A safety protection system structure and method for pressure-bearing equipment

Abstract

A safety protection method for pressure equipment comprises the following steps: Step 1: Data acquisition, collecting and importing data from different sources and from different structured and unstructured sources, and storing them in a database; Step 2: Constructing a safety protection scenario model from four dimensions: hazard source, hazard, danger, and safety protection level; Step 3: Safety protection measure analysis, customizing safety protection measures based on the collected data and the safety protection scenario analysis. This invention achieves the acquisition of characteristic data of pressure equipment from three levels: hazard source, hazard, and danger. Based on the acquisition and analysis of characteristic data from these three levels, this invention collects and extracts massive amounts of multi-source heterogeneous data in different ways, and processes and stores structured and unstructured data in different ways, thus achieving the acquisition of characteristic data from these three levels.

Description

Technical Field

[0001] This invention relates to the field of pressure equipment safety technology, specifically to a structure and method for a pressure equipment safety protection system. Background Technology

[0002] Pressure equipment stores pressurized or even toxic and harmful media, which carries the risk of safety accidents. Once an accident occurs, it will cause serious consequences such as casualties, property damage, and environmental destruction. Therefore, sufficient and effective safety protection measures must be set up for pressure equipment to ensure its safe use.

[0003] In the pressure equipment industry, big data analytics represents an engineering extension of current big data analytics models, primarily utilizing massive amounts of data to explore potential future patterns. Therefore, there is an urgent need to leverage big data technology to mine existing data and derive corresponding patterns in pressure equipment safety incidents. The installation, use, inspection, and maintenance of pressure equipment generate a large amount of data. In-depth analysis of this data may reveal patterns in pressure equipment failures, thereby enabling the development of safety measures.

[0004] Existing technologies mainly focus on data mining and analysis of individual components of pressure equipment, or employ unsupervised learning methods, which makes it impossible to comprehensively and effectively predict the hazards, dangers, and risks associated with pressure equipment. Summary of the Invention

[0005] This invention addresses the shortcomings of existing technologies by proposing a multi-source heterogeneous data collection approach based on four levels: prevention measures, mitigation measures, procedural safety, and personal protective equipment, encompassing types such as hazard sources, hazards, and risks. The data is preprocessed through data cleaning and classification. Then, the processed data undergoes correlation analysis and safety-oriented scenario construction. By matching scenarios with the hierarchical levels of safety measures, a complete and customized safety protection system structure and method for pressure equipment is achieved. The specific technical solution is as follows:

[0006] A safety protection method for pressure equipment includes the following steps:

[0007] Step 1: Data Acquisition. Collect and import data from different sources, as well as different structured and unstructured data, and store them in the database.

[0008] Step 2: Construct a safety protection scenario model from four dimensions: hazard source, hazard, danger, and safety protection level;

[0009] Step 3: Security protection measures analysis. Based on the collected data and the analysis of security protection scenarios, customized security protection measures are developed.

[0010] As an optimization, step one specifically involves:

[0011] Methods for acquiring multi-source heterogeneous data include:

[0012] Reading directly from the database using SQL scripts: First, create a data source repository using SQL statements. Then, read data from different data source databases by writing database SQL scripts. Access the business data from different systems through interfaces or directly through database SQL views. Finally, write the acquired data sources into the data acquisition database using SQL scripts for later use in the big data preprocessing, analysis, and prediction stages of pressure equipment.

[0013] Manual data entry: Data is collected manually.

[0014] Transformation of unstructured data: By extracting relevant metadata from unstructured data files and converting it into binary database fields, unstructured data is transformed into structured data.

[0015] As an optimization, step two specifically involves:

[0016] 2.1: Equipment classification and categorization: Data from which features need to be selected is analyzed for business relevance and then classified.

[0017] 2.2: Select characteristic items based on four dimensions: hazard source, hazard, danger, and safety protection level, including equipment type, type of medium in the equipment, phase of medium, whether the medium is flammable, flammability rating, whether it is toxic, toxicity rating, quantity of medium, process temperature, process pressure, leakage, combustion and explosion, other hazardous elements, surrounding population, equipment, geographical location, and safety protection level.

[0018] 2.3: For discrete data processing of feature items, the character type of feature items with sequential meaning is converted into numerical type. For feature items with fixed options, one-hot encoding is performed. For feature items with actual numerical values, they are classified into intervals and then discretized in the same way as feature items with fixed options.

[0019] 2.4: Correlation coefficient. If both variables are continuous and have a linear relationship, the Pearson correlation coefficient can be used; otherwise, the Spearman correlation coefficient should be used.

[0020] 2.5: Safety Protection Scenario Construction. A safety protection scenario model is constructed from four dimensions: hazard source, hazard, danger, and safety protection level, including:

[0021] Data association connects the feature items into a wide table. The associated data table includes equipment PID number, equipment type, equipment operating time, type of medium in the equipment, phase of medium, whether the medium is flammable, flammability rating, whether it is toxic, toxicity rating, medium quantity, process temperature, process pressure, leakage, explosion, other hazardous factors, surrounding population, equipment, geographical location, and safety protection level.

[0022] Associate dynamic attribute parameters to construct the dynamic historical cumulative number of inspections, possible accident causes, and safety protection measures for pressure equipment;

[0023] Constructing scene feature codes: The feature items are sorted according to the hazard source, hazard, danger, and safety protection level, and then secondary encoding is performed using the ASCII character set to generate unique feature codes for different scenes.

[0024] As an optimization, step three specifically involves:

[0025] 3.1 Automatically construct a multi-level safety protection measure view from hazard source to hazard to harm based on feature codes;

[0026] 3.2 Matching of safety protection measures: Extract the data on safety protection, hazard nature and severity of hazard from each of the multi-level safety protection measures views, and combine them to form a complete set of safety protection measures including four levels: prevention measures, mitigation measures, procedural safety and personal protective equipment;

[0027] 3.3 Ranking of Safety Protection Measures: Safety protection capabilities are comprehensively analyzed using the Safety Integrity Levels (SILs) analysis method for prevention and mitigation measures; safety protection capabilities are analyzed using the Risk-Based Inspection (RBI) management factor analysis method for procedural safety measures; and safety protection capabilities are analyzed using the Gaussian analysis model of combustion, explosion, and toxicity diffusion for individual protection measures. The safety measures are then ranked based on the sum of the results.

[0028] The beneficial effects of this invention are as follows:

[0029] 1. This invention enables the collection of characteristic data from three levels—hazard source, hazard, and hazard—for pressure equipment. Traditional safety measure analyses often focus on only one level, failing to comprehensively mine and analyze data from all three levels, thus hindering the development of customized, scientifically effective safety measures. This invention collects and analyzes characteristic data from these three levels, employing various methods to collect and extract massive amounts of multi-source heterogeneous data, and processing and storing structured and unstructured data in different ways, thereby achieving the collection of characteristic data from all three levels.

[0030] 2. A multi-level database of safety measures for pressure equipment and a four-level feature database covering hazard sources, hazards, dangers, and causes of hazards were established. Independent databases were constructed from eight dimensions: preventive measures, mitigation measures, procedural safety, personal protective equipment, hazard sources, hazards, dangers, and causes of hazards. This not only supports the development of complete safety measures but also meets the safety requirements of pressure equipment in different scenarios.

[0031] 3. Rules for classifying and labeling pressure equipment have been established. Based on the standard GB / T 13861-2022 Classification and Code of Hazardous and Harmful Factors in Production Processes, safety measures have been defined and classified hierarchically from the dimensions of human factors, material factors, management factors, and environmental factors.

[0032] 4. Innovative Data Preprocessing and Cleaning Methods. Traditional Excel methods are cumbersome when dealing with massive amounts of unstructured data. This method utilizes Python's flexible and efficient data structures to process missing, duplicate, and abnormal data from pressure equipment, significantly improving data cleaning efficiency.

[0033] 5. Multiple big data mining and analysis methods are integrated. A decision model combining logistic regression, decision trees, and GBDT gradient boosting decision trees is adopted to perform feature analysis and feature selection on the data, establish a complete security measure decision model, and repeatedly train the model through supervised learning to formulate security measures for different scenarios.

[0034] 6. By performing data preprocessing, simulation analysis, and comparison of safety measures on massive amounts of data on safety accidents involving pressure equipment, the effectiveness of safety measures in response to safety accidents has been improved. At the same time, a supervised learning approach has been adopted to effectively improve the multi-level safety measure database for pressure equipment. Attached Figure Description

[0035] Figure 1 This is a flowchart illustrating the process of the invention;

[0036] Figure 2 This is a structural diagram of the hierarchical safety measures system in this invention;

[0037] Figure 3 This is a schematic diagram of the framework structure for acquiring multi-source heterogeneous data in this invention; Detailed Implementation

[0038] The preferred embodiments of the present invention will be described in detail below so that the advantages and features of the present invention can be more easily understood by those skilled in the art, thereby making a clearer and more explicit definition of the scope of protection of the present invention.

[0039] A structure and method for a safety protection system for pressure equipment, comprising the following specific steps:

[0040] I. Safety Measures Scenario Data Collection

[0041] 1.1 Data Source

[0042] Data characterizing hazards, dangers, and risks will generate completely different structured and unstructured data depending on the context, location, and production process. We need to extract hidden, valuable information and knowledge from these numerous, dispersed, and heterogeneous data sources.

[0043] (1) Hazard source data for pressure equipment

[0044] Hazard data for pressure equipment mainly comes from the ERP system or equipment management system of the pressure equipment user, including equipment process data and manufacturing information, such as the type of medium in the equipment, the phase of the medium, whether the medium is flammable, its flammability rating, whether it is toxic, and its toxicity rating.

[0045] (2) Hazard data for pressure equipment

[0046] Hazardous data for pressure equipment mainly comes from production process data, LIMS (Laboratory Management Information System) data, and equipment production environment data. This includes the accidental release of energy, fire, leakage and diffusion of toxic substances, exposure to dusty environments, equipment damage, leakage or exposure of high-temperature substances, etc., caused by temperature, pressure, medium, medium phase, thickness, corrosion rate, material, etc. during the use of pressure equipment.

[0047] (3) Hazard data for pressure equipment

[0048] Hazard data for pressure equipment mainly comes from manual input and includes explosion injuries, collapses, environmental emissions, combustible material explosion injuries, burns, fires, explosive material explosion injuries, and personnel poisoning.

[0049] 1.2 Data Structures and Types

[0050] In the analysis of multi-level safety measures for pressure equipment, the types of data collected include both structured and unstructured data.

[0051] Structured data is data that can be logically expressed and implemented using a two-dimensional table structure. This type of data strictly adheres to specific data format and length specifications and is primarily stored and managed in relational databases. Examples include registration information for pressure equipment, information on property owners, user units, hazard sources, hazards, dangers, and hazard causes, etc.

[0052] Unstructured data refers to data that is inconvenient to represent using the logic of a two-dimensional table. This type of data generally has an irregular structure and lacks a predefined data model, including text, tables, and images. Examples include structural images of pressure equipment, failure images, design specifications, and operation manuals.

[0053] 1.3 Acquisition of Multi-Source Heterogeneous Data

[0054] (1) Framework for acquiring multi-source heterogeneous data

[0055] In the process of data acquisition for pressure equipment, the differences in data sources are generally quite large. Since we need to store the processed data in the database, it is necessary to collect and import data from different sources and different structured and unstructured data separately and store them in the database.

[0056] Due to differences in data storage media, data storage types, and data transmission methods, different import tools and methods are needed to import data from various sources separately during data acquisition. Considering the characteristics of data from pressure equipment, such as significant differences in data structure and a wide range of data sources, a multi-source heterogeneous data acquisition framework provides effective means and methods for data acquisition, organization, and utilization. Through this framework, the acquisition of data from different sources and different structured and unstructured data has been achieved, such as... Figure 2 As shown, methods for collecting multi-source heterogeneous data include directly reading from the database using SQL scripts, manual data entry, and conversion of unstructured data.

[0057] (2) SQL script extraction and database access methods

[0058] By using the SQL scripting language, including database definition statements, database manipulation language, database query statements, and database processing language, the contents of the database can be read directly.

[0059] First, SQL statements are used to create data views in different multi-source heterogeneous databases, such as pressure equipment production data views, production process data views, storage media views, and installation environment views.

[0060] Then, by writing SQL scripts, data is read from databases of different sources. The business data from different systems is accessed via interfaces or directly through database SQL views. Finally, the acquired data sources are written into the data acquisition database using SQL scripts for later analysis and customization of multi-level safety measures for pressure equipment.

[0061] (3) Manual data entry method

[0062] For some basic information about pressure equipment, such as paper materials or pictures, including structural pictures, failure pictures, design specifications, and operation manuals, data can be collected manually.

[0063] (4) Data acquisition methods for transforming unstructured data

[0064] For information stored in some text or image files, the relevant metadata can be extracted and converted into a one-to-one correspondence between device ID and binary storage, thereby transforming unstructured data into structured data.

[0065] 1.5 Database Setup

[0066] Establish a large database of multi-level safety measures for pressure equipment, including basic information, production process information, operating environment data, and storage media information. The database should meet the requirements of safety, uniformity, and reliability. The database construction will be carried out through the design of the physical structure, content, functional modules, database information language, and physical structure design for database storage and backup.

[0067] II. Analysis of Multi-Level Security Measures Based on One-Hot Coding

[0068] 2.1 Multi-level security measures and sparse data structures

[0069] One-hot encoding is a method of encoding that replaces a categorical variable in a dataset with one or more new features. Multi-level safety measures in sparse data structures use this encoding to vectorize and identify the feature data of hazards, dangers, and risks. This includes:

[0070] 1. Hazard source coding

[0071] Hazard source coding includes two levels of coding, with the first level coding being:

[0072] (Energetic substances, toxic substances, asphyxiating substances, dust, others) have unique thermal codes (10000, 01000, 00100, 00010, 00001).

[0073] The second-level code is:

[0074] (Pressure energy, chemical energy, thermal energy, kinetic energy) have unique thermal codes of (1000, 0100, 0010, 0001).

[0075] Hazards with pressure energy can be identified using the code 100001000.

[0076] 2. Pressure Energy Hazard Code

[0077] (Compressed gas, liquefied gas, supercritical fluid) has an unique thermal code of (100, 010, 001).

[0078] 3. Chemical Energy Hazard Code

[0079] (Flammable materials, explosives, corrosives, and spontaneously combustible materials) have unique thermal codes (1000, 0100, 0010, 0001).

[0080] 4. Chemical Energy Hazard Code

[0081] The unique thermal code for (high-temperature and low-temperature objects) is (10, 01).

[0082] 5. Hazard code for accidental energy release

[0083] (Explosion, combustion, escape or exposure of chemical hazards, leakage or exposure of high-temperature substances, movement of objects or exposure of moving parts) Its unique thermal codes are (10000, 01000, 00100, 00010, 00001).

[0084] 6. Hazard coding

[0085] Hazard coding includes two levels of coding, with the first level being:

[0086] (Personal injury, equipment damage, environmental pollution, equipment damage / equipment shutdown) Its unique thermal code is (1000, 0100, 0010, 0001).

[0087] The second-level code is:

[0088] (Injuries from explosions of pressure vessels (including gas cylinders), boilers, pressure pipelines, collapses, environmental emissions, combustible material explosions, burns, fires, and explosive material explosions) are unique thermal codes (10000000, 01000000, 00100000, 000100000, 00001000, 00000100, 000000100, 00000010, 00000010, 00000001).

[0089] 7. Hazard Cause Coding

[0090] (Unsafe acts of people, unsafe conditions of objects, environmental factors, management factors, chemical hazards, physical hazards) have unique thermal codes (100000, 010000, 001000, 000100, 000010, 000001).

[0091] 2. Multi-level security measures library

[0092] The multi-level security measures library comprises four levels: prevention measures, mitigation measures, procedural security, and personal protection.

[0093] (1) Prevention Measures Library

[0094] Serial Number Field Name code 1 Preventive measures signage ProtectMeasure_ID 2 Security protection scenario feature code Scene_ID 3 Prevention measures ProtectMeasure

[0095] (2) Mitigation Measures Library

[0096] Serial Number Field Name code 1 Mitigation measures sign ReduceMeasure_ID 2 Security protection scenario feature code Scene_ID 3 Mitigation measures ReduceMeasure

[0097] (3) Security Library

[0098] Serial Number Field Name code 1 Security Procedure Identification SaftyPrg_ID 2 Security protection scenario feature code Scene_ID 3 Security procedures SaftyPrg

[0099] (4) Personal Protective Equipment Storage

[0100]

[0101]

[0102] 3. Security Protection Scenario Feature Code Rules

[0103] The safety protection scenario feature code is formed by converting binary to hexadecimal based on the hazard source code, hazard code, danger code, and hazard cause code.

[0104] First, the codes are merged in the order of hazard source code, hazard code, harm code, and harm cause code to form a binary code of 0 and 1.

[0105] Then, the generated binary code is converted into hexadecimal every 4 bits to obtain the scene code. When performing scene simulation analysis, the inverse transformation is required to obtain the hazard source code, hazard code, danger code, hazard cause code, etc.

[0106] 4. Obtaining multi-level security protection measures

[0107] For each pressure-bearing device scenario, an analysis is performed. After obtaining the scenario feature code, SQL statements are used to query four-level databases (prevention measures, mitigation measures, program security, and personal protection) to form a safety protection measures dataset.

[0108] III. Multi-level security protection feature data analysis

[0109] 1) Selection of basic information characteristic data items for pressure equipment

[0110] The basic information characteristic data item for pressure equipment is used to identify the specific equipment targeted by multi-level safety protection measures, including equipment category, equipment name, equipment model, equipment installation address, and user unit.

[0111] 2) Selection of Hazardous Source Characteristic Data Items for Pressure Equipment

[0112] The characteristic data items of the hazard source of pressure equipment are used to analyze and characterize the characteristic items of the hazard source, and further generate data for hazard source coding, including the composition of the equipment storage medium, the phase state of the medium, the material, the installation method, etc.

[0113] 3) Selection of hazardous characteristic data items for pressure equipment

[0114] The hazardous characteristic data items of pressure equipment are used to analyze and characterize the characteristics of hazards, and further generate hazard codes. These data include installation location, process temperature, process pressure, equipment wall thickness, commissioning date, corrosion rate, etc.

[0115] 4) Selection of hazard characteristic data items for pressure equipment

[0116] The hazard characteristic data items of pressure equipment are used to analyze and characterize the characteristics of hazards, and further generate hazard source coding data, including production process, surrounding environment, personnel density, equipment density, atmospheric environment, etc.

[0117] 5) Safety protection feature data analysis

[0118] First, using the aforementioned input feature data items, the feature data items of hazard source, hazard, hazard and hazard cause are sorted manually. Second, based on the historical accident statistics of pressure equipment, a fuzzy neural network is used to calculate the corrected weighting coefficient of the parameters. Then, using the sorted value of the same feature item and the value of its corrected weighting coefficient, 1-2 main indicator items of hazard source, hazard, hazard and hazard cause are selected to participate in the analysis of multi-level safety protection measures.

[0119] IV. Output of Multi-Level Safety Protection Measures

[0120] This system uses an Excel data interface to output the multi-level safety protection measures for pressure equipment to an xls file, forming the file content shown in the figure below, which is provided to users for review and execution.

Claims

1. A method for safety protection of pressure equipment, characterized in that, The specific steps are as follows: Step 1: Data Acquisition. Collect and import data from different sources, as well as different structured and unstructured data, and store them in the database. Step 2: Construct a safety protection scenario model from four dimensions: hazard source, hazard, danger, and safety protection level; 2.1: Equipment classification and categorization: Data from which features need to be selected is analyzed for business relevance and then classified. 2.2: Select characteristic items based on four dimensions: hazard source, hazard, danger, and safety protection level, including equipment type, type of medium in the equipment, phase of medium, whether the medium is flammable, flammability rating, whether it is toxic, toxicity rating, quantity of medium, process temperature, process pressure, leakage, combustion and explosion, other hazardous elements, surrounding population, equipment, geographical location, and safety protection level. 2.3: For discrete data processing of feature items, the character type of feature items with sequential meaning is converted into numerical type. For feature items with fixed options, one-hot encoding is performed. For feature items with actual numerical values, they are classified into intervals and then discretized in the same way as feature items with fixed options. 2.4: Correlation coefficient. If both variables are continuous and have a linear relationship, the Pearson correlation coefficient can be used; otherwise, the Spearman correlation coefficient should be used. 2.5: Safety Protection Scenario Construction. A safety protection scenario model is constructed from four dimensions: hazard source, hazard, danger, and safety protection level, including: Data association connects the feature items into a wide table. The associated data table includes equipment PID number, equipment type, equipment operating time, type of medium in the equipment, phase of medium, whether the medium is flammable, flammability rating, whether it is toxic, toxicity rating, medium quantity, process temperature, process pressure, leakage, explosion, other hazardous factors, surrounding population, equipment, geographical location, and safety protection level. Associate dynamic attribute parameters to construct the dynamic historical cumulative number of inspections, possible accident causes, and safety protection measures for pressure equipment; Constructing scene feature codes: The feature items are sorted according to the hazard source, hazard, danger, and safety protection level, and then a second encoding is performed using the ASCII character set to generate unique feature codes for different scenes; Step 3: Security protection measures analysis. Based on the collected data and the analysis of security protection scenarios, customized security protection measures are developed.

2. The safety protection method for pressure-bearing equipment according to claim 1, characterized in that: Step one specifically involves: Methods for acquiring multi-source heterogeneous data include: Reading directly from the database using SQL scripts: First, create a data source repository using SQL statements. Then, read data from different data source databases by writing database SQL scripts. Access the business data from different systems through interfaces or directly through database SQL views. Finally, write the acquired data sources into the data acquisition database using SQL scripts for later use in the big data preprocessing, analysis, and prediction stages of pressure equipment. Manual data entry: Data is collected manually. Transformation of unstructured data: By extracting relevant metadata from unstructured data files and converting it into binary database fields, unstructured data is transformed into structured data.

3. The safety protection method for pressure-bearing equipment according to claim 1, characterized in that: step three specifically comprises: 3.1 Automatically construct a multi-level safety protection measure view from hazard source to hazard to harm based on feature codes; 3.2 Matching of safety protection measures: Extract the data on safety protection, hazard nature and severity of hazard from each of the multi-level safety protection measures views, and combine them to form a complete set of safety protection measures including four levels: prevention measures, mitigation measures, procedural safety and personal protective equipment; 3.3 Ranking of Safety Protection Measures: Safety protection capabilities are comprehensively analyzed using the Safety Integrity Levels (SILs) analysis method for prevention and mitigation measures; safety protection capabilities are analyzed using the Risk-Based Inspection (RBI) management factor analysis method for procedural safety measures; and safety protection capabilities are analyzed using the Gaussian analysis model of combustion, explosion, and toxicity diffusion for individual protection measures. The safety measures are then ranked based on the sum of the results.

Images