Mechanical and electrical system debugging method and system based on digital twinning and artificial intelligence algorithm

Abstract

The application provides a kind of electromechanical system debugging method and system based on digital twinning and artificial intelligence algorithm, to deeply integrate digital twinning and artificial intelligence technology, build full-process intelligent debugging management platform, realize the automation and intelligentization of preparation confirmation, state tracking, parameter comparison, progress control, change analysis and problem diagnosis, so as to improve the efficiency, quality and management level of electromechanical system debugging, involving the cross technology of wisdom construction and industrial automation, including: collecting the preparation work credentials of each professional debugging of electromechanical system; based on the simulation verification result, the debugging process is automatically optimized, and the optimal parameter calibration strategy is generated; field operation personnel receive matching debugging tasks through unique identity identification, and real-time upload field debugging data to the system; a multi-dimensional benchmarking system is established, which includes the design standard value of the electromechanical system, the field measured value and the historical optimal value; match and push the corresponding correction measure suggestion from the strategy library.

Description

Technical Field

[0001] This invention relates to the intersection of smart construction and industrial automation technologies, specifically a method and system for debugging electromechanical systems based on digital twins and artificial intelligence algorithms. Background Technology

[0002] In large and complex electromechanical engineering projects, commissioning is a crucial step in ensuring the normal operation of the system. Electromechanical system commissioning refers to the process of systematically testing, adjusting, and optimizing each piece of equipment and system to meet design requirements and achieve optimal operating conditions after installation and before formal operation. This process involves multiple disciplines, including HVAC, water supply and drainage, power supply and distribution, and automatic control, requiring close collaboration among these professional units.

[0003] Traditional electromechanical commissioning methods rely heavily on manual experience and paper records, which have several technical limitations. During the pre-commissioning preparation phase, the completion status of preparation work by various professional units is typically confirmed through manual reports and paper documents, resulting in information delays and a lack of real-time information sharing mechanisms, leading to low collaborative efficiency in preparation work. During commissioning, commissioning personnel mainly record test data manually, lacking real-time data acquisition and analysis methods. Parameter deviations are often only discovered in the later stages of commissioning, making problem localization difficult and time-consuming. In terms of schedule management, traditional methods struggle to quantify commissioning progress, relying primarily on experience-based judgment and lacking scientific early warning mechanisms, resulting in delayed detection of schedule deviations. When design changes occur, the impact of these changes on commissioning parameters is difficult to quickly identify, leading to slow response times for parameter adjustments and a high risk of rework. Furthermore, the experience and solutions accumulated during commissioning rely mainly on personal records, lacking systematic knowledge management, causing similar problems to recur in different projects and requiring lengthy training periods for new employees.

[0004] In recent years, digital twin technology has seen some application in the field of architectural engineering, mainly focusing on 3D visualization during the design phase and progress simulation during the construction phase. However, the application of digital twin technology in the commissioning phase of electromechanical systems remains relatively limited. Existing technologies primarily use digital twin models for static display, lacking deep integration with commissioning management processes and failing to fully leverage the advantages of digital twin technology in virtual pre-simulation, real-time interaction, and parameter optimization. Meanwhile, the application of artificial intelligence technology in engineering project management is still in the exploratory stage, and its level of intelligence in areas such as commissioning problem diagnosis, progress prediction, and decision support needs improvement.

[0005] Therefore, a new technical solution is needed to deeply integrate digital twin technology with artificial intelligence algorithms to build an intelligent management system covering the entire debugging process. This system can achieve systematic confirmation of pre-debugging preparations, real-time tracking of the debugging process status, intelligent comparison and analysis of debugging parameters, quantitative control of debugging progress, automatic analysis of the impact of design changes, and intelligent diagnosis of debugging problems, thereby improving the efficiency, quality, and management level of electromechanical system debugging. Summary of the Invention

[0006] The technical problem to be solved by this invention is to provide a method and system for debugging electromechanical systems based on digital twin and artificial intelligence algorithms. By deeply integrating digital twin and artificial intelligence technologies, a fully intelligent debugging management platform is built to realize the automation and intelligence of preparation confirmation, status tracking, parameter comparison, progress control, change analysis and problem diagnosis, thereby improving the efficiency, quality and management level of electromechanical system debugging.

[0007] The technical solution adopted by this invention to solve its technical problem is: a debugging method for electromechanical systems based on digital twins and artificial intelligence algorithms, comprising the following steps: S1. Collect the pre-debugging preparation documents of each discipline of the electromechanical system and complete the data storage based on the preset standardized checklist library; calculate the comprehensive readiness index based on the multi-dimensional data of checklist completion, equipment status, and personnel qualifications; when the comprehensive readiness index reaches the preset threshold, the commissioning can be started; otherwise, the commissioning is prohibited and the unqualified items are output. S2. After the startup conditions are met, load the electromechanical system geometric model and physical parameter model into the digital twin platform, import the debugging scheme and simulate the entire debugging process; verify the feasibility of the debugging scheme through simulation, identify potential risk points such as collision, overload and out-of-tolerance; automatically optimize the debugging process based on the simulation verification results, and generate the optimal parameter calibration strategy. S3. On-site operators receive matching debugging tasks through a unique identifier and upload on-site debugging data in real time. Based on the uploaded data, a multi-level status management mechanism is established for tasks that are pending, in progress, pending confirmation, have anomalies, and have been completed, and task progress is tracked in real time. When data anomalies or status stagnation anomalies are detected, an anomaly work order is automatically generated and bound to a preset processing time limit. S4. Construct a multi-dimensional benchmarking system that includes the design standard value of the electromechanical system, the measured value on site, and the historical best value; collect the measured value on site in real time, and calculate the deviation between the measured value and the benchmark value through a deviation analysis algorithm; when the deviation exceeds the preset allowable range, automatically trigger a graded early warning and push the early warning information. S5. Based on the multi-dimensional evaluation model of commissioning progress, the comprehensive index of commissioning progress is calculated in real time by combining the planned duration, the number of completed processes, and the number of equipment sets; according to the difference between the comprehensive index of progress and the target value, the corresponding level of progress warning is triggered; after the warning is triggered, the corresponding corrective action suggestions are automatically matched from the strategy library and pushed. S6. When a design change occurs, based on a pre-built change impact propagation network model, a graph traversal algorithm is used to automatically identify the equipment scope and commissioning procedures affected by the change; for the affected equipment, the calibration values ​​of the commissioning parameters are automatically recalculated, and the corresponding commissioning task requirements and digital twin model parameters are updated synchronously. S7. After receiving the debugging problem report, the problem description text is parsed using natural language processing technology; the most similar historical cases are matched from the debugging problem knowledge base; at the same time, a root cause diagnosis model is constructed based on fault tree analysis; the model is run to obtain the fault tree analysis results, which include a set of possible causes of the problem and their logical relationships; the optimal solution is output by combining the case matching results and the fault tree analysis results. S8 records debugging data, exception handling records, and solution execution effect data throughout the process; it synchronously updates effective problem handling experience and optimal solutions to the debugging problem knowledge base and solution base; based on the updated data, it automatically optimizes the debugging process template and parameter calibration strategy to form a closed-loop iteration.

[0008] Furthermore, the formula for calculating the comprehensive readiness index is as follows: ; in, As the weight of the inspection items, The status is now complete.

[0009] Furthermore, the deviation analysis algorithm is a similarity algorithm, where similarity... ;in, These are measured values. For design values, This refers to the parameter range.

[0010] Furthermore, the multi-dimensional evaluation model for debugging progress is a four-dimensional evaluation model, including task completion rate, quality pass rate, resource input rate, and risk quantity. The comprehensive debugging progress index (CPI) is calculated using the following formula: ; Where T is the task completion rate, Q is the quality pass rate, R is the resource input rate, and R0 is the resource input rate. i Let α represent the number of risks, and β, γ, and δ represent weighting coefficients.

[0011] Furthermore, blockchain technology is used to store the collected pre-debugging preparation documents for each specialty.

[0012] Furthermore, the change impact propagation network model uses a graph traversal algorithm to calculate the propagation path and degree of impact of the change, and trains a machine learning model based on historical change data to predict the impact of the change on debugging quality.

[0013] Furthermore, step S7 also includes: based on matching historical cases and constructing a fault tree analysis model, using a Bayesian network to calculate the probability of each fault cause, and outputting the optimal solution after sorting them by probability from highest to lowest. The formula for calculating the probability of fault causes using the Bayesian network is: ; in, Posterior probability, representing the cause of the failure after observing the current symptom. i The actual probability of occurrence; Conditional probability represents the probability that the cause of the failure is... i When it occurs, the probability of observing the current symptoms; Prior probability, representing the probability of failure based on historical statistical data. i Probability of occurrence when no current symptom information is available; Total probability.

[0014] Furthermore, the optimal parameter calibration strategy includes: The adjustable parameters in the debugging scheme are iteratively optimized using the PID control principle, so that the virtual measured values ​​in the digital twin model gradually approach the design target values. The iterative formula for the PID control principle is as follows: ; in, For the first k The calibration parameter values ​​at the next iteration; For the first k The deviation between the design target value and the virtual measured value during the next iteration; , , These are the proportional, integral, and derivative control coefficients, respectively.

[0015] An electromechanical system debugging system based on digital twin and artificial intelligence algorithms is used to execute the aforementioned electromechanical system debugging method based on digital twin and artificial intelligence algorithms, including: Intelligent confirmation module for data acquisition and adaptation preparation: configured in execution step S1; Digital twin simulation and optimization module: configured in execution step S2 Multi-level state management module: configured in execution step S3; Parameter comparison and early warning module: configured in execution step S4; Progress control module: configured to execute step S5; Change impact analysis module: configured in execution step S6; Intelligent problem diagnosis module: configured to execute step S7; and, Knowledge closed-loop iteration module: configured in execution step S8.

[0016] The beneficial effects of this invention are as follows: This invention realizes intelligent closed-loop management of the entire process of electromechanical system commissioning. Its beneficial effects are reflected in the following aspects: By using a standardized checklist and a mandatory threshold based on the comprehensive readiness index in step S1, commissioning interruptions and rework caused by insufficient preparation are eliminated from the source, significantly improving cross-disciplinary collaboration efficiency; Step S2 utilizes a digital twin environment for full-process simulation verification and risk identification, and automatically optimizes the commissioning process and generates optimal parameter calibration strategies based on the verification results, greatly reducing on-site trial-and-error costs and commissioning cycles; Step S3 establishes a five-level status management mechanism and a closed-loop system for abnormal work orders, making the status of each commissioning task transparent in real time, problems traceable, and responsibilities accountable, completely solving the problems of process loss of control and problem omissions; Step S4 constructs a three-dimensional benchmarking system of design values, measured values, and historical best values, using deviation... The analysis algorithm performs real-time comparison and triggers tiered early warnings, significantly improving the timeliness of parameter deviation detection and preventing operation with defects. Step S5 calculates the comprehensive index (CPI) in real time based on a multi-dimensional progress assessment model and automatically pushes corrective measures, achieving quantitative early warning and proactive intervention to mitigate schedule risks in advance. Step S6 uses a change impact propagation network model and graph traversal algorithm to automatically identify the scope of change impact, recalculate parameters, and synchronously update the twin model, significantly shortening change response time and reducing rework. Step S7 integrates natural language processing, historical case matching, and fault tree analysis to output the optimal solution, improving the efficiency and accuracy of problem diagnosis. Step S8 records data and experience throughout the process and automatically updates the knowledge base, optimizes process templates and calibration strategies, forming a closed-loop iteration that continuously improves experience reuse rate and system intelligence. In summary, this invention comprehensively overcomes the inherent defects of traditional debugging modes, such as difficult preparation and collaboration, opaque processes, uncontrollable progress, slow change response, and lack of knowledge sharing, significantly improving debugging efficiency, quality, and safety. Attached Figure Description

[0017] Figure 1 This is a flowchart of the present invention. Detailed Implementation

[0018] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0019] This invention provides an electromechanical system debugging system based on digital twins and artificial intelligence algorithms. This system is used to execute an electromechanical system debugging method based on digital twins and artificial intelligence algorithms, such as... Figure 1 As shown, the method includes the following steps: S1. Collect the pre-debugging preparation documents of each discipline of the electromechanical system and complete the data storage based on the preset standardized checklist library; calculate the comprehensive readiness index based on the multi-dimensional data of checklist completion, equipment status, and personnel qualifications; when the comprehensive readiness index reaches the preset threshold, the commissioning can be started; otherwise, the commissioning is prohibited and the unqualified items are output. S2. After the startup conditions are met, load the electromechanical system geometric model and physical parameter model into the digital twin platform, import the debugging scheme and simulate the entire debugging process; verify the feasibility of the debugging scheme through simulation, identify potential risk points such as collision, overload and out-of-tolerance; automatically optimize the debugging process based on the simulation verification results, and generate the optimal parameter calibration strategy. S3. On-site operators receive matching debugging tasks through a unique identifier and upload on-site debugging data to the system in real time. Based on the uploaded data, the system establishes a multi-level status management mechanism for tasks, including tasks pending, in progress, pending confirmation, abnormal, and completed, and tracks task progress in real time. When data abnormalities or status stagnation abnormalities are detected, an abnormal work order is automatically generated and bound to a preset processing time limit. S4. Construct a multi-dimensional benchmarking system that includes the design standard value of the electromechanical system, the measured value on site, and the historical best value; collect the measured value on site in real time, and calculate the deviation between the measured value and the benchmark value through a deviation analysis algorithm; when the deviation exceeds the preset allowable range, automatically trigger a graded early warning and push the early warning information. S5. Based on the multi-dimensional evaluation model of commissioning progress, the system calculates the comprehensive index of commissioning progress in real time by combining the planned duration, number of completed processes, and number of equipment sets; and triggers a corresponding level of progress warning based on the difference between the comprehensive progress index and the target value; after the warning is triggered, the system automatically matches and pushes corresponding corrective action suggestions from the strategy library. S6. When a design change occurs, based on a pre-built change impact propagation network model, a graph traversal algorithm is used to automatically identify the equipment scope and commissioning procedures affected by the change; for the affected equipment, the calibration values ​​of the commissioning parameters are automatically recalculated, and the corresponding commissioning task requirements and digital twin model parameters are updated synchronously. S7. After receiving the debugging problem report, the problem description text is parsed using natural language processing technology; the most similar historical cases are matched from the debugging problem knowledge base; at the same time, a root cause diagnosis model is constructed based on fault tree analysis; the model is run to obtain the fault tree analysis results, which include a set of possible causes of the problem and their logical relationships; the optimal solution is output by combining the case matching results and the fault tree analysis results. S8 records debugging data, exception handling records, and solution execution effect data throughout the process; it synchronously updates effective problem handling experience and optimal solutions to the debugging problem knowledge base and solution base; based on the updated data, it automatically optimizes the debugging process template and parameter calibration strategy to form a closed-loop iteration.

[0020] The following is a detailed explanation of each step.

[0021] Detailed explanation of step S1: The standardized checklist library covers all disciplines of electromechanical systems. This library is generally compiled based on national standards, industry standards, and enterprise-level construction process standards. In this embodiment of the invention, it covers eight major categories, totaling 156 inspection items: equipment installation, electrical wiring, pipeline pressure testing, system cleaning, instrument calibration, safety measures, personnel qualifications, and document preparation. To achieve standardized identification and differentiated quantitative evaluation of each preparatory work, and to provide a structured data foundation for the accurate calculation of the comprehensive readiness index, this invention assigns a unique code to each inspection item (e.g., INST-001 represents "water pump installation levelness inspection"), and configures a corresponding weighting coefficient. The weighting coefficient is determined based on the degree of influence of this item on the debugging quality.

[0022] To achieve specialized and access-based mobile management of pre-commissioning preparations, ensuring that each professional unit can only access and process inspection items within its own scope of responsibility and avoiding data confusion and unauthorized operations, this invention allows each professional unit to log in to the system via a dedicated mobile terminal APP and view their respective list of inspection items according to their role permissions. After completing a preparatory task on-site, the operator must select the corresponding inspection item code in the APP and upload the completion status. (1 indicates completion, 0 indicates incomplete), and upload supporting documents (including on-site photos, videos, test report PDF, signed scanned copies, etc.), while filling in metadata such as operator ID, GPS location coordinates, equipment QR code scan results, and completion timestamp. The system automatically verifies the completeness of the documents; incomplete documents will be rejected and require resubmission.

[0023] To ensure the authenticity, immutability, and traceability of the credentials, this embodiment of the invention employs blockchain technology to store the collected credentials for pre-debugging preparation work from various disciplines. Specifically, the mobile app will check the item ID and completion status. Data such as file hash value, operator ID, GPS coordinates, and timestamp are packaged, digitally signed using the private key of the subcontracting unit, and then broadcast to the blockchain network. After the consensus node verifies the signature and permissions, the transaction is packaged and uploaded to the chain. The system returns the transaction hash and block height as proof of evidence.

[0024] Based on the completion status of each inspection item and their weights Calculate the comprehensive readiness index, the formula for which is: ; in, As the weight of the inspection items, The status is now complete.

[0025] In this embodiment of the invention, the comprehensive readiness index The preset threshold is 85%, when When the percentage is ≥85%, the system determines that the debugging start conditions are met, automatically unlocks subsequent debugging tasks, and generates a debugging start license; when When the threshold is less than 85%, the system will prevent debugging from starting and will automatically output a list of non-compliant items, the responsible unit and person, missing voucher prompts and recommended rectification measures, and push a pending notice to the relevant responsible persons until all non-compliant items are completed and the threshold is recalculated.

[0026] Detailed explanation of step S2: After the startup conditions are met, the system loads the geometric model of the electromechanical system (including three-dimensional spatial information of equipment, pipes, valves, instruments, etc.) from the BIM database, and simultaneously loads the corresponding physical parameter models (such as equipment performance curves, fluid dynamics models, thermodynamic models, etc.) from the physical model library. It then imports a pre-prepared commissioning plan and simulates the entire commissioning process in the digital twin environment, including equipment start-up and shutdown sequences, valve adjustment sequences, and parameter measurement points.

[0027] Through simulation, the system automatically verifies the feasibility of the debugging scheme and identifies potential collision risks (such as insufficient operating space), overload risks (such as excessive equipment starting current), and out-of-tolerance risks (such as adjusted parameters exceeding the design range). Based on the simulation verification results, the system automatically optimizes the debugging process, such as adjusting the operation sequence, modifying the adjustment step size, and reallocating test points.

[0028] Specifically, collision risk is identified through a spatial collision detection algorithm; overload risk is judged based on the instantaneous current of each device when it starts up during the simulation; and out-of-tolerance risk is determined by comparing the simulation output value with the design standard value extracted from BIM attributes or design documents and judging it according to the deviation tolerance rules (such as flow deviation not exceeding ±5% and temperature deviation not exceeding ±1℃). If the simulation value exceeds the allowable range, out-of-tolerance risk is triggered.

[0029] In this embodiment of the invention, the operation sequence is adjusted as follows: The system uses simulation data to construct a dependency graph between equipment and processes, where nodes represent operation steps and edges represent time or logical dependencies. The critical path method is used to analyze the total time consumption and bottleneck processes under the current sequence, and a topological sorting algorithm is run to generate all feasible legal sequences. Based on this, the system uses tabu search or a genetic algorithm to iteratively adjust the sequence with the goal of minimizing the total debugging time or minimizing the peak load. For example, if the simulation shows that "starting two chiller units first and then opening the terminal valve" will cause the current to exceed the limit, the system will automatically modify the sequence to "start unit 1 → open valve 1 → start unit 2 → open valve 2", decomposing the serial risk into interleaved parallel operations. Simultaneously, if the simulation shows that two operations do not share resources and are physically isolated, the system will mark them as "parallelizable," automatically splitting the tasks and assigning them to different work groups.

[0030] The adjustment step size is modified as follows: The system records the system response curve after each adjustment action in the simulation and extracts the local sensitivity (i.e., output change / input change). For high-sensitivity ranges (e.g., a sharp increase in flow when the valve opening is 30%–40%), the system uses a small step size (e.g., 1% increment) to avoid overshoot; for low-sensitivity ranges (e.g., a smooth flow when the opening is 70%–100%), a large step size (e.g., 10% increment) is used to quickly approximate the target. In specific implementation, the system runs a variable step size adaptive algorithm: the initial step size is set to a preset value; if the deviation sign is the same after two consecutive adjustments and the absolute value does not decrease, the step size is increased; if the deviation sign alternates (oscillates), the step size is decreased and the waiting time is increased. Furthermore, the system utilizes the PID parameter tuning concept, based on the deviation e(k) and the deviation change rate... (k) Dynamically calculate the next adjustment amount: Δu=f(e, This allows the adjustment process to converge quickly without overshoot.

[0031] Test point reallocation: The system treats each measuring point (such as pressure sensor, thermometer) in the commissioning scheme as a monitoring node. After simulation, the system outputs parameter values ​​and time series for all potential locations (such as pipe nodes and equipment interfaces in the BIM model). Then, the following analysis is performed: Redundancy removal: Calculate the Pearson correlation coefficient for all measurement point pairs. If the correlation coefficient is >0.95, it means that the two measurement points provide almost the same information. The system will automatically remove the one with the lower weight and suggest that the sensor be removed or reused on site.

[0032] Add key measuring points: Calculate the parameter gradient (such as temperature change rate, pressure drop gradient) at each spatial location where no measuring point is set. If the gradient exceeds the threshold, it indicates that the dynamics in the area are significant, and the system recommends adding measuring points. At the same time, principal component analysis is used to identify the spatial coordinates with the highest contribution rates, and measuring points are preferentially placed at these locations.

[0033] Optimize acquisition frequency: For the retained measurement points, analyze their signal spectrum. For high-frequency fluctuation components (such as pressure fluctuations >10Hz), high-frequency acquisition (100Hz) is recommended; for slowly changing signals (such as water temperature), low-frequency acquisition (0.1Hz) is recommended to reduce the amount of data.

[0034] In this invention, various methods such as reinforcement learning and model predictive control (MPC) can be used to iteratively optimize the adjustable parameters in the debugging scheme. In this embodiment, the PID control principle is used to iteratively optimize the adjustable parameters in the debugging scheme, so that the virtual measured values ​​in the digital twin model gradually approach the design target values. The iterative formula for the PID control principle is: ; in, For the first k The calibration parameter values ​​at the next iteration; For the first k The deviation between the design target value and the virtual measured value during the next iteration; , , These are the proportional, integral, and derivative control coefficients, respectively. Through multiple iterations, the system generates the optimal parameter calibration strategy for field implementation.

[0035] The proportional, integral, and derivative control coefficients can be determined through empirical trial-and-error methods, attenuation curve methods, or automatic tuning via digital twin simulation. In this embodiment of the invention, the automatic tuning method using digital twin simulation is preferred: the step response of the controlled object is simulated in a virtual environment, and algorithms such as genetic algorithms or particle swarm optimization are used to search for the proportional coefficient that minimizes the overall performance index (such as the integral of absolute error or the time multiplied by the integral of absolute error). Integral coefficient and differential coefficients The tuned coefficients are then incorporated into the optimal parameter calibration strategy. During actual field debugging, if the system's actual characteristics match the simulation model, no further adjustments are needed; if deviations exist, field personnel can manually correct them according to the system's recommended fine-tuning range. This approach balances the efficiency of offline optimization with the flexibility of online correction.

[0036] Detailed explanation of step S3: On-site personnel can log in to the system by scanning a task QR code via a mobile app or by using a unique identifier such as facial recognition to receive a matching debugging task. Each debugging task is bound to a unique task code, equipment code, debugging procedure instructions, and preset standard parameter range.

[0037] The system establishes a five-level status management mechanism: Pending Start, In Progress, Pending Confirmation, Abnormal, and Completed. After arriving on-site, the operator scans the code to start the task, and the status changes to "In Progress." After completing the debugging operation and uploading the data, the status changes to "Pending Confirmation," which is automatically verified by the system or manually confirmed by quality inspectors. If the data is qualified, the status changes to "Completed." If the data is abnormal or the operator proactively reports a problem, the status changes to "Abnormal."

[0038] When data anomalies are detected (such as measured values ​​deviating significantly from design values) or status stagnation (such as a task remaining in the "in progress" state for more than a preset time limit without updates), the system automatically generates an anomaly work order, binds a processing time limit (such as 2 hours, 24 hours, etc.), and assigns it to the corresponding responsible unit or person. In this invention, anomaly issues employ a "three-review, one-verification" closure mechanism: self-inspection by the responsible unit, review by the supervision unit, approval by the owner unit, and finally, closure only after on-site verification. The entire state transition process is controlled by a finite state machine, defining legal state transition paths and triggering events to ensure the traceability of state changes.

[0039] Detailed explanation of step S4: The system constructs a multi-dimensional benchmarking system that includes design standard values, on-site measured values, and historical best values. Design standard values ​​are derived from BIM models or design drawings; on-site measured values ​​are collected in real time through IoT sensors, handheld instruments, or manual input; and historical best values ​​are extracted from historical commissioning databases to record parameters of similar equipment under optimal operating conditions.

[0040] After real-time acquisition of on-site measured values, the system in this invention uses a similarity algorithm to calculate the deviation between the measured values ​​and the benchmark values ​​(prioritizing comparison with design standard values, and comparing with historical best values ​​if design values ​​are missing): Similarity ; in, These are measured values. For design values, For parameter range. Similarity The closer to 1, the smaller the deviation.

[0041] The system has preset allowable ranges (e.g., similarity ≥ 0.95 is considered acceptable). When the deviation exceeds the allowable range, a tiered warning system is automatically triggered: Level 1 warning (yellow) indicates attention, Level 2 warning (orange) indicates adjustment is needed, and Level 3 warning (red) indicates severe deviation requiring immediate shutdown and inspection. Warning information is sent to relevant personnel via APP push notifications, SMS, and large-screen displays. The system also establishes a parameter deviation pattern library to identify systematic deviations (e.g., all similar equipment is too small), random deviations (irregular fluctuations), and periodic deviations (e.g., periodic changes with temperature), providing a basis for subsequent optimization.

[0042] Step S5 explained in detail: The system is based on a multi-dimensional evaluation model of commissioning progress, combining planned duration, number of completed processes, and number of equipment sets to calculate a comprehensive index of commissioning progress in real time. In this embodiment of the invention, the model is a four-dimensional evaluation model, including task completion rate T, quality pass rate Q, resource input rate R, and risk quantity. The Commissioning Progress Composite Index (CPI) is calculated using the following formula: ; Among them, α, β, γ, and δ are weighting coefficients that can be dynamically adjusted according to the characteristics of the project, and the sum of α, β, γ and δ is 1.

[0043] In this embodiment of the invention: for conventional balanced projects, the values ​​of α, β, γ, and δ are 0.35, 0.35, 0.20, and 0.10, respectively; for schedule-priority projects, the values ​​of α, β, γ, and δ are 0.50, 0.25, 0.15, and 0.10, respectively; for quality-priority projects, the values ​​of α, β, γ, and δ are 0.25, 0.55, 0.10, and 0.10, respectively; and for high-risk, high-complexity projects, the values ​​of α, β, γ, and δ are... The values ​​are 0.30, 0.30, 0.10, and 0.30 respectively; for resource-constrained projects, the values ​​of α, β, γ, and δ are 0.30, 0.30, 0.30, and 0.10 respectively; for the early debugging stage, the values ​​of α, β, γ, and δ are 0.45, 0.30, 0.15, and 0.10 respectively; and for the later closing stage, the values ​​of α, β, γ, and δ are 0.25, 0.45, 0.10, and 0.20 respectively.

[0044] The system triggers corresponding progress warnings based on the difference between the CPI and the target value (usually 1.0): a yellow warning (slight lag) for CPI < 0.9, an orange warning (moderate lag) for CPI < 0.8, and a red warning (severe lag) for CPI < 0.7. After a warning is triggered, the system automatically matches and pushes corrective action suggestions from the strategy library, such as: increasing personnel, extending work time, adjusting the sequence of processes, and prioritizing critical path tasks. All warning records and the effects of corrective actions are recorded for subsequent optimization of the strategy library.

[0045] Detailed explanation of step S6: When design changes occur (such as equipment model replacement, pipeline routing adjustment, or control logic modification), the system first parses the change content, extracting the equipment, parameters, and relationships involved. Then, based on a pre-built change impact propagation network model, a graph traversal algorithm is used to automatically identify the equipment scope and commissioning procedures affected by the change. In this embodiment of the invention, the change impact propagation network model expresses the dependencies between equipment, parameters, and procedures in the form of a directed graph. Breadth-first search is used to automatically identify the equipment scope and commissioning procedures affected by the change, and to calculate the impact propagation path and the degree of impact (such as direct impact, indirect impact, and impact weight).

[0046] For affected equipment, the system automatically recalculates the calibration values ​​of commissioning parameters, such as recalculating setpoints for flow rate, pressure, and temperature based on the performance curves of the new equipment. Simultaneously, the system updates the corresponding commissioning task requirements (e.g., modifying target values ​​in commissioning steps) and digital twin model parameters (e.g., updating equipment attributes in the BIM model and parameters in the physical model). To better provide quantitative evidence for design change approval decisions and commissioning plan adjustments, the system also trains machine learning models (e.g., random forests, gradient boosting trees) based on historical change data to predict the impact of changes on commissioning quality (e.g., the magnitude of the decrease in parameter compliance rate, potential risk level), providing support for decision-making.

[0047] Detailed explanation of step S7: Upon receiving a reported debugging issue, the system first uses Natural Language Processing (NLP) to parse the problem description text, extracting keywords, device names, symptom characteristics, etc. Then, it matches the most similar historical cases from the debugging issue knowledge base, calculates text similarity using cosine similarity or a vector space model, and returns the closest historical case and its solution.

[0048] Simultaneously, the system constructs a root cause diagnosis model based on fault tree analysis. Taking the problem symptom as the top event, it decomposes all possible causes (intermediate events, bottom events) leading to the top event layer by layer, connecting them with logic gates (AND gates, OR gates) to form a fault tree. Running this model yields the fault tree analysis results, including the set of possible causes leading to the problem and their logical relationships (e.g., "loose belt OR substandard water quality OR clogged condenser").

[0049] The system integrates case matching results and fault tree analysis results to output the optimal solution. Specifically, the system uses similar historical cases and their solutions (with similarity scores and success rates) obtained from case matching, and the set of possible causes and their logical relationships (with importance or probability) obtained from fault tree analysis as two information sources. It merges these sources using a weighted scoring or rule-based hierarchical approach: cases with high similarity (e.g., ≥0.85) and high success rates are directly prioritized for recommendation; for cases with medium similarity, the case solutions are output by overlaying them with the fault tree troubleshooting order; if no high-similarity cases exist, the troubleshooting list is output entirely based on the fault tree cause probability ranking. The final optimal solution output includes recommended actions, explanations, and alternative solutions, ensuring that the diagnostic results are supported by both historical experience and logical reasoning, thus improving the accuracy and efficiency of on-site problem solving.

[0050] As a further optimization, based on matching historical cases and constructing a fault tree analysis model, a Bayesian network is used to calculate the probability of each fault cause, and the optimal solution is output after sorting them by probability from highest to lowest. The formula for calculating the probability of fault causes using a Bayesian network is: ; in Let be the posterior probability, representing the cause of the failure after observing the current symptom. i The actual probability of occurrence; Conditional probability represents the probability that the cause of the failure is... i When it occurs, the probability of observing the current symptoms; Prior probability, representing the probability of failure based on historical statistical data. i Probability of occurrence when there is no current symptom information; The probability is set to the total probability and used for normalization. The system sorts the results from highest to lowest posterior probability and recommends prioritizing the investigation of the causes with the highest probability on-site.

[0051] Detailed explanation of step S8: The system records all debugging data (including start time, end time, operator, measured value, deviation, etc. for each task), exception handling records (exception type, handling process, handling time, handling result), and solution execution effect data (whether the problem was solved in one go, whether new problems were generated, etc.). All data is stored in a central database and indexed for quick retrieval.

[0052] Effective problem-solving experience and optimal solutions are synchronously updated in the debugging problem knowledge base and solution library. For example, a case of successfully solving the "high condensing pressure" problem has its symptom description, diagnosis process, measures taken, and final results stored in a structured manner for future matching and use in similar problems.

[0053] Based on the updated data, the system automatically optimizes the debugging process template and parameter calibration strategy. For example, if historical data analysis reveals that the balancing valve calibration of a certain type of equipment is generally too high, the system will automatically adjust the target value or initial parameters in the PID calibration strategy. Similarly, if statistics show that a solution to a certain type of problem has a high success rate, the system will increase the recommendation priority of that solution in the matching results. This forms a closed-loop iteration of "practice—record—optimization—reuse," enabling the system's intelligence level to continuously improve with increased usage.

[0054] This invention, through the synergistic effect of the above eight steps, provides a systematic solution to the core pain points of traditional electromechanical commissioning models, such as difficult preparation and coordination, lack of process transparency, weak progress control, slow change response, and low knowledge reuse: Before commissioning, standardized checklists and blockchain-based evidence storage are used to quantitatively assess and enforce thresholds for preparation work, preventing start-up delays due to insufficient preparation; full-process simulation and parameter self-optimization are performed in a digital twin environment to identify collision, overload, and out-of-tolerance risks in advance and generate optimal calibration strategies, significantly reducing on-site trial-and-error costs; during on-site execution, a five-level closed-loop management system and an abnormal work order mechanism are established to ensure… Each task is transparent and traceable; real-time comparison of multi-dimensional parameters and hierarchical early warning significantly improve the timeliness of deviation detection; quantitative early warning and intelligent correction are achieved based on a four-dimensional progress assessment model and a comprehensive index (CPI), mitigating schedule risks in advance; when design changes occur, graph traversal algorithms are used to automatically identify the scope of impact and update twin parameters accordingly, combined with machine learning to predict the impact of changes on quality, significantly improving response speed; fault diagnosis integrates NLP, case matching, fault trees, and Bayesian probability ranking to output the most likely cause, improving resolution efficiency; the entire process is data-driven, with a closed-loop iteration of the knowledge base, continuously enhancing experience reuse and system intelligence. In summary, this invention achieves intelligent management of the entire process of electromechanical system commissioning, significantly improving efficiency, quality, and safety, and has outstanding industrial practical value.

[0055] In this embodiment of the invention, specifically, the electromechanical system debugging system based on digital twin and artificial intelligence algorithms of the present invention includes: Intelligent confirmation module for data acquisition and adaptation preparation: configured in execution step S1; Digital twin simulation and optimization module: configured in execution step S2 Multi-level state management module: configured in execution step S3; Parameter comparison and early warning module: configured in execution step S4; Progress control module: configured to execute step S5; Change impact analysis module: configured in execution step S6; Intelligent problem diagnosis module: configured to execute step S7; and, Knowledge closed-loop iteration module: configured in execution step S8.

Claims

1. A debugging method for electromechanical systems based on digital twins and artificial intelligence algorithms, characterized in that: Includes the following steps: S1. Collect the pre-commissioning documents of various disciplines of electromechanical system and complete the data storage based on the preset standardized checklist library; Calculate the comprehensive readiness index based on multi-dimensional data including checklist completion, equipment status, and personnel qualifications. When the overall readiness index reaches a preset threshold, debugging is allowed to start; otherwise, debugging is prohibited and unqualified items are output. S2. After the startup conditions are met, load the electromechanical system geometric model and physical parameter model into the digital twin platform, import the debugging scheme and simulate the entire debugging process; verify the feasibility of the debugging scheme through simulation, identify potential risk points such as collision, overload and out-of-tolerance; automatically optimize the debugging process based on the simulation verification results, and generate the optimal parameter calibration strategy. S3. On-site operators receive matching debugging tasks through a unique identifier and upload on-site debugging data in real time. Based on the uploaded data, a multi-level status management mechanism is established for tasks to be started, in progress, pending confirmation, with anomalies, and completed, and task progress is tracked in real time. When data anomalies or status stagnation anomalies are detected, an anomaly work order is automatically generated and bound to a preset processing time limit. S4. Construct a multi-dimensional benchmarking system that includes the design standard values, field measured values, and historical best values ​​of the electromechanical system. Real-time acquisition of on-site measured values; calculation of the deviation between the measured values ​​and the benchmark values ​​using a deviation analysis algorithm; when the deviation exceeds the preset allowable range, automatic triggering of graded early warning and push notification information. S5. Based on the multi-dimensional evaluation model of commissioning progress, the comprehensive index of commissioning progress is calculated in real time by combining the planned construction period, the number of completed processes, and the number of equipment sets; the corresponding level of progress warning is triggered according to the difference between the comprehensive progress index and the target value. Once an alert is triggered, the system will automatically match and push corresponding corrective action recommendations from the policy library. S6. When a design change occurs, based on a pre-built change impact propagation network model, a graph traversal algorithm is used to automatically identify the equipment scope and debugging procedures affected by the change. For affected equipment, the calibration values ​​of debugging parameters are automatically recalculated, and the corresponding debugging task requirements and digital twin model parameters are updated synchronously. S7. After receiving the debugging problem report, use natural language processing technology to parse the problem description text; The system matches the most similar historical cases from the debugging problem knowledge base, and constructs a root cause diagnosis model based on fault tree analysis. The model is then run to obtain the fault tree analysis results, which include a set of possible causes of the problem and their logical relationships. Based on the combined case matching results and fault tree analysis results, the optimal solution is output. S8: Records all debugging data, exception handling records, and solution execution effect data throughout the process; Effective problem-solving experience and optimal solutions will be updated simultaneously to the debugging problem knowledge base and solution base; Based on the updated data, the debugging process template and parameter calibration strategy are automatically optimized to form a closed-loop iteration.

2. The electromechanical system debugging method based on digital twin and artificial intelligence algorithms as described in claim 1, characterized in that, The formula for calculating the comprehensive readiness index is as follows: ； in, As the weight of the inspection items, The status is now complete.

3. The electromechanical system debugging method based on digital twin and artificial intelligence algorithms as described in claim 1, characterized in that, The deviation analysis algorithm is a similarity algorithm, and the similarity... ;in, These are measured values. For design values, This refers to the parameter range.

4. The electromechanical system debugging method based on digital twin and artificial intelligence algorithms as described in claim 2, characterized in that, The multi-dimensional evaluation model for debugging progress is a four-dimensional evaluation model, including task completion rate, quality pass rate, resource input rate, and risk quantity. The comprehensive debugging progress index (CPI) is calculated using the following formula: ; Where T is the task completion rate, Q is the quality pass rate, R is the resource input rate, and R0 is the resource input rate. i Let α represent the number of risks, and β, γ, and δ represent weighting coefficients.

5. The electromechanical system debugging method based on digital twin and artificial intelligence algorithms as described in claim 1, characterized in that, Blockchain technology is used to store the collected documents for pre-debugging preparation work in various disciplines.

6. The electromechanical system debugging method based on digital twin and artificial intelligence algorithms as described in claim 1, characterized in that, The change impact propagation network model uses a graph traversal algorithm to calculate the propagation path and degree of impact of the change, and trains a machine learning model based on historical change data to predict the impact of the change on debugging quality.

7. The electromechanical system debugging method based on digital twin and artificial intelligence algorithms as described in claim 1, characterized in that, Step S7 further includes: based on matching historical cases and constructing a fault tree analysis model, using a Bayesian network to calculate the probability of each fault cause, and outputting the optimal solution after sorting them by probability from highest to lowest. The formula for calculating the probability of fault causes using the Bayesian network is as follows: ; in, Posterior probability, representing the cause of the failure after observing the current symptom. i The probability of it actually happening; Conditional probability represents the probability that the cause of the failure is... i When it occurs, the probability of observing the current symptoms; Prior probability, representing the probability of failure based on historical statistical data. i Probability of occurrence when no current symptom information is available; Total probability.

8. The electromechanical system debugging method based on digital twin and artificial intelligence algorithms as described in claim 1, characterized in that, The strategy for generating optimal parameter calibration includes: The adjustable parameters in the debugging scheme are iteratively optimized using the PID control principle, so that the virtual measured values ​​in the digital twin model gradually approach the design target values. The iterative formula for the PID control principle is as follows: ; in, For the first k The calibration parameter values ​​at the next iteration; For the first k The deviation between the design target value and the virtual measured value during the next iteration; , , These are the proportional, integral, and derivative control coefficients, respectively.

9. A debugging system for electromechanical systems based on digital twins and artificial intelligence algorithms, used to execute the debugging method for electromechanical systems based on digital twins and artificial intelligence algorithms as described in any one of claims 1 to 8, characterized in that, include: Intelligent confirmation module for data acquisition and adjustment preparation: configured in execution step S1; Digital twin simulation and optimization module: configured in execution step S2 Multi-level state management module: configured in execution step S3; Parameter comparison and early warning module: configured in execution step S4; Progress control module: configured to execute step S5; Change impact analysis module: configured in execution step S6; Intelligent problem diagnosis module: configured to execute step S7; and, Knowledge closed-loop iteration module: configured in execution step S8.

Images