CN-121997760-A - Method, system, equipment and medium for generating detection scheme of special equipment

CN121997760ACN 121997760 ACN121997760 ACN 121997760ACN-121997760-A

Abstract

The invention provides a method, a system, equipment and a medium for generating a detection scheme of special equipment, which belong to the technical field of special equipment and artificial intelligence application, and the method comprises the steps of collecting data of target equipment to construct a data file, respectively extracting time sequence, image and text characteristics by utilizing LSTM, CNN and a transducer encoder, and fusing the characteristics to realize intelligent evaluation of the equipment; based on the evaluation result, combining rule constraint and historical case recommendation to generate an initial test scheme, then carrying out simulation prediction on the scheme through a digital twin environment, carrying out multi-objective optimization by adopting reinforcement learning to obtain an optimal test scheme with balanced safety, economy and efficiency, finally outputting the scheme, and carrying out feedback update on an evaluation model according to the actual test result to form a closed-loop learning mechanism. The invention realizes the conversion of the inspection scheme from experience driving to data intelligent driving, and improves the capability of personalized formulation, prospective optimization and continuous evolution.

Inventors

ZHANG TAO
LIU JUNBO
LIAO DONGMING
XIAO GUOFENG
HE SHUISHENG
WU XIAOMING

Assignees

江西省检验检测认证总院特种设备检验检测研究院

Dates

Publication Date: 20260508
Application Date: 20260129

Claims (10)

1. The method for generating the detection scheme of the special equipment is characterized by comprising the following steps of: Collecting inherent set attributes, dynamic operation data, historical diagnosis reports and regulation standard data of target special equipment, wherein the dynamic operation data comprises temperature and pressure curves of all parts in equipment operation, and ultrasonic images and radiographic images of all parts of the equipment; Extracting time sequence change characteristics of temperature and pressure curves of all parts, extracting defect morphological characteristics of ultrasonic images and radiographic images of all parts, extracting text semantic characteristics of historical diagnosis reports of all parts, and carrying out dynamic weighted fusion on the time sequence change characteristics, the defect morphological characteristics and the text semantic characteristics to form a unified equipment state characteristic vector; Based on the security assessment score, inherent set attribute and constraint conditions of rule standard data, generating an initial detection scheme comprising a detection project, a recommended detection method, a key detection part and a recommended detection period of each component; Performing simulation execution on an initial detection scheme in a preset digital twin system, constructing a reward function by taking safety, economy and efficiency as comprehensive optimization targets by adopting a reinforcement learning algorithm, solving parameters of the initial detection scheme, and outputting an optimal detection scheme which meets the minimum detection cost and the minimum detection duration.
2. The method for generating the detection scheme of the special equipment according to claim 1 is characterized in that a reinforcement learning algorithm is adopted, a bonus function is built by taking safety, economy and efficiency as comprehensive optimization targets, and parameters of an initial detection scheme are solved, specifically, the reinforcement learning algorithm takes safety maximization, detection cost minimization and detection duration minimization as core design bonus functions, detection scheme parameters are adjusted through interaction iteration of an agent and a digital twin system, and the pareto optimal detection scheme is output after the bonus value converges, wherein the detection scheme parameters comprise detection method combination, detection sequence, detection range and operation parameters.
3. The method for generating the detection scheme of the special equipment according to claim 1 is characterized in that time sequence change characteristics of temperature and pressure curves of all components are extracted by utilizing a long-short-term memory network LSTM, defect morphological characteristics of ultrasonic images and radiographic images of all the components are extracted by utilizing a convolutional neural network CNN, text semantic characteristics of historical diagnosis reports of all the components are extracted by utilizing a transducer encoder, and the time sequence change characteristics, the defect morphological characteristics and the text semantic characteristics are dynamically weighted and fused by adopting an attention mechanism to form a unified equipment state feature vector.
4. The method for generating the detection scheme of the special equipment according to claim 1 is characterized by further comprising the steps that after the optimal detection scheme which meets the minimum detection cost and the minimum detection duration is output, a digital twin system generates a prediction result of the optimal detection scheme, the optimal detection scheme is detected by the target equipment in site, actual detection result data is collected, and the reinforcement learning is updated in a feedback mode by means of the difference between the actual detection result data and the prediction result.
5. The method for generating a detection scheme for special equipment according to claim 1, wherein the inherent attribute parameters comprise design parameters, manufacturing information, material marks, structural drawings and service life, the dynamic operation data comprise time sequence parameters of pressure, temperature, flow, vibration, acoustic emission and corrosion monitoring, the historical diagnosis report comprises historical detection report, defect record, maintenance history and nondestructive detection original data and images, and the regulation standard data comprise TSG series regulations, GB/T national standards and industry standards which are stored in a structured mode.
6. The method for generating a detection scheme for a special device according to claim 1, wherein the security assessment score includes a component risk level, a health score, and a comprehensive evaluation of potential failure probability.
7. The method for generating a detection scheme for special equipment according to claim 1, wherein an initial detection scheme including a test item, a recommended test method, a key test part and a recommended test period of each component is generated by a rule engine and a case recommendation system, the rule engine ensures compliance of the initial detection scheme, the case recommendation system searches a successful case most similar to a current problem from a historical case library, and generates a preliminary detection scheme adapting to the current scene by referring to the solution thereof.
8. A detection scheme generation system for a special device, comprising: The system comprises a data acquisition module, a target special device, a data processing module and a control module, wherein the data acquisition module is used for acquiring inherent set attributes, dynamic operation data, historical diagnosis reports and regulation standard data of the target special device, and the dynamic operation data comprises temperature and pressure curves of all parts in the operation of the device, ultrasonic images and radiographic images of all the parts of the device; The device comprises a characteristic extraction module, a dynamic weighting fusion module, a safety evaluation score, a recommendation test method, a key test part and a recommendation test period, wherein the characteristic extraction module is used for extracting time sequence change characteristics of temperature and pressure curves of all parts, extracting defect morphological characteristics of ultrasonic images and radiographic images of all parts and extracting text semantic characteristics of historical diagnosis reports of all parts; the application module is used for performing simulation execution on the initial detection scheme in a preset digital twin system, constructing a reward function by taking safety, economy and efficiency as comprehensive optimization targets by adopting a reinforcement learning algorithm, solving parameters of the initial detection scheme, and outputting an optimal detection scheme which meets the minimum detection cost and the shortest detection duration.
9. A computer device comprising a memory, a processor and a computer program stored on the memory, characterized in that the processor executes the computer program to carry out the steps of the method according to any one of claims 1 to 7.
10. A computer readable storage medium, on which a computer program is stored, characterized in that the computer program, when loaded by a processor, is able to carry out the steps of the method according to any one of claims 1 to 7.

Description

Method, system, equipment and medium for generating detection scheme of special equipment Technical Field The invention belongs to the technical field of special equipment and artificial intelligence application, and particularly relates to a method, a system, equipment and a medium for generating a detection scheme of special equipment. Background Special equipment (such as pressure vessels, boilers, pressure pipelines, elevators, hoisting machinery and the like) is an important infrastructure for national economy and people's life, the operation safety of the special equipment is critical, and the detection of the special equipment is an execution standard for ensuring safe operation. Traditional inspection and detection scheme establishment mainly depends on personal experience of inspection staff and interpretation of national forced regulations, so that massive historical operation data of equipment, on-line monitoring data and inspection records are in an 'information island' state, cannot be effectively integrated and used for predicting equipment performance degradation trend, and experience of senior specialists is difficult to quantify and solidify into scheme establishment flow. In order to break the information island, a device management information system (such as EAM), an electronic inspection report library and an online monitoring data platform are established, and the digitalized centralized storage and inquiry of device files, historical records and part of real-time data are realized. However, the system mainly solves the problem of data access, but lacks the capability of deep analysis, intelligent association and decision support, and the scheme generation still needs the inspector to manually extract information from mass data and rely on experience judgment, and does not substantially convert the data into fuel for driving decisions, so that the data value is not intelligently applied, and the prospective and the accuracy of the inspection scheme are restricted. To overcome subjective differences and solidify some expert experience, a rule engine based decision-making aid system was developed. The system encodes important regulatory provisions and explicit expert judgment logic (e.g. "medium is extremely compromised, the verification period is shortened to X years") into computer-executable rules. The rule engine, while promoting standardization and consistency of solutions, sacrifices the necessary flexibility and adaptability. The method lacks understanding and adapting capability of real world complexity and dynamics, does not have advanced intelligence of continuous learning from data and active optimization under constraint, and therefore cannot generate a truly prospective, accurate and economical dynamic optimal scheme. Disclosure of Invention In order to solve the background problem, the invention provides a method, a system, equipment and a medium for generating a detection scheme of special equipment. In order to achieve the above object, the present invention provides a method for generating a detection scheme of a special device, including: And collecting inherent set attributes, dynamic operation data, historical diagnosis reports and regulation standard data of the target special equipment, wherein the dynamic operation data comprises temperature and pressure curves of all parts in the operation of the equipment, and ultrasonic images and radiographic images of all the parts of the equipment. Extracting time sequence change characteristics of temperature and pressure curves of all components, extracting defect morphological characteristics of ultrasonic images and radiographic images of all components, extracting text semantic characteristics of historical diagnosis reports of all components, and carrying out dynamic weighted fusion on the time sequence change characteristics, the defect morphological characteristics and the text semantic characteristics to form a unified equipment state characteristic vector. And generating an initial detection scheme comprising a detection project, a recommended detection method, a key detection part and a recommended detection period of each component based on the safety evaluation score, the inherent set attribute and the constraint condition of the rule standard data. Performing simulation execution on an initial detection scheme in a preset digital twin system, constructing a reward function by taking safety, economy and efficiency as comprehensive optimization targets by adopting a reinforcement learning algorithm, solving parameters of the initial detection scheme, and outputting an optimal detection scheme which meets the minimum detection cost and the minimum detection duration. The method comprises the steps of constructing a reward function by taking safety, economy and efficiency as comprehensive optimization targets, and solving parameters of an initial detection scheme, wherein the reinforcement learning algorithm adopts