Intelligent fault diagnosis method and system for heat exchange equipment based on digital twinning

Abstract

The invention discloses a heat exchange equipment intelligent fault diagnosis method and system based on digital twinning, relates to the technical field of digital twinning intelligent diagnosis, and is used for solving the problem that concurrent valve jamming and multi-equipment coupling faults are difficult to quickly and accurately position and safely dispose. Design configuration, maintenance history and real-time working condition data are written into a twin library through semantic mapping and unified time synchronization, and virtual-real synchronization is kept. Secondly, fusing the flow, the pressure difference, the temperature sequence and the color-infrared image, generating a health vector by means of a dimension reduction network, and performing real-time early warning; and after early warning occurs, a twin copy is cloned by using the combination of the abnormal equipment and the regulating valve, parallel simulation is performed on the jam disturbance of the injection valve, and a concurrent fault is locked through comprehensive similarity. Finally, a knowledge base is called to evaluate candidate disposal strategies, an optimal scheme is selected to be issued and executed, dynamic correction is monitored through residual errors, and a diagnosis-decision-execution closed loop is achieved. According to the method, the equipment reliability is improved, and the energy utilization rate is increased.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of digital twin intelligent diagnosis, and more particularly, to a heat exchange equipment intelligent fault diagnosis method and system based on digital twin.

[0002] The tube heat exchanger, reboiler and heat tracing pump in the petrochemical device long-term bear steam-material coupling high load, and valve wear, fouling and throttling fluctuation easily cause heat exchange efficiency to attenuate. The first-line operation depends on the "pressure difference + temperature limit" static alarm and manual troubleshooting, which cannot distinguish between system energy supply wave and single degradation, and is prone to false positives and missed reports. Data-driven residual monitoring lacks physical constraints and is prone to inaccuracy in the face of flow-heat oscillation; offline CFD simulation requires a large amount of calculation and can only be used for post-analysis, which is difficult to meet the hourly diagnosis window. The mother pipe sharing leads to high coupling of the equipment, and the traditional single algorithm directly fails in the scene of increasing correlation; although the multivariate statistical method can capture the cooperative behavior, it is difficult to provide accurate root cause and disposal scheme. Standards such as API682, TEMA and IEC61511 focus on hardware redundancy and lack a fast identification method for concurrent valve-equipment cooperative failure. Although OPC-UA and edge AI have been deployed on site, there is a lack of unified semantic mapping and cross-source timing mechanism, and model drift is serious; the cloud-side large model is also limited by bandwidth and delay.

[0003] In view of the above problems, the present application provides a solution. SUMMARY

[0004] In order to overcome the above-mentioned defects of the prior art, the embodiments of the present application provide a heat exchange equipment intelligent fault diagnosis method and system based on digital twin to solve the problems raised in the background art.

[0005] To achieve the above object, the present application provides the following technical scheme:

[0006] In a preferred embodiment, it comprises:

[0007] Collecting design configuration, maintenance history and real-time working condition three-source data, writing into the digital twin library after unifying fields and time scales, and keeping virtual and real state full-time synchronization by parallel coupling calculation;

[0008] Aligning and fusing the flow-pressure difference-temperature sequence and color-infrared image, sending the cleaned anomaly into the dimension reduction network, and outputting a single health vector and triggering an early warning mark in real time;

[0009] Cloning twin copies based on abnormal equipment set and regulating valve combination, solving in parallel by batch injecting stickiness disturbance, and quickly locking the most consistent fault combination by matching the on-site curve;

[0010] Call knowledge base to generate multiple sets of treatment strategies, select the optimal solution after evaluating the pros and cons in the twin environment, and issue it for execution, and use residual monitoring to iteratively correct to achieve a data-model-decision closed loop.

[0011] In a preferred embodiment, a digital twin model is constructed, a shared digital twin library is established, and the design configuration file, maintenance history, and real-time operating curve of the equipment are fused, and then the physical object state is virtually mapped.

[0012] In a preferred embodiment, the overall perception monitoring of the equipment operating state converts the fused equipment data in the digital twin model into a unique health vector.

[0013] In a preferred embodiment, the health vector sequence of the continuous time window is retrieved from the digital twin database and dimensionally reduced, and then the equipment degradation state is evaluated, the health condition is quantified, and abnormal detection and warning are realized.

[0014] In a preferred embodiment, when the warning occurs, the coupling strength is determined by retrieving the equipment operating parameters, the coupling information is output, and then the possible fault propagation chain is traced, which is recorded as the implicit feature;

[0015] The coupling information and the implicit feature are combined and input into the diagnosis engine for fault diagnosis.

[0016] In a preferred embodiment, during the steady-state operation phase of the high-pressure steam main pipe heat exchange network, the flow rate, shell side pressure difference, and outlet temperature data are sampled, then all the original sampling frames are rewritten and mapped to a unified sampling window, a three-component tensor is generated and preprocessed to obtain a standardized sequence, and then the synchronous amplitude of each device and the correlation coefficient between each other are solved.

[0017] The synchronization amplitude threshold and the correlation threshold are calculated, the new sequence in the unified sampling window is compared with the threshold, and the energy supply resonance synchronization abnormal set is determined.

[0018] In a preferred embodiment, a concurrent fault hypothesis is generated based on the energy supply resonance synchronization abnormal set and the main pipe inlet regulating valve entity, and a virtual twin of the concurrent fault hypothesis is cloned;

[0019] A throttling angle disturbance is applied to the main pipe inlet regulating valve entity, and injected into a three-dimensional multi-field solver, which dynamically subdivides and iterates steps according to local pressure gradients, and performs block sparse solving to control the calculation error below the stability constraint while compressing the total calculation amount.

[0020] The virtual twin outputs the simulation response curve in real time and uploads it to the simulation result buffer, then generates the comprehensive similarity, sorts it in ascending order, sets the coincidence threshold, selects the concurrent valve fault hypothesis with the smallest comprehensive similarity, and locks the corresponding virtual twin for root cause analysis.

[0021] In a preferred embodiment, the locked valve disturbance amplitude, energy loss rate curve and flow-pressure difference coupling stage incremental are combined to fit the output fault severity scalar and assign a fault type label, then a differentiated collaborative treatment scheme is constructed, and the collaborative treatment scheme is optimally screened;

[0022] Calculate other similarity and concurrent valve fault hypothesis key feature coincidence, energy imbalance opening and second order sensitivity of pipe energy supply stability, generate list from high to low, when the comprehensive similarity of the concurrent valve fault hypothesis with the minimum anomaly, automatically select the concurrent valve fault hypothesis with the highest confidence to replace the concurrent valve fault hypothesis with the minimum comprehensive similarity.

[0023] In a preferred embodiment, after root cause confirmation, a candidate treatment sequence is automatically generated, a twin simulation multi-objective optimization is selected to select the optimal scheme, and a safety instruction is compiled and issued for execution, a residual monitor continuously compares the real-time curve with the twin prediction curve, and when the residual exceeds the limit, a backup scheme is switched, the execution result is written back to the knowledge graph and the model self-learning is triggered.

[0024] In a preferred embodiment, it comprises a data synchronization fusion module, a state feature generation module, a concurrent fault simulation screening module and a collaborative treatment closed loop module, and the modules are signal connected;

[0025] The data synchronization fusion module is mainly used for collecting three source data of design configuration, maintenance history and real-time working condition, writing into the digital twin library after unifying fields and time scales, and maintaining virtual and real state full-time synchronization by parallel coupling solution;

[0026] The state feature generation module is mainly used for aligning and fusing the flow-pressure difference-temperature sequence with the color-infrared image, cleaning the anomaly and sending it into the dimension reduction network, and outputting a single health vector in real time and triggering an early warning mark;

[0027] The concurrent fault simulation screening module is mainly used for cloning twin copies based on abnormal equipment set and regulating valve combination, batch injecting sticking disturbance and solving in parallel, and matching the field curve to quickly lock the most consistent fault combination;

[0028] The collaborative treatment closed loop module is mainly used for calling the knowledge base to generate multiple sets of treatment strategies, selecting the optimal scheme after evaluating the pros and cons in the twin environment, issuing for execution, and realizing data-model-decision closed loop by residual monitoring iteration correction.

[0029] The technical effects and advantages of the heat exchange equipment intelligent fault diagnosis method and system based on digital twinning of the application are as follows:

[0030] The application realizes millisecond early warning and minute root cause positioning by constructing a closed-loop system of "data fusion-health vector-concurrent simulation-collaborative treatment". The three-component synchronous amplitude-relevance adaptive threshold greatly reduces the misjudgment of system-level power supply waves; TwinCopy parallel simulation combined with Mahalanobis-DTW composite similarity shortens the concurrent valve fault screening time to one tenth of the traditional method. The multi-objective optimizer automatically selects the treatment scheme with the highest energy recovery rate and the lowest shutdown cost under the premise of ensuring the SIL safety level; the residual monitor can switch the backup strategy online to avoid risk diffusion. Device-level tests show that the false alarm rate is reduced to 0.8%, the total length of diagnosis-treatment is compressed by 60%, the annual steam rate is increased by 3.7%, the maintenance cost is reduced by 28%, and the threshold and rules are continuously optimized relying on incremental learning to realize the self-evolution of fault self-healing ability. BRIEF DESCRIPTION OF DRAWINGS

[0031] Figure 1 A timing diagram of the heat exchange equipment intelligent fault diagnosis method and system based on digital twinning.

[0032] Figure 2 A structure schematic diagram of the heat exchange equipment intelligent fault diagnosis method and system based on digital twinning. DETAILED DESCRIPTION

[0033] The technical solutions in the embodiments of the application will be clearly and completely described below with reference to the drawings in the embodiments of the application. Obviously, the described embodiments are only part of the embodiments of the application, rather than all the embodiments of the application. Based on the embodiments in the application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the application.

[0034] EMBODIMENT

[0035] The application discloses a heat exchange equipment intelligent fault diagnosis method based on digital twinning, as shown in the figure, comprising: Figure 1

[0036] Collecting design configuration, maintenance history and real-time working condition three-source data, writing into a digital twin library after unifying fields and time scales, and keeping virtual and real state full-time synchronization by parallel coupling calculation;

[0037] Aligning and fusing flow-pressure difference-temperature sequences and color-infrared images, sending into a dimension reduction network after cleaning anomalies, and outputting a single health vector and triggering an early warning mark in real time;

[0038] Based on abnormal equipment set and adjusting valve combination cloning twin copies, batch injecting stickiness disturbance and parallel solving, and matching field curves to quickly lock the most consistent fault combination;

[0039] ​The knowledge base is called to generate multiple sets of treatment strategies, and the optimal solution is selected and issued for execution after evaluating the pros and cons in the twin environment, and the residual error is monitored to realize the data-model-decision closed loop.

[0040] Firstly, the design configuration file, maintenance history and real-time working condition curve of the equipment are taken as the only original data source, the fields, units and time scales are accurately aligned through unified semantic mapping rules, and then the three types of data are aggregated into a shared digital twin database; then, with the help of parallel physical field solver, the heat-flow-force coupled equations are solved at fixed period and the database records are updated in real time, so that the virtual object and the physical equipment are kept in fine-grained synchronization in the full time domain.

[0041] Then, the flow, pressure and temperature sequences collected by the field sensors, and the color images and infrared temperature matrix obtained by industrial vision are jointly injected into the digital twin model.

[0042] Specifically, the hardware clock is calibrated to ensure that the three data of sensors, industrial vision and infrared temperature matrix are in the same time window, and then the visible light pixels and thermal image pixels are reprojected to the digital twin coordinate system using the camera calibration results; then the false temperature rise caused by oil stains or shadows is removed by using the temperature difference threshold combined with the texture gradient, and the temperature abnormal area of the equipment surface is directly depicted in the virtual mapping to realize accurate early warning of equipment overheating without false alarm.

[0043] After the space-temperature fusion in the digital twin model, the state feature extraction is performed on the fusion tensor. Specifically, the improved standardization-hyperbolic tangent strategy is used to suppress outlier noise, and then the processed multi-dimensional data is sent to the stack auto-encoding network for automatic dimension reduction.

[0044] After the network converges, the output of the bottleneck layer is extracted as the only health vector, and a continuous feature stream is formed by a sliding time window and an exponential smoother and written back to the digital twin database.

[0045] Further, the stack auto-encoding network SAE is used to retrieve the health vector sequence of the continuous time window in the digital twin database, and the nonlinear dimension reduction is performed in the form of layer-by-layer unsupervised pre-training combined with overall fine-tuning reconstruction. In the training stage, the reconstruction error is taken as the loss index, and each hidden layer is suppressed by weight sparsity constraint to suppress collinear interference, so that the network automatically filters noise and redundancy while retaining key dynamics.

[0046] Since the bottleneck layer vector extracted after convergence not only condenses the joint evolution law of temperature, pressure and flow ternary curve, but also amplifies the micro-drift signal, the early degradation features of the equipment can be captured without manual feature selection.

[0047] Then, the reduced implicit features are input into the support vector data descriptor with Morlet wavelet kernel. The population search strategy of whale optimization algorithm (WOA) is used to perform adaptive optimization in the joint space of penalty coefficient and kernel bandwidth. After iterations, the parameter combination that minimizes the volume of hypersphere and the false alarm rate is automatically locked.

[0048] In the online evaluation stage, the distance of each new feature point to the hypersphere surface is mapped into the device reliability index in real time, and compared with the threshold value updated by the exponential weighted moving average. Once the index is lower than the threshold value, an abnormality warning is triggered.

[0049] When the warning occurs, the digital twin network model is called to record the operating parameters of each device in the same time window, and the correlation coefficient matrix is used to determine the coupling strength. Then, the path search algorithm is used to trace back the possible fault propagation chain, which is recorded as implicit features.

[0050] After that, the implicit features and coupling information are combined and input into the diagnosis engine. The specific diagnosis is as follows:

[0051] If the feature pattern meets the expert rules in the knowledge base, the corresponding fault mechanism and cause are immediately output.

[0052] If the feature pattern does not meet the expert rules in the knowledge base, the off-line trained artificial intelligence model is called to infer and locate the fault source by using the graph attention mechanism.

[0053] It should be noted that in the device heat exchange network sharing the high-pressure steam main pipe, when the inlet regulating valve of the main pipe intermittently sticks and generates a non-steady-state pressure wave in a short time, the steam-side energy supply will simultaneously cause the side flow, shell-side pressure difference, and outlet temperature curve of multiple tube heat exchangers, reboilers, and heat tracing pumps in the same branch to present highly synchronized sawtooth oscillation in the same sampling window, leading to the digital twin network model misjudging the multi-point synchronous oscillation as a single heat exchanger load fluctuation. It is more difficult to distinguish whether it is a system-level energy supply wave caused by valve sticking or a downstream equipment fault, causing alarm omission and incorrect treatment direction. Therefore, in the present embodiment,

[0054] In the steady-state operation stage of the high-pressure steam main pipe heat exchange network, when analyzing the dependency relationship between devices by using the digital twin network model, first, the high-precision Coriolis flowmeters, differential pressure transmitters, and platinum resistance temperature probes pre-deployed on the tube heat exchangers, reboilers, and heat tracing pumps are called to perform millisecond-level cyclic sampling on the flow F(t), shell-side pressure difference ΔP(t), and outlet temperature T(t) in sequence.

[0055] Then, in order to eliminate the influence of cross-device clock drift on synchronization analysis, all original sampling frames are mapped to a unified sampling window W1 after hardware timestamp rewriting by means of the pulse time base T0 output by the main control PLC of the mother pipe, and a three-component tensor Draw is generated in the digital twin monitoring buffer. Subsequently, the three-component tensor Draw is subjected to background noise suppression, valve position opening and closing spike removal, and low drift segmented baseline correction by means of an adaptive band-pass filter chain, to obtain a stabilized normalized sequence Dstd. Then, in a window sliding manner, the synchronization amplitude indicators Ai and the pairwise correlation coefficients ρij of each device are solved in real time.

[0056] The solving results are written into the statistical queue Qsta at the first time, and the mean value μ and the standard deviation σ in the last N healthy operation periods are read by the threshold updater, which dynamically calculates the synchronization amplitude threshold Athr and the correlation threshold ρthr without human intervention, so as to ensure that the thresholds always reflect the latest cooperative fluctuation benchmark of the device group.

[0057] Further, the monitoring logic is used to compare the new sequence in the unified sampling window W1 with the thresholds. When and only when the conditions of Ai≥Athr and ρij≥ρthr are met at the same time, the device subset corresponding to the current sampling window is immediately marked as the energy resonance synchronization abnormal set Ω, and is written into the digital twin database together with the trigger threshold snapshot;

[0058] Then, in order to avoid false alarms triggered by transient noise, a minimum duration constraint is embedded in the marking link. When the energy resonance synchronization abnormal set Ω still remains in the threshold overrun state in the continuous two sampling windows, it enters the next stage of processing.

[0059] Then, based on the physical topology graph maintained by the digital twin platform, the energy resonance synchronization abnormal set Ω is expanded with the mother pipe inlet regulating valve entity Vin in Cartesian, to automatically generate multiple valve sticking-device response concurrent fault hypotheses hk, and to call the twin copy replication instruction TwinCopy to clone a virtual twin Twk for each concurrent fault hypothesis hk in the distributed computing pool Cpool, which is completely consistent with the on-site working condition;

[0060] At the same time, in order to ensure that the virtual twin Twk and the physical body are numerically isomorphic in the thermal, flow, and force three-field coupling relationship, the current transient boundary conditions and the inputs of each driving source are replicated at the same time, so as to fundamentally eliminate the secondary modeling error.

[0061] When the virtual twin Twk is ready, the Vin is subjected to a throttle angle disturbance according to the preset sticking disturbance curve δk, and is injected into the three-dimensional multi-field solver as a time-dependent boundary;

[0062] In the process of parallel computing, the three-dimensional multi-field solver dynamically subdivides the iteration step according to the local pressure gradient with an adaptive step strategy, and solves the heat-flow-force control equation in a block sparse manner through GPU thread bundles, thereby reducing the total amount of calculation while ensuring that the calculation error is lower than the stability constraint εs.

[0063] Further, each virtual twin Twk outputs a simulated response curve Rk(t) in real time, and uploads it to the simulation result buffer Rbuf immediately after the calculation is completed, performs normalized time alignment τ through feature alignment to eliminate the response time lag difference between scenarios, and then generates a comprehensive similarity Sk by combining the weighted Mahalanobis distance DM and the dynamic time warping distance DDTW;

[0064] After the comprehensive similarity calculation is completed, all comprehensive similarities Sk are sorted in ascending order at one time, and the concurrent valve fault hypothesis h with the smallest Sk value and Sk≤Sthr is screened out with the set coincidence threshold Sthr as the limit, and the corresponding virtual twin Tw is locked for root cause analysis process.

[0065] Under the h framework, for each device qj participating in the anomaly, the valve disturbance amplitude δj, the energy loss rate curve Ej(t), and the phase increment ΔF and ΔPj(t) of the flow-pressure difference coupling are extracted from the virtual twin Tw, and a polynomial regression model Rfit with an L2 regularization term is used to jointly fit δj and Ej(t), outputting a fault severity scalar gj, and assigning a fault type label lj according to the valve opening-flow nonlinear response curve matching result.

[0066] All fault severity scalars gj are arranged in descending order to form a fault priority matrix G=[g1,g2,…,gm], and are sent to the knowledge base interface.

[0067] The knowledge base interface retrieves the built-in valve-heat exchanger coupling disposal rule set, comprehensively considers the inter-device thermal load dependence and pipe network energy supply balance, and dynamically constructs a differentiated collaborative disposal scheme candidate set P={P1,P2,…}; then, the twin simulation verifier Simeval is called, and virtual deduction is performed again in the virtual twin Tw for each scheme Pi, and the energy recovery rate ηe, the reliability improvement ΔR, and the disposal time cost τc are used as a three-objective vector fi to perform Pareto optimal screening on the candidate schemes, and finally the scheme P* with the highest comprehensive score is selected and output to the DCS control execution layer, while generating detailed operation instruction sequences and safety check lists.

[0068] Afterwards, to ensure the uncertainty of diagnosis is continuously controlled, the secondary concurrent hypothesis set H' with the top k similarity and Sk > Sthr is saved into the hypothesis monitoring pool, and the key feature fitness Kh', the energy imbalance opening Eh', and the second-order sensitivity of the parent pipe energy supply stability σh' are calculated to synthesize the confidence πh', and a list Π is generated by arranging the confidence πh' from high to low. When P* is executed on site, the residual error ε(t) = S(t)new - R(t) is taken as the closed-loop driving quantity, and if ε(t) deviates unacceptably within the set observation window, the hypothesis h' with the highest confidence in Π is automatically selected to replace h to enter the new simulation alignment process, and the threshold pair is updated synchronously, and self-correction is completed in a periodic manner; if ε(t) converges to zero, h and its simulation-real alignment features are written into the twin knowledge graph as incremental samples for subsequent case inference.

[0069] After the most consistent concurrent valve fault combination is confirmed as the root cause through virtual-site alignment, the decision generation engine is immediately called to convert the fault priority matrix and the corresponding label into a unified feature vector, and the feature vector is weighted and scaled according to the severity and then sent to the strategy retrieval;

[0070] Specifically, based on the valve-heat exchange coupling disposal rules previously deposited in the knowledge base, multiple candidate disposal sequences are automatically spliced, and complete safety interlocking and interlocking constraints are bound for each sequence. Subsequently, the candidate sequences are injected into a dedicated twin deduction container, which simultaneously tracks four types of evaluation indexes, including energy recovery rate, reliability improvement amplitude, potential risk threshold, and estimated downtime, under a heat-flow-force accelerated solving mode, and uses a sliding integrator to eliminate the influence of transient noise on the stability of the indexes, and finally generates a multi-dimensional performance image written back to the decision buffer.

[0071] Then, the performance image is subjected to Pareto partitioning and AHP-entropy weight compound sorting in the multi-objective optimizer: the optimizer first excludes any sequence that has a risk threshold exceeding the limit or a downtime that is too long, then scores the remaining sequences according to the energy recovery rate and reliability improvement amplitude, and the sequence with the highest score is established as the optimal disposal scheme, while the remaining high-potential sequences are saved as backup schemes in case of deviation during operation.

[0072] The optimal disposal scheme is subjected to deep analysis by the operation ticket compiler before being issued, which generates an atomized control instruction stream step by step, and embeds interrupt protection logic and rollback paths that meet the safety integrity level requirements in the instruction stream; then, the signed control instruction stream is issued together with the target performance threshold as an execution token to the process control system and the safety instrument system, realizing seamless connection from diagnosis to disposal.

[0073] During the instruction execution, the residual monitor continuously compares the on-site real-time curve with the twin prediction curve, and once the residual exceeds the adaptive convergence threshold, the standby scheme is triggered to re-enter the twin deduction-performance evaluation-optimization selection cycle, ensuring that the treatment process has self-correction ability without causing secondary risks. After the treatment is completed, the diagnostic features, final scheme, actual execution indicators and residual trajectory are recovered in the form of a quadruple and written into the twin knowledge graph; the incremental learner updates the threshold value updater, similarity weight set and treatment rule library online accordingly, forming a closed loop of data, model and decision, and ultimately achieving continuous improvement of fault diagnosis accuracy and collaborative treatment efficiency, while significantly reducing the cost and energy loss of field trial and error.

[0074] The application further provides a heat exchange equipment intelligent fault diagnosis system based on digital twinning, as shown in the accompanying drawings, comprising a data synchronization fusion module, a state feature generation module, a concurrent fault simulation screening module and a collaborative treatment closed loop module, and the modules are signal connected; Figure 2 The data synchronization fusion module is mainly used for collecting design configuration, maintenance history and real-time working condition three-source data, writing into the digital twin library after unifying fields and time scales, and maintaining virtual and real state full-time synchronization through parallel coupling calculation.

[0075] The data synchronization fusion module is mainly used for collecting design configuration, maintenance history and real-time working condition three-source data, writing into the digital twin library after unifying fields and time scales, and maintaining virtual and real state full-time synchronization through parallel coupling calculation.

[0076] The state feature generation module is mainly used for aligning and fusing the flow-pressure difference-temperature sequence and color-infrared image, cleaning the abnormality, and then sending it into the dimension reduction network to output a single health vector in real time and trigger an early warning mark.

[0077] The concurrent fault simulation screening module is mainly used for cloning twin copies based on the abnormal equipment set and the regulating valve combination, batch injecting stickiness disturbance and solving in parallel, matching the on-site curve to quickly lock the most consistent fault combination.

[0078] The collaborative treatment closed loop module is mainly used for calling the knowledge base to generate multiple sets of treatment strategies, selecting the optimal scheme for execution after evaluating the pros and cons in the twin environment, and realizing the data-model-decision closed loop through residual monitoring iteration correction.

[0079] The above formulas are all dimensionless numerical calculations, the formulas are obtained by collecting a large amount of data to simulate the latest real situation, and the preset parameters in the formula are set by the person skilled in the art according to the actual situation.

[0080] The above embodiments can be realized all or partially by software, hardware, firmware or any other combination. When realized by software, the above embodiments can be realized all or partially in the form of a computer program product.

[0081] Those skilled in the art can understand that the modules and algorithm steps of each example described in combination with the embodiments disclosed herein can be realized by electronic hardware or a combination of computer software and electronic hardware. Whether the functions are realized in hardware or software manner depends on the specific application of the technical solution and the constraints of the invention. The skilled person can use different methods to realize the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

[0082] In addition, each functional module in each embodiment of the present application can be integrated in one processing module, or each module can exist physically alone, or two or more modules can be integrated in one module.

[0083] The above is merely specific embodiments of the present application, but the protection scope of the present application is not limited thereto, any person skilled in the art can easily think of changes or replacements within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

[0084] Finally: the above is only the preferred embodiment of the present application, and is not used to limit the present application, any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included in the protection scope of the present application.

Claims

1. A heat exchange equipment intelligent fault diagnosis method based on digital twinning, characterized in that, Comprise: Collecting design configuration, maintenance history and real-time working condition three source data, writing into digital twin library after unifying field and time scale, and keeping virtual and real state full-time synchronization by parallel coupling solution; Aligning and fusing flow-pressure difference-temperature sequence with color-infrared image, sending into dimension reduction network after cleaning abnormality, and outputting single health vector and triggering early warning mark in real time; Based on abnormal device set and regulating valve combination cloning twin copy, batch injecting stuck disturbance and solving in parallel, matching field curve to quickly lock the most consistent fault combination; Calling knowledge base to generate multiple sets of treatment strategies, selecting the optimal scheme after evaluating the pros and cons in the twin environment, and issuing for execution, and realizing data-model-decision closed loop by residual monitoring and iterative correction.

2. The heat exchange equipment intelligent fault diagnosis method based on digital twinning according to claim 1, characterized in that: Building a digital twin model, establishing a shared digital twin library, and fusing the design configuration file, maintenance history and real-time working condition curve of the equipment, and then virtually mapping the physical object state.

3. The heat exchange equipment intelligent fault diagnosis method based on digital twinning according to claim 2, characterized in that: Comprehensive perception and monitoring of equipment operating state, converting the fused equipment data in the digital twin model into a unique health vector.

4. The heat exchange equipment intelligent fault diagnosis method based on digital twinning according to claim 3, characterized in that; Retrieve the health vector sequence of the continuous time window in the digital twin database and perform dimension reduction processing, then evaluate the degradation state of the equipment, quantify the health status and realize abnormal detection and early warning.

5. The heat exchange equipment intelligent fault diagnosis method based on digital twinning according to claim 4, characterized in that: When the early warning occurs, retrieve the equipment operating parameters to determine the coupling strength, output the coupling information, and then trace back the possible fault propagation chain, which is recorded as the implicit feature; Merge the coupling information and implicit features to input the diagnostic engine for fault diagnosis.

6. The heat exchange equipment intelligent fault diagnosis method based on digital twinning according to claim 5, characterized in that: In the steady-state operation stage of the high-pressure steam main pipe heat exchange network, sample the flow, shell side pressure difference and outlet temperature data, then rewrite all the original sampling frames and map them to a unified sampling window, generate a three-component tensor and pre-process to get a standardized sequence, then solve the synchronization amplitude of each device and the correlation coefficient of each other; Calculate the synchronization amplitude threshold and correlation threshold, compare the new sequence in the unified sampling window with the threshold, and determine the energy supply resonance synchronization abnormal set.

7. The heat exchange equipment intelligent fault diagnosis method based on digital twinning according to claim 6, characterized in that: Based on the energy supply resonance synchronization abnormal set and the main pipe inlet regulating valve entity, generate concurrent fault hypotheses and clone virtual twins of concurrent fault hypotheses; Apply throttle angle disturbance to the main pipe inlet regulating valve entity, inject it into the three-dimensional multi-field solver, which dynamically subdivides and iterates steps according to local pressure gradient, and performs block sparse solving to control the calculation error below the stability constraint while compressing the total calculation amount; The virtual twin outputs the simulation response curve in real time and uploads it to the simulation result buffer, then generates the comprehensive similarity and performs ascending sorting, sets the fitting threshold, selects the concurrent valve fault hypothesis with the smallest comprehensive similarity, and locks the corresponding virtual twin for root cause analysis.

8. The heat exchange equipment intelligent fault diagnosis method based on digital twinning according to claim 7, characterized in that; Extract the locked valve disturbance amplitude, energy loss rate curve and flow-pressure difference coupling phase increment, jointly fit the fault severity scalar, assign the fault type label, then build a differentiated collaborative treatment scheme, and select the optimal collaborative treatment scheme. The calculation of other similarity degrees and the coincidence of concurrent valve fault hypotheses, energy imbalance opening, and the second-order sensitivity of the stability of the parent pipe energy supply, and the synthesis of confidence, arranged from high to low to generate a list, when the comprehensive similarity of the concurrent valve fault hypothesis with the minimum anomaly, automatically select the highest confidence concurrent valve fault hypothesis to replace the concurrent valve fault hypothesis with the minimum comprehensive similarity.

9. The heat exchange equipment intelligent fault diagnosis method based on digital twinning according to claim 8, characterized in that: After root cause identification, candidate treatment sequences are automatically generated, twin simulation multi-objective optimization selects the optimal scheme, and safety instructions are compiled and issued for execution. The residual monitor continuously compares the real-time curve with the twin prediction curve. When the residual exceeds the limit, the standby scheme is switched, the execution result is written back to the knowledge graph, and the model self-learning is triggered.

10. A heat exchange equipment intelligent fault diagnosis system based on digital twinning, characterized in that, It includes: Data synchronization fusion module, state feature generation module, concurrent fault simulation screening module, and collaborative treatment closed loop module, signal connection between modules; The data synchronization fusion module is mainly used to collect design configuration, maintenance history and real-time working condition three source data, write into digital twin library after unified field and time scale, and use parallel coupling solution to keep virtual and real state full-time synchronization; The state feature generation module is mainly used to align and fuse the flow-pressure difference-temperature sequence and color-infrared image, clean the abnormality, and send it to the dimension reduction network. A single health vector is output in real time and the warning mark is triggered; The concurrent fault simulation screening module is mainly used to clone twin copies based on abnormal equipment set and regulating valve combination, batch inject stickiness disturbance and solve in parallel, match the field curve to quickly lock the most consistent fault combination; The collaborative treatment closed loop module is mainly used to call the knowledge base to generate multiple sets of treatment strategies, evaluate the pros and cons in the twin environment, select the optimal scheme, issue for execution, and use residual monitoring to realize data-model-decision closed loop.

Images