CN-122021140-A - Ethylene low-temperature pipeline heat insulation protection state evaluation and control method

CN122021140ACN 122021140 ACN122021140 ACN 122021140ACN-122021140-A

Abstract

The invention discloses an ethylene low-temperature pipeline heat insulation protection state evaluation and control method, and relates to the technical field of pipeline heat insulation protection. The invention solves the problems of incomplete monitoring of abnormal events, inaccurate damage evaluation, single risk judgment and passive control response in the prior art. The method is characterized in that abnormal vibration triggers multi-sensor high-frequency diagnosis and infrared thermal image synchronous acquisition to form event full-dimensional data, characteristics are extracted, a digital twin model is utilized to reversely calculate heat insulation layer performance parameters, a multi-dimensional analysis result is fused to generate a comprehensive evaluation report with risk level and confidence level, hierarchical response is executed according to the risk level, and when the risk level reaches a preset level, an operation parameter adjustment or active thermal compensation control instruction is automatically generated based on multi-objective optimization. The method is used for realizing the accurate assessment and active safety control of the adiabatic state of the ethylene low-temperature pipeline.

Inventors

WANG WENJIE
PENG WEI
ZHANG YAOTIAN
MENG DEXIANG

Assignees

中国化学工程第七建设有限公司

Dates

Publication Date: 20260512
Application Date: 20260116

Claims (9)

1. The method for evaluating and controlling the thermal insulation protection state of the ethylene low-temperature pipeline is characterized by comprising the following steps of: S1, judging an abnormal vibration event based on monitored vibration sensor data, when judging that the abnormal vibration event occurs, switching a high-frequency vibration sensor, a strain sensor and a temperature sensor of a trigger event related area to a high-frequency diagnosis acquisition mode, and controlling an infrared thermal imager to acquire infrared thermal images within a preset time period after the event triggering, and forming event full-dimension data together with the high-frequency diagnosis acquisition data; S2, preprocessing the event full-dimension data and extracting features to obtain event feature data, and carrying out reverse thermal engineering calculation by utilizing a digital twin model and combining the event feature data with real-time pipeline working condition data to obtain equivalent heat conductivity distribution and damage probability index of the heat insulation structure; S3, comparing the equivalent thermal conductivity distribution with a dynamic allowable threshold value, and carrying out decision-level fusion on an identification result of event feature data and a damage probability index trend analysis result by using a machine learning model to generate a comprehensive health assessment report containing risk level and assessment confidence; And S4, executing grading response according to the risk level of the comprehensive health evaluation report, and automatically generating and issuing a control strategy based on multi-objective optimization when the risk level reaches a preset action level or an emergency level, wherein the control strategy comprises a pipeline operation parameter adjustment instruction and/or an adiabatic structure intervention instruction with an active thermal compensation function.
2. The method for evaluating and controlling the adiabatic protection state of an ethylene cryogenic pipeline according to claim 1, wherein in step S1, after determining that an abnormal vibration event occurs and before triggering the high-frequency diagnostic acquisition mode, a physical verification process is further performed, the process comprising: S11, extracting strain data of a strain sensor and temperature data of a temperature sensor in a preset window before and after an abnormal event time stamp; S12, performing trend fitting on the strain data after the event, and calculating a determination coefficient R 2 of the strain data and the pre-event baseline trend, wherein if R 2 is lower than a preset threshold X, the strain change is judged to be inconsistent with the baseline trend, and the strain verification is not passed; S13, calculating a theoretical minimum temperature rise delta T min according to a formula delta T min =E damage /(m.c), wherein E damage is a damage energy threshold value calibrated according to an impact test of a heat insulation layer material, m and c are respectively the mass and specific heat capacity of a heat insulation structure of a related area, and if the actually measured temperature rise of a temperature sensor in a characteristic time window tau after an event does not reach Y% of delta T min , the temperature verification is failed; And S14, executing the triggering high-frequency diagnosis acquisition mode and the subsequent steps only when the strain verification and the temperature verification are passed, and marking the event as instantaneous interference and only storing locally if any verification is not passed.
3. The method for evaluating and controlling the adiabatic protection state of an ethylene cryogenic pipeline according to claim 1, wherein the abnormal vibration event determination in step S1 is required to satisfy the following conditions simultaneously: the first condition is that the monitored vibration sensor data is subjected to wavelet packet transformation, and the probability of exceeding the historical baseline statistical control upper limit in N continuous sampling periods is more than 95% on energy in at least three characteristic frequency bands corresponding to the pipeline heat insulation layer-supporting structure coupling vibration modes; The second condition is that the peak value of the time domain energy envelope of the monitored vibration sensor data and the integrated energy in the side band range formed by the frequency multiplication of the natural frequency of the pipeline after the fast Fourier transformation are exceeded, and the rolling average value calculated based on the recent historical data is added with 3 times sigma.
4. The method for evaluating and controlling the adiabatic protection state of an ethylene cryogenic pipeline according to claim 1, wherein the feature extraction in the step S2 is realized by a physical constraint encoder, the encoder trains a loss function L to be L= |T rec -T real || 2 +λ×||L D (T rec )|| 2 , wherein T rec is reconstruction temperature data of the self-encoder, T real is measured temperature data, L D (S) is a differential operator constraint term constructed according to a pipeline heat conduction physical equation, and lambda is a super parameter for balancing two weights.
5. The method for evaluating and controlling the adiabatic protection state of an ethylene cryogenic pipeline according to claim 1, wherein the inverse thermal process calculation in step S2 comprises a dynamic calibration strategy, specifically: carrying out quality scoring on the real-time pipeline working condition data according to the degree that the scoring includes data delay, sequence smoothness and numerical value are in a historical reasonable interval; Selecting an inversion algorithm according to the scoring result, executing the high-precision inversion algorithm to output high-resolution equivalent heat conductivity distribution when the scoring is higher than a high-quality threshold, executing the probability inversion algorithm to output the mean value and the confidence interval of the distribution when the scoring is between the high-quality threshold and a low-quality threshold, and executing the robust inversion algorithm to output the conservative estimated value and the possible range of the heat flow leakage when the scoring is lower than the low-quality threshold.
6. The method for evaluating and controlling the adiabatic protection state of an ethylene cryogenic pipeline according to claim 1, wherein the decision-level fusion in the step S3 is implemented based on the schiff evidence theory, and specifically comprises the following steps: S31, respectively based on an equivalent thermal conductivity distribution and a dynamic allowable threshold comparison result, a recognition result of event feature data by using a machine learning model and a damage probability index trend analysis result, constructing a first basic probability distribution function, a second basic probability distribution function and a third basic probability distribution function, wherein the first basic probability distribution function, the second basic probability distribution function and the third basic probability distribution function all comprise probability values of no damage, slight damage, serious damage and uncertain four propositions; S32, correcting the first basic probability distribution function based on the quality of the infrared thermal image data, correcting the second basic probability distribution function based on the quality of the high-frequency vibration data, and correcting the third basic probability distribution function based on the continuity of the historical data to obtain three corrected basic probability distribution functions; S33, combining the three corrected basic probability components into a comprehensive basic probability distribution by using a synthesis rule in a Mooney-Sheff evidence theory; s34, determining risk levels according to probability values of slight damage and severe damage propositions in comprehensive basic probability distribution, and outputting the highest probability value as evaluation confidence coefficient according to preset mapping rules, wherein the mapping rules are that probability threshold values corresponding to the slight damage propositions and the severe damage propositions are preset, the probability values of the severe damage propositions are mapped into emergency risks when the probability values of the severe damage propositions exceed the corresponding threshold values, the probability values of the severe damage propositions are mapped into action risks when the probability values of the severe damage propositions do not exceed the threshold values but the probability values of the slight damage propositions exceed the corresponding threshold values, and otherwise, the probability values of the severe damage propositions are mapped into observation risks.
7. The method for evaluating and controlling the adiabatic protection state of an ethylene cryogenic pipeline according to claim 1, wherein the step S4 of automatically generating the control strategy based on the multi-objective optimization specifically comprises: s41, invoking a pre-trained thermal engineering, mechanics and process reduced order proxy model, and constructing a multi-objective optimization problem with the aim of minimizing heat flow leakage and the constraint of a pipeline stress safety limit value and a downstream process parameter stability range; s42, solving an optimization problem by adopting an evolutionary algorithm to obtain a pareto optimal strategy set, selecting a final strategy from the pareto optimal strategy set according to the current production priority, simulating the execution of the final strategy by utilizing a reduced-order agent model, and predicting the associated influence on pipeline stress and a process system; S43, the predicted association influence, the corresponding monitoring suggestions and the recommended alleviation measures are explicitly listed in the control strategy.
8. The method for evaluating and controlling the adiabatic protection state of an ethylene cryogenic pipeline according to claim 1, further comprising S5, closed loop verification and system self-optimization, comprising: s51, after the control strategy is executed, collecting feedback data of an event related area, and performing space-time alignment and comparison with historical data when the event is triggered so as to quantitatively calculate the comprehensive efficiency index of the control strategy; S52, integrating all-link data and comprehensive efficiency indexes from monitoring triggering to feedback execution of the event into a structured case, and storing the structured case in a case library; s53, periodically calling new case data in the case library, performing parameter inversion correction on the digital twin model, and performing incremental training on the machine learning model.
9. The method for evaluating and controlling the adiabatic protection state of an ethylene low-temperature pipeline according to claim 8, wherein the comprehensive performance index is calculated by a double difference method, specifically: a) Presetting or dynamically selecting a control area with similar working conditions but without executing a control strategy in the pipeline system; b) Calculating the variation delta event and delta control of key indexes representing the adiabatic state in the same time window before and after executing the control strategy of the event area and the control area respectively; c) And calculating according to a formula eta=delta event -Δ control to obtain the comprehensive performance index eta.

Description

Ethylene low-temperature pipeline heat insulation protection state evaluation and control method Technical Field The invention relates to the technical field of pipeline heat insulation protection. More particularly, the invention relates to an ethylene cryogenic pipeline insulation protection state evaluation and control method. Background In the field of low-temperature medium conveying such as ethylene, the integrity of heat insulation protection of a pipeline is important for maintaining medium parameter stability, preventing cold loss and basic frost heaving and guaranteeing safe and economic operation of a system. Therefore, real-time and accurate assessment and active control of the adiabatic state of a pipeline are long-standing technical demands. At present, the field mainly depends on periodic manual inspection and fixed-point low-frequency on-line monitoring. The manual inspection has blind areas and interval periods, and sudden damage is difficult to discover in time. Although the conventional online monitoring can provide continuous data, a fixed sampling strategy is generally adopted, and high-fidelity capturing cannot be carried out on transient events such as abnormal vibration and the like, so that event characteristic information is lost. This limitation of data acquisition makes early warning and accurate diagnosis difficult, since abnormal vibrations are often a cause or precursor of damage inside the insulating layer. On the evaluation level, the existing method is mainly used for forward calculation based on stable heat transfer assumption and sparse temperature measuring points, and the overall heat preservation effect is evaluated. When the heat insulating layer is damaged locally and unevenly due to impact, aging and the like, the method is difficult to reversely invert the thermal performance (such as equivalent heat conductivity) spatial distribution of the damaged area, so that the judgment on the damage positioning, range and severity degree is fuzzy, and a quantitative basis cannot be provided for accurate maintenance. In decision and control level, the existing system depends on the threshold value of single parameter (such as wall temperature) to alarm, each parameter information is isolated, fusion analysis is lacking, false alarm or missing alarm is easy to generate, and the confidence of risk assessment is not high. The corresponding control response is also often passive and single, either only an alarm is triggered, or a full line shutdown check is required, and an intervention means (such as flexible adjustment of operating parameters or local thermal compensation) which is classified, active and optimized according to the real-time risk level is lacking, so that it is difficult to implement accurate maintenance while ensuring continuous production. In summary, the prior art has shortcomings in dealing with the fine diagnosis of sudden events, quantitative evaluation of heterogeneous damage, fusion decision of multi-source information, hierarchical optimization control and the like, and an integrated solution capable of realizing accurate perception, quantitative evaluation, credible decision and active control is needed. Disclosure of Invention It is an object of the present invention to solve at least the above problems and to provide at least the advantages to be described later. The invention also aims to provide an ethylene low-temperature pipeline heat insulation protection state evaluation and control method, which solves the problems that the existing ethylene low-temperature pipeline heat insulation protection state evaluation method is incomplete in abnormal event data acquisition, difficult to quantitatively evaluate non-uniform damage of a heat insulation layer, insufficient in risk evaluation reliability due to dependence on a single-dimensional threshold value and single in control response strategy. To achieve these objects and other advantages and in accordance with the purpose of the invention, there is provided a method for evaluating and controlling an adiabatic protection state of an ethylene cryogenic pipeline, comprising the steps of: S1, judging an abnormal vibration event based on monitored vibration sensor data, when judging that the abnormal vibration event occurs, switching a high-frequency vibration sensor, a strain sensor and a temperature sensor of a trigger event related area to a high-frequency diagnosis acquisition mode, and controlling an infrared thermal imager to acquire infrared thermal images within a preset time period after the event triggering, and forming event full-dimension data together with the high-frequency diagnosis acquisition data; S2, preprocessing the event full-dimension data and extracting features to obtain event feature data, and carrying out reverse thermal engineering calculation by utilizing a digital twin model and combining the event feature data with real-time pipeline working condition data to obtain equiva