single-point signals or simple fluctuations for judgment, lacking the ability to model long-term evolution processes, making them prone to false alarms or missed alarms, especially under conditions of high-frequency small deformations or slow temperature changes, resulting in low alarm signal-to-noise ratios. In large-scale pipe networks, existing systems struggle to achieve unified data management and risk normalization analysis across regions and projects, hindering systematic maintenance and resource scheduling.

[0004] Therefore, it is necessary to design a cloud-based multi-project management and early warning analysis system for corrugated compensators to solve the problems existing in the current technology. Summary of the Invention

[0005] In view of this, the present invention proposes a cloud platform-based multi-project management and early warning analysis system for corrugated compensators, which aims to solve the problems of low efficiency in the utilization of current monitoring data, poor accuracy in risk identification, and lack of multi-project collaborative management mechanism.

[0006] This invention proposes a cloud-based multi-project management and early warning analysis system for ripple compensators, comprising:

[0007] The sensor module includes a temperature sensor and a displacement sensor;

[0008] A wireless communication module is connected to the sensor module, and the wireless communication module is used to transmit temperature data and / or displacement data;

[0009] The data processing module is configured to collect the expansion and contraction data of the bellows and obtain the number of fatigue cycles of the bellows, and determine the remaining number of fatigue cycles of the bellows compensator; collect temperature data and compare it with a temperature threshold to determine whether there is a risk of leakage; and when it is determined that there is a risk of leakage or fatigue damage, collect data from all sensors within a preset time period, extract feature vector data groups, process the feature vector data groups based on a multi-head temporal attention mechanism, and determine comprehensive temporal features.

[0010] The early warning module is configured to provide risk warnings based on the comprehensive time-series characteristics.

[0011] Furthermore, the sensor module includes:

[0012] At least four temperature sensors are provided, and the temperature sensors are located on both sides of the external movable end of the bellows compensator, with at least two temperature sensors provided on each side.

[0013] At least four displacement sensors are provided, and the displacement sensors are located inside the bellows compensator on both sides of the protective bellows.

[0014] Furthermore, when the data processing module collects the expansion and contraction data of the bellows and obtains the number of fatigue cycles of the bellows, and determines the remaining fatigue cycles of the bellows compensator, it includes:

[0015] Based on a preset sampling period, the expansion and contraction data of the bellows in the bellows compensator are collected to obtain the axial expansion and contraction displacement sequence. The axial expansion and contraction displacement sequence is divided into sliding windows to construct a displacement change dataset within the time window, identify peaks and valleys, and obtain tension and compression cycles.

[0016] The amplitude and average displacement of each pair of tension-compression half-cycles are extracted based on rainflow counting.

[0017] The elastic modulus of the bellows and the stress-strain transformation coefficient of the compensator structure are collected to obtain the equivalent stress amplitude.

[0018] The standard number of fatigue cycles is obtained based on the equivalent stress amplitude and the SN fatigue life curve corresponding to the bellows material.

[0019] The fatigue consumption of the identified tension-compression cycles is accumulated based on the Miner linear damage accumulation method, and compared with the standard fatigue cycle count to obtain the remaining fatigue count.

[0020] Furthermore, after determining the remaining fatigue cycles of the corrugated compensator, the data processing module further includes:

[0021] The data processing module compares the remaining fatigue counts with the remaining fatigue threshold, and determines whether the corrugated compensator is at risk of fatigue damage based on the comparison results.

[0022] When the remaining fatigue count is less than the remaining fatigue threshold, the data processing module determines that the corrugated compensator is at risk of fatigue damage; when the remaining fatigue count is greater than or equal to the remaining fatigue threshold, the data processing module determines that the corrugated compensator is not at risk of fatigue damage.

[0023] Furthermore, when the data processing module collects temperature data and compares it with a temperature threshold to determine whether there is a risk of leakage, it includes:

[0024] When the temperature data is less than or equal to the temperature threshold, the data processing module determines that there is no risk of leakage; when the temperature data is greater than the temperature threshold, the data processing module determines that there is a risk of leakage.

[0025] Furthermore, the data processing module obtains the temperature threshold using the following method:

[0026] The data processing module collects temperature data for N consecutive sampling cycles and constructs a reference temperature curve for each sensor measuring point to obtain the normal temperature change baseline of the compensator under leak-free conditions.

[0027] The soil temperature outside the compensator is collected, and the baseline of normal temperature change is offset and corrected based on the soil temperature. The correction factor is obtained by linear regression.

[0028] The standard temperature maximum and minimum values ​​are obtained based on the corrected normal temperature change baseline. The temperature threshold is obtained based on the standard temperature minimum value and the safety margin, wherein the safety margin is 2-5℃.

[0029] Furthermore, the data processing module processes the feature vector data set based on a multi-head temporal attention mechanism to determine the comprehensive temporal features, including:

[0030] The feature vector data set includes displacement data, temperature data, displacement data acceleration change, and temperature data acceleration change.

[0031] The feature vector data set is used to construct a time-series feature matrix;

[0032] Configure the number of attention heads, and sequentially perform multi-head temporal attention processing on the temporal feature matrix. Each attention head includes an independent Query, Key, and Value linear mapper.

[0033] Calculate the attention weight of each time frame in the time series to other frames;

[0034] The representations of all heads are concatenated and linearly combined to generate a comprehensive temporal feature.

[0035] Furthermore, the data processing module processes the feature vector data set based on a multi-head temporal attention mechanism to determine the comprehensive temporal features, and also includes:

[0036] For the h-th attention head, its attention output is:

[0037] ;

[0038] in, Represents the similarity scoring matrix; Indicates the scaling factor. Represents the Value matrix. Represents the Query matrix. Represents the Key matrix. Represents the normalization function. This represents the attention output of the h-th attention head;

[0039] The softmax result is the attention weight matrix:

[0040] ;

[0041] in, This indicates the degree to which the i-th time frame in the h-th attention head sequence pays attention to the j-th frame when generating its output representation. Represents the Query vector at the i-th time step. This represents the Key vector at the j-th time step; j represents the target time frame number for attention calculation. This represents the index variable used for normalization in the denominator of the softmax function; T represents the length of the time series.

[0042] Furthermore, when the early warning module performs risk warning based on the comprehensive time-series characteristics, it includes:

[0043] The early warning module is based on a fully connected network and a softmax activation function, and performs risk warnings according to the comprehensive temporal characteristics.

[0044] ;

[0045] in, This represents the predicted probability for each risk level. Indicates the comprehensive time series characteristics, This represents the bias vector, and C represents the risk level number, C=4;

[0046] The early warning module outputs the risk level with the highest predicted probability as the risk level and simultaneously outputs the judgment result, which includes whether there is a risk of leakage and / or whether there is a risk of fatigue damage.

[0047] Furthermore, it also includes:

[0048] The data compression module is configured to perform aggregation processing on the collected temperature and displacement data and generate representative sampling points using a moving average when the data processing module determines that there is no risk of leakage or fatigue damage, in order to replace the original high-frequency sampling sequence; and to keep the collected temperature and displacement data uncompressed when the data processing module determines that there is a risk of leakage or fatigue damage.

[0049] The sampled data sequence is stored after being structured and encoded, and the structured encoding includes difference encoding, interval statistical encoding, and lightweight temporal encoding.

[0050] Compared with existing technologies, the advantages of this invention are as follows: By setting up a sensor module composed of temperature and displacement sensors, real-time acquisition of key physical parameters of the corrugated compensator (such as axial expansion and contraction and soil temperature changes) is achieved, and remote, high-frequency, low-power data transmission is realized through a wireless communication module, effectively adapting to the widely distributed urban heating pipe network scenario; the data processing module constructs a corrugated pipe expansion and contraction displacement sequence based on the sampled data, obtains the current fatigue state of the corrugated pipe through the rainflow counting method and the Miner fatigue damage model, and combines temperature data and a dynamic threshold model to identify and judge potential leakage risks; when risk signs are triggered, multi-channel sensor data within a certain time window is automatically called, and key frame change features are extracted through a multi-head temporal attention mechanism to generate a fused comprehensive temporal feature representation for accurate identification of minor trend anomalies or multi-variable interaction correlations; finally, the early warning module outputs the risk level based on the fused representation, realizing a closed-loop architecture from "physical acquisition - intelligent analysis - proactive early warning". Compared with existing solutions that rely on static thresholds and cannot model evolution trends, this solution improves the ability to identify early fatigue degradation and gradual leakage, reduces false alarm and false negative rates, and achieves unified management of multiple projects and across regions through centralized cloud deployment. Attached Figure Description

[0051] Various other advantages and benefits will become apparent to those skilled in the art upon reading the following detailed description of preferred embodiments. The accompanying drawings are for illustrative purposes only and are not intended to limit the invention. Furthermore, the same reference numerals denote the same parts throughout the drawings. In the drawings:

[0052] Figure 1The structural block diagram of the cloud platform-based multi-project management and early warning analysis system for corrugated compensators provided in the embodiments of the present invention;

[0053] Figure 2 A schematic diagram of the temperature sensor arrangement in a cloud-based multi-project management and early warning analysis system for corrugated compensators provided in an embodiment of the present invention;

[0054] Figure 3 This is a schematic diagram of the displacement sensor arrangement in a cloud-based corrugated compensator multi-project management and early warning analysis system provided in an embodiment of the present invention.

[0055] Among them, 100 is the internal protective bellows; 200 is the external moving end; 300 is the displacement sensor; and 400 is the temperature sensor. Detailed Implementation

[0056] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided to enable a more thorough understanding of the present disclosure and to fully convey the scope of the disclosure to those skilled in the art. It should be noted that, unless otherwise specified, embodiments and features in the embodiments of the present invention can be combined with each other. The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0057] In traditional, existing bellows compensator monitoring systems based on static threshold judgments, the reliance on fixed temperature and displacement thresholds for anomaly detection makes it difficult to capture the multidimensional temporal correlation features hidden in sensor data. This makes it impossible to effectively identify slowly changing risk signals caused by material fatigue accumulation or minor leaks. The system lacks the ability to analyze dynamic evolution patterns over long time series, leading to high-frequency minute displacement fluctuations and gradual temperature rises being easily misjudged as noise. It also fails to accurately distinguish between normal thermal expansion and contraction and early damage signs. In pipeline network scenarios with multiple compensators operating collaboratively, isolated data storage and analysis further exacerbate the probability of missing cross-project risk characteristics.

[0058] For example, in inter-regional heating networks, multiple corrugated compensators are subjected to cyclical thermal stress and soil pressure over long periods. Traditional monitoring systems independently compare the displacement and temperature data of each compensator against thresholds. When a compensator experiences a single displacement fluctuation due to material fatigue, and the fluctuation frequency increases within adjacent time windows, this abnormal pattern cannot be identified through isolated data points. Furthermore, due to diurnal temperature fluctuations, the surface temperature monitoring values ​​of the compensators exhibit baseline drift, and the static threshold cannot be adaptively adjusted, causing the initial characteristics of slowly changing temperature leaks to be masked by environmental noise. In multi-project management scenarios, the operational data of compensators in different geographical areas lack unified time-series modeling, making it impossible to discover abnormal evolution trends through historical data comparison.

[0059] If the above issues are not addressed, gradually changing fatigue damage and minor leaks cannot be effectively identified in the early stages, leading to the continuous accumulation of structural damage to the compensator until sudden failure. In multi-project management scenarios, isolated data analysis will prevent the horizontal comparison of risk characteristics of compensators in different areas, delaying systematic maintenance decisions. The cumulative damage caused by long-term, high-frequency, minor displacements may trigger a chain reaction of structural failures, significantly increasing the risk of pipeline leaks and outages. At the same time, false alarms caused by static thresholds will reduce the efficiency of operation and maintenance response and increase the cost of manual verification.

[0060] To address the aforementioned challenges, this application first considers how to overcome the limitations of traditional static thresholds by capturing multi-dimensional temporal correlation features to identify gradually changing risks. Existing single-point monitoring cannot distinguish between normal fluctuations and early damage; therefore, a dynamic analysis model needs to be constructed to jointly model displacement and temperature data over time. For high-frequency, minute displacements and gradually changing temperatures, this application attempts to combine the displacement change dataset within a time window with the temperature change baseline to extract tension-compression cycle features and environmental correction factors. A multi-head temporal attention mechanism is introduced to simultaneously process multi-dimensional features such as displacement acceleration and temperature change rate. Through self-attention weights, abnormal patterns in key time frames are dynamically captured, forming a comprehensive temporal analysis framework that integrates fatigue damage assessment and leakage risk determination.

[0061] See Figure 1As shown, this application proposes a cloud-based multi-project management and early warning analysis system for bellows compensators, comprising: a sensor module, including a temperature sensor and a displacement sensor; a wireless communication module connected to the sensor module, used to transmit temperature data and / or displacement data; a data processing module configured to collect bellows expansion and contraction data and obtain the number of bellows fatigue cycles, determining the remaining fatigue cycles of the bellows compensator; collecting temperature data and comparing it with a temperature threshold to determine if there is a leakage risk; and, when a leakage risk or fatigue damage risk is determined, collecting data from all sensors within a preset time period, extracting feature vector data sets, processing the feature vector data sets based on a multi-head temporal attention mechanism to determine comprehensive temporal features; and an early warning module configured to provide risk warnings based on the comprehensive temporal features.

[0062] Specifically, temperature and displacement sensors are devices used to monitor the operating status of the bellows compensator and environmental changes in real time. Temperature detection can be achieved using thermocouples, resistance temperature detectors (RTDs), or fiber Bragg grating temperature sensors, while displacement measurement can be achieved using laser displacement sensors or capacitive displacement sensors. Multi-point deployment improves the comprehensiveness of data acquisition and the sensitivity of anomaly detection. The wireless communication module transmits temperature and displacement data, employing LoRa, NB-IoT, or 5G communication protocols to achieve low-power, long-distance data transmission, ensuring real-time uploading of monitoring data from multiple projects to the cloud platform. The bellows fatigue cycle calculation in the data processing module assesses the degree of structural fatigue damage based on axial expansion and contraction displacement sequences and material properties. Specifically, the rainflow counting method can be used to extract the number of tension and compression cycles, combined with the SN fatigue life curve and Miner's linear damage accumulation method to predict the remaining life, addressing the problem that traditional static thresholds cannot capture long-term fatigue degradation. Leakage risk assessment compares temperature data with dynamically corrected temperature thresholds. Historical temperature data can be used to construct a baseline for normal temperature changes, combined with linear regression correction using ambient soil temperature, generating a safety margin threshold adapted to different operating conditions, avoiding misjudgments caused by a single threshold. Comprehensive temporal feature generation refers to processing multi-dimensional sensor data based on a multi-head temporal attention mechanism. Specifically, multiple independent attention heads can be used to capture the temporal correlation of displacement, temperature, and acceleration changes. The dependency relationship between different time frames is quantified through an attention weight matrix, improving the ability to identify early leaks and minor deformations. The risk warning module performs multi-level risk classification based on comprehensive temporal features. Specifically, a fully connected neural network combined with a softmax activation function can be used to output a risk probability distribution, achieving an end-to-end mapping from data features to risk levels and enhancing the accuracy of warnings under complex operating conditions.

[0063] Understandably, by integrating multi-project management functions through a cloud platform, combined with dynamically corrected temperature thresholds, fatigue damage accumulation models, and multi-head time-series attention mechanisms, refined monitoring of the bellows compensator's operating status and early risk warnings can be achieved. Through time-series analysis of multi-dimensional sensor data and machine learning algorithms, the problems of poor adaptability to static thresholds and weak correlation of multi-dimensional data in traditional solutions are effectively solved, improving the detection accuracy and alarm timeliness of leaks and fatigue damage, while also supporting unified management and analysis of large-scale pipeline network data.

[0064] The working process and principle of this application are as follows: The multi-project management and early warning analysis system for bellows compensators based on a cloud platform includes a sensor module, a wireless communication module, a data processing module, and an early warning module. The sensor module includes temperature and displacement sensors for collecting temperature and displacement data from the bellows compensator. The wireless communication module connects to the sensor module and is responsible for transmitting the collected temperature and displacement data. The data processing module first collects the expansion and contraction data of the bellows and obtains the number of fatigue cycles, then calculates the remaining fatigue cycles of the bellows compensator. Next, it collects temperature data and compares it with a preset temperature threshold to determine if there is a risk of leakage. When a risk of leakage or fatigue damage is determined, the data processing module collects data from all sensors within a preset time period and extracts a feature vector data set. Then, it processes the feature vector data set based on a multi-head temporal attention mechanism to determine comprehensive temporal features. Finally, the early warning module issues a risk warning based on the comprehensive temporal features.

[0065] By collecting bellows expansion and contraction data and temperature data, combined with a multi-head temporal attention mechanism, the temporal correlation features in the data can be effectively captured, enabling early identification of potential risks in bellows compensators. The multi-head temporal attention mechanism can simultaneously process multi-dimensional features such as displacement and temperature, and dynamically captures abnormal patterns at key time points by calculating attention weights between different time frames. Compared to traditional static threshold judgment, it can more accurately identify slowly changing risk signals caused by material fatigue accumulation or minor leaks.

[0066] Understandably, employing a multi-head temporal attention mechanism to process feature vector data sets can effectively extract long-term dependencies in time-series data, overcoming the problem that traditional methods struggle to capture dynamic evolution patterns in long-term series. Through comprehensive analysis of multi-dimensional features, it is possible to more accurately distinguish between normal thermal expansion and contraction and early signs of damage, thus improving the accuracy of risk warnings.

[0067] As a preferred embodiment, the solution of this application is specifically implemented as follows:

[0068] The sensor module includes multiple temperature and displacement sensors, installed at key locations on the bellows compensator. The temperature sensors monitor the surface temperature of the bellows, and the displacement sensors measure the axial displacement of the bellows.

[0069] The wireless communication module uses Bluetooth Low Energy technology to periodically transmit the data collected by the sensors to the cloud platform.

[0070] The data processing module first processes the expansion and contraction data of the bellows. It analyzes the axial expansion and contraction displacement sequence using the sliding window method to identify tension-compression cycles, and extracts the amplitude and average displacement of each pair of tension-compression half-cycles using the rainflow counting method. Combining the bellows' material parameters, it calculates the equivalent stress amplitude and determines the standard fatigue cycle number based on the SN fatigue life curve. The Miner linear damage accumulation method is used to calculate fatigue consumption, ultimately yielding the remaining fatigue cycles.

[0071] For temperature data, a baseline for normal temperature variation is established. Temperature data from multiple consecutive sampling periods are collected, and combined with external environmental soil temperature, a correction factor is obtained through linear regression to correct the baseline shift. A temperature threshold is then determined based on the corrected baseline for leak risk assessment.

[0072] When a risk is identified, a multi-head temporal attention mechanism is triggered. This mechanism combines displacement data, temperature data, and their acceleration changes into a feature vector data set to establish a temporal feature matrix. By configuring multiple attention heads, each containing an independent Query, Key, and Value linear mapper, the attention weight of each time frame in the time series to other frames is calculated. The representations of all heads are concatenated and linearly combined to generate a comprehensive temporal feature.

[0073] The early warning module uses a fully connected network and a softmax activation function to predict risk levels based on comprehensive temporal features. It outputs the risk level with the highest predicted probability and provides a judgment result indicating whether there is a risk of leakage or fatigue damage.

[0074] The above solution effectively addresses the problems existing in traditional corrugated compensator monitoring systems. By employing multi-dimensional temporal correlation feature analysis, it overcomes the limitations of static thresholds, accurately identifying risk signals of gradual deformation caused by material fatigue accumulation or minor leaks. The introduction of a multi-head temporal attention mechanism enables the system to simultaneously process multi-dimensional features such as displacement acceleration and temperature change rate. Through self-attention weights, it dynamically captures abnormal patterns in key time frames, improving the ability to identify high-frequency minute displacement fluctuations and gradual temperature rise phenomena.

[0075] The cloud-based design enables unified data management and risk normalization analysis across regions and projects, overcoming the problems of isolated data storage and analysis in traditional systems. It can identify abnormal trends through historical data comparison, effectively supporting systemic maintenance decisions.

[0076] This application's solution integrates fatigue damage assessment and leakage risk determination to form a comprehensive time-series analysis framework, improving the accuracy and timeliness of early risk identification for bellows compensators. This helps prevent the continuous accumulation of structural damage to the compensator until sudden failure, reduces false alarm rates, improves operational and maintenance response efficiency, and reduces manual verification costs.

[0077] In some of the solutions described above in this application, the sensor module has insufficient number of temperature and displacement sensors or incomplete position coverage, which makes it impossible to effectively capture the dynamic changes of the external moving end 200 and the internal protective bellows 100 of the bellows compensator. Especially under complex working conditions, it is difficult to distinguish between normal fluctuations and abnormal offsets, resulting in local deviations in monitoring data and affecting the accuracy of leakage and fatigue damage judgment.

[0078] See Figure 2-3 As shown, this application further proposes that at least four temperature sensors 400 are provided, and the temperature sensors 400 are provided on both sides of the external movable end 200 of the bellows compensator, with at least two temperature sensors 400 provided on each side. At least four displacement sensors 300 are provided, and the displacement sensors 300 are provided on both sides of the protective bellows 100 inside the bellows compensator.

[0079] Temperature sensors 400 are distributed on both sides of the external movable end 200 of the bellows compensator, with at least two temperature sensors 400 on each side, forming a symmetrical monitoring layout. Displacement sensors 300 are distributed on both sides of the internal protective bellows 100 of the bellows compensator, with at least two displacement sensors 300 on each side, forming a synchronous monitoring structure for axial and radial displacement. The displacement sensors 300 are distributed on both sides of the bellows compensator. Through this symmetrical layout, the temperature sensors 400 can capture temperature gradient changes on both sides of the movable end, and the displacement sensors 300 can simultaneously acquire deformation differences on both sides of the protective bellows, eliminating errors caused by single-sided data acquisition. For example, with two temperature sensors 400 on each side of the external movable end 200, a four-sensor redundant monitoring system is formed. When one sensor malfunctions, the sensor on the other side can still provide valid data. With two displacement sensors 300 on each side of the internal protective bellows 100, both axial expansion and contraction and radial displacement can be monitored simultaneously, avoiding the bias of displacement data in only one direction.

[0080] Specifically, the temperature sensors 400 on both sides of the external active end 200, through multi-point coverage, can identify localized temperature anomalies caused by seal failure, such as temperature rise caused by the accumulation of leaked medium on one side. The displacement sensors 300 on both sides of the internal protective bellows 100, through symmetrical arrangement, can simultaneously record displacement differences on both sides of the bellows, such as axial offset and radial torsion caused by uneven soil settlement. When the data processing module collects data, the data from both sensors are weighted averaged or compared by difference to eliminate environmental interference and extract the true displacement and temperature changes. For example, when the data from one side of the displacement sensor 300 deviates significantly from the other side, it can be determined that there is a localized deformation anomaly. The redundant arrangement of the temperature sensors 400 on both sides further reduces the risk of misjudgment due to single-point failure and improves the robustness of the monitoring system. Thus, by optimizing the number and position of sensors, the complementarity and reliability of multi-dimensional data are enhanced, providing high-precision input for subsequent fatigue damage and leakage risk analysis.

[0081] As a preferred embodiment, the solution of this application is specifically implemented as follows:

[0082] The sensor module includes temperature sensors 400 and displacement sensors 300. Four temperature sensors 400 are provided, distributed on both sides of the external movable end 200 of the bellows compensator, with two temperature sensors 400 on each side. Four displacement sensors 300 are provided, distributed on both sides of the protective bellows 100 inside the bellows compensator.

[0083] Specifically, the temperature sensor 400 is a PT100 platinum resistance temperature sensor with a measurement range of -50℃ to 200℃ and an accuracy of ±0.1℃. Four temperature sensors 400 are installed on both sides of the external movable end 200 of the bellows compensator, two on each side, in a symmetrical distribution. The displacement sensor 300 is a linear displacement sensor with a measurement range of 0-100mm and a resolution of 0.01mm. Four displacement sensors 300 are installed on both sides of the protective bellows 100 inside the bellows compensator, two on each side, in a symmetrical distribution.

[0084] Through the above technical solution, this application achieves comprehensive monitoring of temperature and displacement in key components of the bellows compensator. The symmetrical arrangement of temperature sensors 400 accurately captures the temperature distribution at the external moving end 200 of the bellows compensator, helping to promptly detect localized abnormal temperature rises. The symmetrical arrangement of displacement sensors 300 precisely measures the axial and radial deformation of the bellows, facilitating the assessment of the bellows' fatigue state. The multi-point measurement arrangement improves the reliability and representativeness of the data, providing comprehensive basic data support for subsequent data processing and risk analysis.

[0085] In some of the solutions described above in this application, the data processing module performs fatigue analysis by collecting the expansion and contraction data of the bellows. However, existing methods are difficult to accurately quantify the fatigue consumption process of the bellows compensator, resulting in insufficient accuracy in assessing the remaining fatigue cycles and an inability to effectively determine the risk of early fatigue damage.

[0086] This application further proposes a data processing module for collecting bellows expansion and contraction data and obtaining the number of bellows fatigue cycles to determine the remaining fatigue cycles of the bellows compensator. This includes: collecting expansion and contraction data of the bellows in the compensator based on a preset sampling period to obtain an axial expansion and contraction displacement sequence; dividing the axial expansion and contraction displacement sequence into a sliding window to construct a displacement change dataset within the time window; identifying peaks and valleys and obtaining tension-compression cycles; extracting the amplitude and average displacement of each pair of tension-compression half-cycles based on rainflow counting; collecting the elastic modulus of the bellows and the stress-strain transformation coefficient of the compensator structure to obtain the equivalent stress amplitude; obtaining the standard fatigue cycle number based on the equivalent stress amplitude and the SN fatigue life curve corresponding to the bellows material; and accumulating fatigue consumption of the identified tension-compression cycles based on the Miner linear damage accumulation method and comparing it with the standard fatigue cycle number to obtain the remaining fatigue cycles.

[0087] The sliding window division employs a fixed-time-length or event-triggered dynamic window mechanism, with the window length dynamically adjusted based on the compensator's design life and operating environment. The rainflow counting algorithm, combined with peak-valley detection and cycle closure conditions, transforms asymmetric fluctuations into equivalent symmetric cycles. The equivalent stress amplitude is calculated by multiplying the elastic modulus by the stress-strain conversion coefficient, converting displacement changes into stress amplitudes within the material. The SN curve is fitted based on material fatigue test data, and the standard fatigue cycle number is determined through interpolation or extrapolation. The Miner linear damage accumulation method weights and sums the damage levels corresponding to different stress amplitudes, determining fatigue life exhaustion when the cumulative damage level reaches 1.

[0088] Specifically, by dividing the displacement sequence through a sliding window, the expansion and contraction fluctuations of the bellows under different operating conditions can be dynamically captured, eliminating noise interference. The rainflow counting algorithm decomposes complex displacement fluctuations into independent tension-compression cycles, accurately quantifying the number of cycles in actual operation. The introduction of elastic modulus and stress-strain transformation calculation coefficients converts displacement data into material stress parameters, matching the input requirements of the SN curve. Damage accumulation calculation based on the SN curve and Miner criterion comprehensively considers the influence of different stress amplitudes on fatigue life, avoiding the limitations of traditional single threshold judgment. For example, when a fluctuation with a peak value of +8mm and a valley value of -5mm is detected in the axial expansion and contraction displacement sequence, rainflow counting decomposes it into a symmetrical cycle with an amplitude of 6.5mm, combined with the elastic modulus... With a conversion factor of 0.15, the calculated equivalent stress amplitude is 195 MPa, corresponding to the standard number of cycles in the SN curve. If the accumulated damage level during actual operation is 0.85, then the remaining fatigue cycles are: This method, by combining dynamic data acquisition with material mechanical properties, enables a refined assessment of fatigue life.

[0089] As a preferred embodiment, the solution of this application is specifically implemented as follows:

[0090] The data processing module collects the expansion and contraction data of the bellows in the bellows compensator based on a preset sampling period to obtain an axial expansion and contraction displacement sequence. The axial expansion and contraction displacement sequence is divided into sliding windows to construct a displacement change dataset within the time window. Tension-compression cycles are obtained by identifying peaks and valleys.

[0091] Furthermore, the amplitude and average displacement of each pair of tension-compression half-cycles are extracted based on the rainflow counting method. Specifically, a four-point method is used for rainflow counting, and displacement data points are read sequentially to determine whether a closed loop is formed. For each closed loop, its amplitude and average displacement values ​​are recorded.

[0092] Therefore, by collecting the elastic modulus of the bellows and the stress-strain transformation coefficients of the compensator structure, the equivalent stress amplitude can be obtained. For example, the stress-strain transformation coefficients can be determined through finite element analysis to convert the displacement amplitude into the stress amplitude.

[0093] The standard number of fatigue cycles is obtained by using the equivalent stress amplitude and the corresponding SN fatigue life curve of the bellows material. The SN curve, obtained through material fatigue testing, represents the number of cycles the material can withstand at a specific stress level.

[0094] Finally, fatigue wear is accumulated for the identified tension-compression cycles using the Miner linear damage accumulation method, and compared with the standard fatigue cycle count to obtain the remaining fatigue life. Specifically, the fatigue damage amount for each cycle is calculated, and the total damage amount is obtained by summing them up. Subtracting the total damage amount from 1 gives the remaining life percentage. Multiplying this percentage by the standard fatigue cycle count yields the remaining fatigue life.

[0095] Through the above technical solution, this application achieves accurate assessment of fatigue damage in bellows compensators. By collecting real-time expansion and contraction data and performing rainflow counting analysis, stress cycles under actual working conditions are accurately identified. Combining material SN curves and Miner's cumulative damage theory, the remaining fatigue life is reliably predicted. This method avoids the limitations of traditional static threshold judgment, enabling dynamic tracking of the fatigue damage evolution process and timely detection of potential risks. Furthermore, this scheme considers the actual stress amplitude distribution, making it more accurate than simple cycle count statistics. This provides a reliable basis for preventative maintenance and replacement decisions for bellows compensators, effectively reducing the risk of sudden failures.

[0096] In some of the solutions mentioned above in this application, a method is proposed to evaluate the condition of the compensator by collecting bellows expansion and contraction data and calculating the remaining fatigue cycles. However, since the fatigue consumption of the compensator has nonlinear cumulative characteristics in actual operation, simply calculating the remaining fatigue cycles cannot directly determine whether it is close to the critical damage state, which can easily lead to maintenance delays or misjudgment risks.

[0097] This application further proposes a data processing module that compares the remaining fatigue cycles with the remaining fatigue threshold, and determines whether the bellows compensator is at risk of fatigue damage based on the comparison result. When the remaining fatigue cycles are less than the remaining fatigue threshold, the data processing module determines that the bellows compensator is at risk of fatigue damage. When the remaining fatigue cycles are greater than or equal to the remaining fatigue threshold, the data processing module determines that the bellows compensator is not at risk of fatigue damage.

[0098] The remaining fatigue threshold is set based on the fatigue life safety factor of the bellows material, which is determined by combining material test data and engineering experience. The comparison process employs a real-time dynamic threshold adjustment mechanism, dynamically correcting the threshold range based on historical data of temperature and pressure fluctuations in the compensator's environment. The risk assessment logic incorporates a multi-verification mechanism, including triggering a judgment when the remaining fatigue counts for three consecutive sampling cycles are all below the threshold.

[0099] Specifically, after calculating the remaining fatigue cycles based on the rainflow counting method and the Miner linear damage accumulation model, the data processing module compares the calculation results with a preset dynamic threshold in real time. When the remaining fatigue cycles are lower than the dynamically adjusted remaining fatigue threshold, the system automatically triggers a fatigue damage risk marker and transmits the risk level data to the early warning module. By matching the threshold with the real-time value of the remaining fatigue cycles, misjudgments caused by material performance degradation or sudden environmental changes are effectively avoided. For example, under conditions of frequent pressure fluctuations, the remaining fatigue threshold can be adjusted downwards based on the average stress amplitude of the previous sampling period, making the risk assessment more consistent with the actual operating conditions. After the assessment result is generated, the data storage module is triggered to record all displacement and temperature data before and after the assessment, providing data support for subsequent maintenance.

[0100] As a preferred embodiment, the solution of this application is specifically implemented as follows:

[0101] The data processing module compares the remaining fatigue cycles with the remaining fatigue threshold to determine whether the bellows compensator is at risk of fatigue damage. Specifically, if the remaining fatigue cycles are less than the remaining fatigue threshold, the data processing module determines that the bellows compensator is at risk of fatigue damage. If the remaining fatigue cycles are greater than or equal to the remaining fatigue threshold, the data processing module determines that the bellows compensator is not at risk of fatigue damage.

[0102] For example, the remaining fatigue threshold can be set to 1000 cycles. The data processing module calculates that the remaining fatigue cycles of the bellows compensator are 800. Therefore, the data processing module compares 800 cycles with 1000 cycles. Since 800 cycles is less than 1000 cycles, it determines that the bellows compensator has a risk of fatigue damage. Further, if the calculated remaining fatigue cycles are 1200 cycles, then since 1200 cycles is greater than 1000 cycles, the data processing module will determine that the bellows compensator does not have a risk of fatigue damage.

[0103] Through the above technical solution, this application can promptly detect the fatigue damage risk of bellows compensators. By setting a reasonable remaining fatigue threshold, a warning can be issued before the bellows compensator reaches its fatigue limit, allowing maintenance personnel sufficient time for repairs. This method of judgment based on the number of remaining fatigue cycles, compared to simple displacement or temperature threshold judgments, can more accurately reflect the actual usage condition and lifespan of the bellows compensator, effectively avoiding misjudgments and omissions. Simultaneously, this method also provides a reliable basis for preventative maintenance of bellows compensators, helping to extend equipment service life and reduce the risk of sudden failures.

[0104] In some of the solutions described above in this application, the data processing module judges the leakage risk by temperature threshold. However, the existing static threshold setting method cannot adapt to changes in ambient temperature, which increases the probability of misjudging risk in scenarios with large fluctuations in soil temperature or large seasonal temperature differences. Especially when the compensator is in an unstable heat exchange state, the fixed threshold is difficult to accurately reflect the abnormal temperature rise caused by the actual leakage.

[0105] This application further proposes a data processing module that collects temperature data and compares it with a temperature threshold to determine whether there is a leakage risk. This includes: when the temperature data is less than or equal to the temperature threshold, the data processing module determines that there is no leakage risk; when the temperature data is greater than the temperature threshold, the data processing module determines that there is a leakage risk.

[0106] The temperature threshold is obtained through a dynamic correction method. Specifically, this involves constructing a baseline for normal temperature changes based on temperature data from continuous sampling periods, adjusting for offsets by incorporating external environmental soil temperature, and finally determining the maximum and minimum standard temperatures based on the corrected baseline. A safety margin is then added to the minimum standard temperature value to generate the temperature threshold. The data comparison process employs logical judgment between real-time temperature data and the dynamic threshold. Once the judgment result is triggered, it links to subsequent feature vector data acquisition and processing procedures.

[0107] Specifically, the data processing module first acquires real-time temperature data from 400 measuring points on the current temperature sensor, and then compares the values ​​with a preset dynamic temperature threshold. If the real-time temperature does not exceed the threshold, it is determined that there is no leakage risk in the current state, and the normal monitoring mode is maintained. If the real-time temperature exceeds the threshold, a leakage risk assessment is triggered, and full acquisition and feature extraction of all sensor data within a preset time period are simultaneously initiated. This assessment logic reduces misjudgments caused by environmental interference through a dynamic threshold correction mechanism, while ensuring early warning for minor temperature anomalies based on a safety margin setting. For example, when soil temperature fluctuates due to diurnal temperature variations, the dynamic threshold is automatically adjusted according to a correction factor to avoid misjudging leakage due to rising ambient temperature. When an actual leakage causes a local temperature rise exceeding the corrected threshold, a risk assessment is immediately triggered. The assessment result further drives the data processing module to perform multi-dimensional feature analysis to ensure the accuracy and timeliness of risk identification.

[0108] As a preferred embodiment, the solution of this application is specifically implemented as follows:

[0109] The data processing module collects temperature data and compares it with a temperature threshold to determine if there is a risk of leakage. Specifically, when the temperature data is less than or equal to the temperature threshold, the data processing module determines that there is no risk of leakage. When the temperature data is greater than the temperature threshold, the data processing module determines that there is a risk of leakage.

[0110] Furthermore, temperature data is acquired by at least four temperature sensors 400 located on both sides of the external movable end 200 of the bellows compensator. For example, two temperature sensors 400 can be installed on each side of the external movable end 200 of the bellows compensator, for a total of four temperature sensors 400. This allows for comprehensive monitoring of temperature changes in the bellows compensator.

[0111] The data processing module first receives temperature data from the temperature sensor 400. This temperature data can be real-time acquired data or pre-processed data. Subsequently, the data processing module compares the received temperature data with a preset temperature threshold. The temperature threshold can be preset based on factors such as the design parameters of the bellows compensator and the operating environment.

[0112] Specifically, the data processing module executes the following judgment logic: if the temperature data collected by any temperature sensor 400 is greater than the temperature threshold, it is determined that there is a risk of leakage. If the temperature data collected by all temperature sensors 400 is less than or equal to the temperature threshold, it is determined that there is no risk of leakage.

[0113] As a preferred implementation, multiple temperature thresholds can be set, corresponding to different risk levels. For example, low-risk, medium-risk, and high-risk thresholds can be set. When the temperature data exceeds the low-risk threshold but does not exceed the medium-risk threshold, it is determined to be a low-risk leak. When the temperature data exceeds the medium-risk threshold but does not exceed the high-risk threshold, it is determined to be a medium-risk leak. When the temperature data exceeds the high-risk threshold, it is determined to be a high-risk leak.

[0114] Through the above technical solution, this application can promptly detect leakage risks in bellows compensators. By comparing temperature data with preset temperature thresholds, the existence of leakage risk can be quickly determined, avoiding energy loss and equipment corrosion caused by leakage. Furthermore, the use of multiple temperature sensors (400°C) for monitoring improves the comprehensiveness and accuracy of monitoring. By setting multi-level temperature thresholds, the degree of leakage risk can be more precisely assessed, facilitating the implementation of corresponding preventative and remedial measures. This method is simple, intuitive, and easy to implement, effectively improving the operational safety and reliability of bellows compensators.

[0115] In some of the solutions described above in this application, the traditional static temperature threshold does not take into account the impact of external ambient temperature fluctuations on leak detection, which may lead to misjudgment or missed detection when soil temperature changes.

[0116] This application further proposes a data processing module that obtains the temperature threshold using the following method: Temperature data is collected for N consecutive sampling periods to construct a baseline temperature curve for each sensor measurement point, obtaining the normal temperature variation baseline of the compensator under leak-free conditions. The ambient soil temperature is collected, and the normal temperature variation baseline is offset and corrected based on the soil temperature; the correction factor is obtained through linear regression. The standard temperature maximum and minimum values ​​are obtained from the corrected normal temperature variation baseline. The temperature threshold is then obtained based on the standard temperature minimum value and a safety margin of 2-5℃.

[0117] The baseline temperature curve is constructed by statistically analyzing temperature data from a leak-free state over N consecutive cycles, ensuring that the baseline reflects the temperature fluctuation range during normal operation of the compensator. Linear regression correction of soil temperature establishes a mapping relationship between the compensator surface temperature and ambient temperature, eliminating the interference of external ambient temperature changes on the baseline curve. The safety margin is determined through a combination of engineering experience and leakage temperature rise rate; for example, the threshold sensitivity is adjusted within a 2-5℃ range based on the type of pipeline medium.

[0118] Specifically, under leak-free conditions, the data processing module continuously collects surface temperature data of the compensator and generates a baseline temperature curve for each measuring point using a moving average algorithm. When the ambient soil temperature changes seasonally, a linear regression model is used with soil temperature as the independent variable and compensator surface temperature as the dependent variable to calculate the temperature offset and update the baseline curve. The corrected baseline curve separates the superposition effect of ambient temperature and leakage temperature rise. The extracted standard temperature minimum value is then superimposed with a 2-5℃ safety margin to form a dynamically adjusted temperature threshold. For example, when the soil temperature drops sharply in winter, the correction factor lowers the baseline temperature minimum value to avoid misjudging leaks in low-temperature environments. Conversely, in the high temperatures of summer, the baseline temperature minimum value is increased to prevent ambient temperature rise from masking leakage anomalies. By combining dynamic thresholds and safety margins, both underreporting of minor leaks and suppressing false alarms caused by environmental interference are avoided.

[0119] As a preferred embodiment, the solution of this application is specifically implemented as follows:

[0120] The data processing module collects temperature data for 30 consecutive sampling periods, constructs a baseline temperature curve for each sensor measuring point, and obtains the normal temperature variation baseline of the compensator under leak-free conditions. Further, it collects the ambient soil temperature outside the compensator and corrects the normal temperature variation baseline based on this soil temperature. The correction factor is obtained through linear regression. Specifically, the correlation coefficient between soil temperature and the temperature at each measuring point is first calculated, and then a linear regression equation is fitted using the least squares method to obtain the correction factor. Thus, the standard temperature maximum and minimum values ​​are obtained based on the corrected normal temperature variation baseline. For example, the 95th percentile of the corrected baseline data is taken as the standard temperature maximum, and the 5th percentile as the standard temperature minimum. Finally, a temperature threshold of 3℃ is obtained based on the standard temperature minimum and a safety margin.

[0121] Through the above technical solution, this application achieves dynamic adjustment of the temperature threshold of the corrugated compensator. By considering the influence of the external soil temperature and making corresponding offset corrections, the obtained temperature threshold is more accurate and reliable. This method can adapt to temperature changes under different environmental conditions, reduce misjudgments caused by environmental factors, and improve the accuracy of leakage risk assessment. At the same time, by setting a safety margin, the reliability and stability of the system are enhanced. This dynamic adjustment mechanism enables the system to better adapt to the actual operating environment and improves the overall performance of the corrugated compensator multi-project management and early warning analysis system.

[0122] In some of the solutions described above in this application, when the data processing module uses feature vector data sets for risk analysis, the single feature dimension and lack of dynamic change information make it impossible to effectively capture the complex dynamic relationships in multi-dimensional time series data, thus affecting the accuracy of early risk identification.

[0123] This application further proposes a feature vector data set including displacement data, temperature data, displacement data acceleration changes, and temperature data acceleration changes. The feature vector data set is constructed as a temporal feature matrix. The number of attention heads is configured, and the temporal feature matrix is ​​sequentially subjected to multi-head temporal attention processing. Each attention head includes an independent Query, Key, and Value linear mapper, calculating the attention weight of each time frame in the time series to other frames. The representations of all heads are concatenated and linearly combined to generate a comprehensive temporal feature.

[0124] The feature vector data set incorporates acceleration changes from displacement and temperature data, reflecting the dynamic trends of physical quantities and supplementing transient features that static data cannot capture. The construction of the temporal feature matrix aligns multi-source heterogeneous data along the time dimension, forming a unified data structure. In multi-head temporal attention processing, each attention head generates different Query, Key, and Value matrices through an independent linear mapper, enabling different heads to focus on association patterns at different time scales. Attention weight calculation quantifies the degree of influence between different time frames through a scaled dot product mechanism, such as the correlation strength between the current moment and historical moments.

[0125] Specifically, the acceleration changes in the feature vector data set are obtained through difference operations; for example, the acceleration change in displacement data is the second-order difference of the displacement values ​​of adjacent sampling points. The row dimension of the temporal feature matrix corresponds to the time step sequence, and the column dimension contains four feature channels: displacement, temperature, and their acceleration changes. During multi-head processing, each attention head performs a linear transformation on the temporal feature matrix, generating three subspace representations: Query, Key, and Value. By calculating the dot product similarity between Query and Key, and normalizing it using softmax, an attention weight matrix is ​​obtained. This matrix reflects the correlation strength between different time frames. Multiplying the weight matrix with the Value matrix yields the attention output of each head. Finally, the outputs of multiple heads are concatenated and linearly projected to form a comprehensive temporal feature. For example, when processing high-frequency small deformations, one attention head focuses on capturing short-term abrupt changes in displacement acceleration, while another focuses on the long-term cumulative trend of temperature changes. Through the collaborative analysis of multi-dimensional temporal features, the accuracy of identifying complex risks is improved.

[0126] As a preferred embodiment, the solution of this application is specifically implemented as follows:

[0127] The data processing module processes the feature vector data set based on a multi-head temporal attention mechanism. When determining the comprehensive temporal features, it includes the following steps:

[0128] First, the feature vector data set includes displacement data, temperature data, displacement data acceleration change, and temperature data acceleration change.

[0129] Secondly, the feature vector data set is constructed into a time-series feature matrix. Specifically, the collected data of various types are arranged in chronological order to form a matrix, where each row represents a time point and each column represents a feature.

[0130] Furthermore, configure the number of attention heads. For example, you can set up 8 attention heads. The temporal feature matrix is ​​then processed sequentially using multi-head temporal attention, with each attention head including an independent Query, Key, and Value linear mapper. Thus, each attention head can focus on different combinations of features.

[0131] Then, the attention weights of each time frame in the time series to other frames are calculated. Specifically, the similarity is calculated by the dot product of the query and the key, and then normalized by the softmax function to obtain the attention weights.

[0132] Finally, the representations of all attention heads are concatenated and linearly combined to generate a comprehensive temporal feature. This step merges the outputs of multiple attention heads to obtain a comprehensive feature representation containing multi-dimensional information.

[0133] Through the above technical solution, this application can effectively capture long-term temporal dependencies and feature interaction information in the multi-dimensional sensor data of the corrugated compensator. This improves the accuracy of identifying early leakage and fatigue degradation signs in the corrugated compensator, reducing false alarms and missed alarms. Furthermore, this solution can adapt to complex scenarios such as high-frequency minute deformation or gradual temperature changes, improving the robustness of the early warning system under different operating conditions. Simultaneously, the cloud platform-based architecture enables unified data management and risk normalization analysis across regions and projects, providing strong support for systematic maintenance and resource scheduling.

[0134] In some of the solutions described above in this application, when the data processing module processes feature vector data groups based on the multi-head temporal attention mechanism, the dependency relationship between different time frames is difficult to be fully captured, resulting in limited accuracy of temporal feature extraction and affecting the accuracy of subsequent risk warning.

[0135] This application further proposes that for the h-th attention head, its attention output is: .

[0136] in, Represents the similarity scoring matrix; Indicates the scaling factor. Represents the Value matrix. Represents the Query matrix. Represents the Key matrix. Represents the normalization function. This represents the attention output result of the h-th attention head.

[0137] The softmax result is the attention weight matrix:

[0138] ;

[0139] in, This indicates the degree to which the i-th time frame in the h-th attention head sequence pays attention to the j-th frame when generating its output representation. Represents the Query vector at the i-th time step. This represents the Key vector at the j-th time step; j represents the target time frame number for attention calculation. This represents the index variable used for normalization in the denominator of the softmax function; T represents the length of the time series.

[0140] Specifically, the product of the Query matrix and the Key matrix generates the original attention score matrix, and the scaling factor controls gradient stability. The normalization function converts the original scores into a probability distribution, ensuring that the attention weights fall within the [0,1] interval. The Value matrix is ​​multiplied by the attention weight matrix to output a weighted feature representation. The similarity between the Query vector of each time frame and the Key vectors of all time frames is calculated to generate a dynamic attention distribution. Through parallel computation of multiple attention heads, the temporal dependency patterns of different subspaces are captured, and the outputs of each head are finally concatenated to form a comprehensive temporal feature. For example, the scaling factor is set to the reciprocal of the square root of the input vector dimension. When the input dimension is 64, the scaling factor is set to 1 / 8 to effectively prevent the gradient vanishing due to excessively large dot product results. The normalization function performs an exponential operation on each row and then sums and normalizes the results row by row to ensure that the weight distribution in the time dimension satisfies probabilistic characteristics.

[0141] As a preferred embodiment, the solution of this application is implemented as follows: In the multi-head temporal attention processing, a temporal feature matrix containing displacement data, temperature data, and acceleration changes is input into the multi-head attention layer. Each attention head is configured with an independent linear mapping matrix, where the Query matrix dimension is set to 64×128, the Key matrix dimension is set to 128×128, and the Value matrix dimension is set to 128×64. The similarity score between the Query vector of each time frame and the Key vector of all time frames is calculated by matrix multiplication. The result is divided by a scaling factor √128 and then a row-level softmax function is applied to generate an attention weight matrix. For the third attention head, when calculating the output of the fifth time frame, its attention weights for the second to seventh time frames are 0.18, 0.22, 0.15, 0.12, 0.20, and 0.13, respectively, indicating that this head focuses on capturing feature interactions in the mid-term time period. The outputs of all attention heads are concatenated and linearly combined through a 256-dimensional fully connected layer to finally generate a 256-dimensional comprehensive temporal feature vector for subsequent risk classification.

[0142] Through the above technical solution, this application effectively improves the temporal correlation modeling capability of risk warning. By capturing the dynamic correlation patterns of displacement and temperature characteristics across different time spans through a multi-head attention mechanism, it solves the problem of insufficient sensitivity of traditional static threshold methods for identifying slowly changing abnormal signals. This method can automatically learn the weight distribution of key frames in the time series, eliminate redundant noise interference, and accurately identify abnormal features even in the stage of minor pipeline media leakage or early material fatigue, reducing the misjudgment rate under complex operating conditions.

[0143] In some of the solutions described above in this application, the data processing module processes feature vector data groups through a multi-head temporal attention mechanism to generate comprehensive temporal features. However, when calculating the correlation between different time frames, traditional methods are unable to effectively capture the dynamic interaction relationship of multi-dimensional features in the time series, resulting in an unbalanced distribution of attention weights, which affects the accuracy of comprehensive temporal features and thus reduces the reliability of risk warning.

[0144] This application further proposes that for the h-th attention head, its attention output is calculated through a similarity scoring matrix, and then normalized after being adjusted by a scaling factor. The output results of each attention head are then concatenated to generate a comprehensive temporal feature.

[0145] The similarity scoring matrix is ​​calculated by multiplying the query matrix and the key matrix, with the scaling factor set to the square root of the key vector dimension to control gradient stability. A row-direction softmax normalization function is used to ensure that the sum of attention weights in each time frame is 1. The attention weight matrix is ​​generated by the exponentially normalized result of the vector dot product between time frames, representing the association strength at different time steps.

[0146] Specifically, by mapping the temporal feature matrix into independent Query, Key, and Value vectors, each attention head can learn temporal association patterns in different dimensions. For example, when the acceleration change in displacement data undergoes a sudden change in a certain time frame, the dot product of the corresponding Key vector and the Query vector in subsequent time frames will generate a high similarity score, enhancing the correlation between that time frame and subsequent frames in the attention weight matrix, thus highlighting the impact of the sudden event when weighting the Value matrix. Simultaneously, the scaling factor avoids the gradient saturation problem of the softmax function by reducing the magnitude of the dot product values ​​in high-dimensional space. The independent mapping of each attention head allows the model to capture different feature patterns such as temperature change trends and periodic displacement fluctuations in parallel. The concatenated comprehensive temporal features are further fused with multi-dimensional information through a fully connected layer, ultimately improving the accuracy of risk level classification.

[0147] In some embodiments of this application, the early warning module is based on a fully connected network and a softmax activation function, and performs risk warning based on comprehensive temporal characteristics.

[0148] .

[0149] in, This represents the predicted probability for each risk level. Indicates the comprehensive time series characteristics, This represents the bias vector, and C represents the risk level number, C=4. The early warning module uses the risk level with the highest predicted probability as the output risk level and simultaneously outputs the judgment result, which includes whether there is a risk of leakage and / or whether there is a risk of fatigue damage.

[0150] As a preferred embodiment, the solution of this application is implemented as follows: The early warning module adopts a fully connected neural network architecture with three hidden layers. The input layer receives a 256-dimensional comprehensive temporal feature vector generated by a multi-head temporal attention mechanism. Each hidden layer is configured with 128, 64, and 32 neuron units respectively, and the activation function is ReLU. The output layer has four neuron nodes, corresponding to four levels: no risk, low risk, medium risk, and high risk, respectively, and the probability distribution of each risk category is calculated using the softmax function. During the model training phase, a cross-entropy loss function with L2 regularization is used, the Adam algorithm is selected as the optimizer, and the initial learning rate is set to 0.001. When the system is running, after the comprehensive temporal feature vector is calculated through forward propagation, the risk level corresponding to the maximum probability of the output layer will trigger an early warning signal. At the same time, the risk level is combined with the fatigue damage risk flag and the leakage risk flag to form a status code containing a four-tuple, which is pushed to the operation and maintenance terminal in real time through the cloud platform interface.

[0151] Through the above technical solution, this application effectively solves the technical defects of traditional static threshold methods, such as coarse risk level classification and high false positive rate. By mapping multidimensional temporal features to a nonlinear classification space, the coupling relationship between temperature, displacement, and their dynamic change patterns can be accurately captured, achieving collaborative discrimination of leakage risk and fatigue damage risk. The risk level classification mechanism based on probability output improves the sensitivity of early minor anomaly identification, avoids the problem of missed detection caused by single threshold triggering, and provides quantitative basis for operation and maintenance decisions through risk level gradient output.

[0152] In some of the solutions described above in this application, the cloud-based multi-project management and early warning analysis system for corrugated compensators, after risk assessment, still needs to continuously collect temperature and displacement data even if it is determined that there is no risk of leakage or fatigue damage. The large volume of high-frequency sampled data means that direct transmission and storage would consume significant cloud platform resources, leading to decreased data transmission efficiency and increased storage costs. However, existing solutions do not effectively compress redundant data in a risk-free state, failing to balance the conflict between data integrity and resource consumption.

[0153] This application further proposes a data compression module configured to perform aggregation processing on the collected temperature and displacement data when the data processing module determines that there is no risk of leakage or fatigue damage, generating representative sampling points using a moving average to replace the original high-frequency sampling sequence. When the data processing module determines that there is a risk of leakage or fatigue damage, the collected temperature and displacement data are recorded without compression. The sampled data sequence is then stored after structured encoding, which includes difference encoding, interval statistical encoding, and lightweight temporal encoding.

[0154] The moving average method generates representative sampling points by calculating the mean of data within a preset time window. The window length is dynamically adjusted according to the sampling frequency; for example, raw data is aggregated into one representative point every 5 seconds. Difference encoding reduces data redundancy by recording the changes between adjacent sampling points instead of the original absolute values. Interval statistical encoding extracts the maximum, minimum, and standard deviation within a specified time period, forming compressed statistical features. Lightweight time-series encoding uses a binary bit allocation strategy to map consecutive timestamps to offsets relative to the start time, reducing the storage space for time-series labels.

[0155] Specifically, when the system determines there is no risk, the data compression module initiates moving average processing, replacing the original high-frequency sequence with low-density representative points, achieving a compression ratio of over 80%. The representative points retain overall trend characteristics, avoiding the loss of details. If a risk is identified, the original data is recorded in its entirety to ensure the integrity of the high-frequency information required for subsequent analysis. During structured coding, difference coding converts continuous changes in temperature or displacement into difference sequences, further compressed using Huffman coding. Interval statistical coding divides data into independent intervals every 10 minutes, extracting statistical indicators to replace the original sequence. Lightweight time-series coding uses a fixed bit width to store time offsets, for example, 32-bit integers representing second-level offsets. Through these combined compression methods, the data volume is reduced to 15%-20% of its original size in a risk-free state, reducing the storage and transmission load on the cloud platform while maintaining data resolvability. When the system detects a risk event, the compression module automatically switches to uncompressed mode to ensure that fault characteristics are not smoothed or omitted, providing a complete data foundation for subsequent multi-head time-series attention mechanism analysis.

[0156] As a preferred embodiment, the solution of this application is implemented as follows: The data compression module executes a differentiated data processing flow based on the risk assessment result output by the data processing module. When it is determined that there is no risk of leakage and no risk of fatigue damage, the currently collected temperature and displacement data are aggregated. A sliding window mechanism is used to perform a sliding average calculation with a window length of 30 seconds to generate a representative sampling point sequence containing the mean and extreme values, replacing the original high-frequency sampling data stream of 10 times per second. For the sampled data sequence, a structured encoding method is used for storage. The difference encoding achieves data compression by recording the numerical differences between adjacent sampling points. The interval statistical encoding divides the data within each 5-minute interval into independent intervals and records their mean, maximum, and minimum values. The lightweight time-series encoding uses Delta encoding to record the timestamp difference. When it is determined that there is a risk, the original temperature and displacement data are recorded completely in uncompressed form. At the same time, the difference encoding and interval statistical encoding in the structured encoding are still executed, but the timestamps and numerical sequences of all original data points are retained.

[0157] Through the above technical solution, this application effectively resolves the contradiction between data storage resource consumption and critical information integrity in traditional monitoring systems. When the equipment is operating safely, a dynamic data compression algorithm reduces storage pressure while preserving trend characteristics. Upon detecting potential risks, it automatically switches to full data recording mode to ensure complete time-series information required for subsequent fault analysis. Furthermore, this solution unifies the data storage format through structured coding, enabling compressed and uncompressed data to have the same parsing interface, thus improving data retrieval and playback efficiency.

[0158] In the above embodiments, a sensor module consisting of temperature and displacement sensors is used to collect key physical parameters of the corrugated compensator (such as axial expansion and contraction and soil temperature changes) in real time. A wireless communication module enables remote, high-frequency, and low-power data transmission, effectively adapting to widely distributed urban heating pipe networks. The data processing module constructs a corrugated pipe expansion and contraction displacement sequence based on the sampled data. It obtains the current fatigue state of the corrugated pipe using rainflow counting and the Miner fatigue damage model, and combines temperature data with a dynamic threshold model to identify potential leakage risks. When risk signs are triggered, multi-channel sensor data within a certain time window is automatically retrieved. A multi-head temporal attention mechanism is used to extract key frame change features, generating a fused comprehensive temporal feature representation for accurate identification of subtle trend anomalies or multi-variable interaction correlations. Finally, the early warning module outputs the risk level based on the fused representation, achieving a closed-loop architecture from "physical acquisition - intelligent analysis - proactive early warning." Compared with existing solutions that rely on static thresholds and cannot model evolution trends, this approach improves the ability to identify early fatigue degradation and gradual leakage, reduces false alarm and false negative rates, and achieves unified management of multiple projects and across regions through centralized cloud deployment.

[0159] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program goods. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program goods embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0160] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program goods according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0161] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0162] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0163] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A cloud-based multi-project management and early warning analysis system for corrugated compensators, characterized in that, include: The sensor module includes a temperature sensor and a displacement sensor; A wireless communication module is connected to the sensor module, and the wireless communication module is used to transmit temperature data and / or displacement data; The data processing module is configured to collect the expansion and contraction data of the bellows and obtain the number of fatigue cycles of the bellows, and determine the remaining number of fatigue cycles of the bellows compensator; collect temperature data and compare it with a temperature threshold to determine whether there is a risk of leakage; and when it is determined that there is a risk of leakage or fatigue damage, collect data from all sensors within a preset time period, extract feature vector data groups, process the feature vector data groups based on a multi-head temporal attention mechanism, and determine comprehensive temporal features. The early warning module is configured to provide risk warnings based on the comprehensive time-series characteristics. The sensor module includes: At least four temperature sensors are provided, and the temperature sensors are located on both sides of the external movable end of the corrugated compensator, with at least two temperature sensors provided on each side. At least four displacement sensors are provided, and the displacement sensors are arranged on both sides of the protective bellows inside the bellows compensator; The data processing module collects the expansion and contraction data of the bellows and obtains the number of fatigue cycles of the bellows. When determining the remaining fatigue cycles of the bellows compensator, it includes: Based on a preset sampling period, the expansion and contraction data of the bellows in the bellows compensator are collected to obtain the axial expansion and contraction displacement sequence. The axial expansion and contraction displacement sequence is divided into sliding windows to construct a displacement change dataset within the time window, identify peaks and valleys, and obtain tension and compression cycles. The amplitude and average displacement of each pair of tension-compression half-cycles are extracted based on rainflow counting. The elastic modulus of the bellows and the stress-strain transformation coefficient of the compensator structure are collected to obtain the equivalent stress amplitude. The standard number of fatigue cycles is obtained based on the equivalent stress amplitude and the SN fatigue life curve corresponding to the bellows material. The fatigue wear of identified tension-compression cycles is accumulated based on the Miner linear damage accumulation method, and compared with the standard fatigue cycle count to obtain the remaining fatigue count; After determining the remaining fatigue cycles of the corrugated compensator, the data processing module further includes: The data processing module compares the remaining fatigue counts with the remaining fatigue threshold, and determines whether the corrugated compensator is at risk of fatigue damage based on the comparison results. When the remaining fatigue cycles are less than the remaining fatigue threshold, the data processing module determines that the corrugated compensator is at risk of fatigue damage; when the remaining fatigue cycles are greater than or equal to the remaining fatigue threshold, the data processing module determines that the corrugated compensator is not at risk of fatigue damage. When the data processing module collects temperature data and compares it with a temperature threshold to determine whether there is a risk of leakage, it includes: When the temperature data is less than or equal to the temperature threshold, the data processing module determines that there is no risk of leakage; when the temperature data is greater than the temperature threshold, the data processing module determines that there is a risk of leakage. The data processing module obtains the temperature threshold using the following method: The data processing module collects temperature data for N consecutive sampling cycles and constructs a reference temperature curve for each sensor measuring point to obtain the normal temperature change baseline of the corrugated compensator under leak-free conditions. The soil temperature outside the corrugated compensator is collected, and the baseline of normal temperature change is offset and corrected based on the soil temperature. The correction factor is obtained by linear regression. The standard temperature maximum and minimum values ​​are obtained based on the corrected normal temperature change baseline. The temperature threshold is obtained based on the standard temperature minimum value and the safety margin, wherein the safety margin is 2-5℃.

2. The cloud-based multi-project management and early warning analysis system for corrugated compensators according to claim 1, characterized in that, The data processing module processes the feature vector data set based on a multi-head temporal attention mechanism to determine the comprehensive temporal features, including: The feature vector data set includes displacement data, temperature data, displacement data acceleration change, and temperature data acceleration change. The feature vector data set is used to construct a time-series feature matrix; Configure the number of attention heads, and sequentially perform multi-head temporal attention processing on the temporal feature matrix. Each attention head includes an independent Query, Key, and Value linear mapper. Calculate the attention weight of each time frame in the time series to other frames; The representations of all heads are concatenated and linearly combined to generate a comprehensive temporal feature.

3. The cloud-based multi-project management and early warning analysis system for corrugated compensators according to claim 2, characterized in that, The data processing module processes the feature vector data set based on a multi-head temporal attention mechanism. When determining the comprehensive temporal features, it further includes: For the h-th attention head, its attention output is: ； in, Represents the similarity scoring matrix; Indicates the scaling factor. Represents the Value matrix. Represents the Query matrix. Represents the Key matrix. Represents the normalization function. This represents the attention output of the h-th attention head; The softmax result is the attention weight matrix: ； in, This indicates the degree to which the i-th time frame in the h-th attention head sequence pays attention to the j-th frame when generating its output representation. Represents the Query vector at the i-th time step; This represents the Key vector at the j-th time step; j represents the target time frame number for attention calculation. This represents the index variable used for normalization in the denominator of the softmax function; T represents the length of the time series.

4. The cloud-based multi-project management and early warning analysis system for corrugated compensators according to claim 3, characterized in that, When the early warning module performs risk warning based on the comprehensive time-series characteristics, it includes: The early warning module is based on a fully connected network and a softmax activation function, and performs risk warnings according to the comprehensive temporal characteristics. ； in, This represents the predicted probability for each risk level. Indicates the comprehensive time series characteristics, This represents the weights of a fully connected neural network classifier. This represents the bias vector, and C represents the risk level number, C=4; The early warning module outputs the risk level with the highest predicted probability as the risk level and simultaneously outputs the judgment result, which includes whether there is a risk of leakage and / or whether there is a risk of fatigue damage.

5. The cloud-based multi-project management and early warning analysis system for corrugated compensators according to claim 1, characterized in that, Also includes: The data compression module is configured to perform aggregation processing on the collected temperature and displacement data and generate representative sampling points using a moving average algorithm when the data processing module determines that there is no risk of leakage or fatigue damage, in order to replace the original high-frequency sampling sequence; and to keep the collected temperature and displacement data uncompressed when the data processing module determines that there is a risk of leakage or fatigue damage. The sampled data sequence is stored after being structured and encoded, and the structured encoding includes difference encoding, interval statistical encoding, and lightweight temporal encoding.

Images