Automatic analysis report generation method

Abstract

In order to solve the problems that the severity of an accident cannot be comprehensively analyzed, studied and judged and the risk of a major accident cannot be timely identified in the traditional technology, the invention provides an automatic analysis report generation method, which comprises the following steps of: acquiring accident report information, constructing an accident knowledge graph and analyzing the cause of the accident; according to the reason hypothesis, using a decision tree algorithm to obtain a prevention strategy; carrying out loss evaluation on an accident result, judging the severity of the accident, and identifying a major accident; analyzing major accident characteristics to obtain a relationship between major accidents and risks; and based on the relationship between the major accident and the risk, predicting and preventing the major accident in advance, and automatically generating a text report. According to the method, the characteristics of time, places, equipment and personnel are combined, the accident prevention and management effect can be improved through the technical means of accident knowledge graph construction, rule analysis, prevention countermeasure generation and the like, and the risks of casualties, economic property losses, environmental pollution and the like are reduced.

Description

Technical Field

[0001] The present invention relates to the field of information technology, and in particular to a method for automatically generating an analysis report. Background Art

[0002] With the continuous accumulation of dispatch accident data, the traditional manual analysis method can no longer meet the needs of generating analysis reports quickly and accurately. The main problems currently exist are as follows: low analysis efficiency. Traditional manual analysis requires a lot of manpower and time investment, and cannot generate comprehensive and accurate analysis reports in a timely manner. At the same time, due to the intervention of human subjective factors, the analysis results may have certain errors. Preventive measures are not comprehensive. Traditional manual analysis methods can often only find some potential problems and rules from limited data, and cannot fully consider the comprehensive impact of multiple factors such as time, location, equipment and personnel. Manual analysis methods require a lot of manpower and time investment, and the cost is high. Manually collating and analyzing data requires a lot of time and energy, and it is impossible to update and iterate in real time. This is a disadvantage for accident prevention, because the occurrence and evolution of accidents are dynamic processes that require timely data support and analysis feedback. Therefore, preventive measures may be omitted or imperfect. The assessment of the severity of the accident is inaccurate. Manual assessment of the severity of the accident is easily affected by subjective factors, and the assessment results may be subjective and inaccurate. At the same time, traditional methods cannot judge the severity of the accident through comprehensive analysis, and cannot identify the risks of major accidents in a timely manner.

[0003] In order to solve the above problems, the Chinese invention patent with patent number CN113220903A proposes a power accident visualization analysis system and method based on knowledge graph, including first, the knowledge graph layer processes the text data of the power accident event and constructs a graph database; the interface interaction layer obtains the data request and sends it to the data access layer; then, the data access layer processes the graph database according to the request, and sends the processing result to the business logic layer. The business logic layer switches modes, edits nodes and relationships, and analyzes the event process according to the request and the processing result, and determines the main responsible persons and power equipment of the event; finally, the interface interaction layer visualizes the processing results of the business logic layer. However, the invention only realizes accident visualization through knowledge graph, and does not connect and analyze the relationship between the accident and the corresponding risk, which may lead to the inability to identify the accident in time, and does not provide accident prediction and prevention methods, so the invention has defects. Summary of the invention

[0004] The present invention mainly solves the problem that the prior art is unable to judge the severity of an accident through comprehensive analysis, and is unable to timely identify the risk of a major accident, and provides a method for automatically generating an analysis report.

[0005] The above technical problems of the present invention are mainly solved by the following technical solutions: A method for automatically generating an analysis report, characterized in that it comprises the following steps: S1. Obtain accident report information, build accident knowledge graph, and analyze the cause of the accident; S2. Based on the cause hypothesis, a preventive measure is obtained using a decision tree algorithm; S3. Evaluate the loss of the accident, determine the severity of the accident, and identify major accidents; S4. Analyze the characteristics of major accidents and obtain the relationship between major accidents and risks; S5. Based on the relationship between the major accidents and risks, major accidents are predicted and prevented in advance, and a text report is automatically generated.

[0006] As a preferred solution, in step S1, obtaining the accident report information includes obtaining the time period, location area, equipment types involved and category information of personnel involved in the accident, and extracting entities and attributes in the text through regular expression pattern matching.

[0007] As a preferred solution, the analysis of the cause of the accident specifically includes the following steps: S1.1. Analyze the time, location, equipment and personnel information of the accident; S1.2. Analyze the frequency and area of ​​accidents; S1.3. Analyze the type and position of accident equipment; S1.4. Analyze the causes of the accident, including time reasons, location reasons, equipment reasons and personnel reasons.

[0008] As a preferred solution, step S1.4 includes obtaining the time period when accidents occur most frequently, the area range, the faulty equipment involved, and the positions where the accidents are caused by operational errors. If the most frequent accident period corresponds to the peak season for product delivery, analyze the time reasons for the mismatch between the period characteristics and the task objectives; if the emergency plan for the most frequent accident area is incomplete, evaluate the location reasons for the expansion of the accident; If the equipment status detection data with the most frequent accidents is abnormal, associate it with the corresponding equipment cause; If the average training score of the post personnel is lower than the preset average score threshold, it is judged to be related to the personnel reasons for the operational behavior error.

[0009] As a preferred solution, the use of a decision tree comprises the following steps: S2.1. Construct corresponding nodes and branches in the decision tree model, classify and predict according to the input attribute values, and generate corresponding preventive measures; S2.2. Adjust and optimize preventive measures for different time periods, locations, equipment and personnel combinations; S2.3. Evaluate and adjust preventive measures based on actual results, and continuously optimize prevention work.

[0010] As a preferred solution, the Apriori algorithm is used to analyze the characteristics of major accidents to obtain the relationship between major accidents and risks, which specifically includes the following steps: S4.1. Extract the features of each major accident and convert the features into binary representation; S4.2, setting a minimum support, and generating all possible frequent item sets from the data set according to the minimum support; S4.3, generating association rules based on the frequent item sets, and screening out several risk factors most relevant to major accidents; S4.4. Based on the screened risk factors, further determine the relationship between the characteristics of major accidents and risks.

[0011] As a preferred solution, in step S4.1, the features of each major accident are extracted using the Apriori algorithm.

[0012] As a preferred solution, step S5 specifically includes the following steps: S5.1. Conduct regular system safety inspections, risk identification and assessments based on the relationship between major accidents and risks; S5.2. Develop a checklist and scoring criteria for the most significant hazard sources, accident-prone areas and systems, and record existing risk hazards; S5.3. Search and match historical cases for major accident types, associate accident characteristics with impact results, and assess the probability of damage caused by major accidents; S5.4. Use the risk matrix and index rule to calculate the severity of individual risks and determine the system's major accident risk value through weighted synthesis; S5.5. When the assessed risk value exceeds the established threshold, take appropriate preventive measures to reduce the possibility of major accidents or losses and implement risk management and control; S5.6. Automatically generate a PDF report of the dispatch accident data and visualize the technical solution for accident handling.

[0013] As a preferred solution, in step S3, the severity of the accident includes the number of casualties, the amount of economic and property losses, and the degree of environmental pollution.

[0014] As a preferred solution, if the comprehensive score of the number of casualties, the amount of economic and property losses, and the degree of environmental pollution exceeds the preset standard, the accident is a major accident.

[0015] Therefore, the advantages of the present invention are: 1. The present invention can automatically generate preventive measures for time, location, equipment and personnel, and conduct loss assessment on accident results, determine the severity of the accident, and then identify major accidents.

[0016] 2. The present invention combines the characteristics of time, place, equipment and personnel, and through technical means such as accident knowledge graph construction, law analysis, and prevention strategy generation, it can improve the effectiveness of accident prevention and management, and reduce the risks of casualties, economic and property losses, and environmental pollution. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 It is a flow chart of the present invention.

[0018] Figure 2 It is a schematic diagram of the present invention. DETAILED DESCRIPTION

[0019] The technical solution of the present invention is further specifically described below through embodiments and in conjunction with the accompanying drawings.

[0020] Embodiment 1: The present application provides a method for automatically generating an analysis report, comprising the following steps: S1. Obtain accident report information, construct an accident knowledge graph, form the time, location, equipment and personnel dimensions of the accident, analyze the accident occurrence patterns with different characteristics for each dimension, determine the most frequent time period, area, equipment and position of the accident, and analyze the cause of the accident based on the characteristics of the time period, area, equipment and position.

[0021] Construct an accident knowledge graph that includes time, location, equipment, and personnel dimensions; if the number of accidents of a certain category of equipment in the acquired time period exceeds the preset accident number threshold, establish an association between the time period and the equipment category in the accident knowledge graph; determine the region to which the accident location belongs based on the location data in the geographic information system, and establish an association with the environmental conditions and personnel distribution related to the region; if the number of accidents caused by operating and maintenance errors of a certain category of personnel exceeds the preset accident number threshold, mark the causal relationship between the personnel category and the corresponding accident in the accident knowledge graph; establish a quantitative correspondence between the accident level and the impact result based on the economic losses and environmental and social impact assessments under different accident levels; if the environmental or social impact exceeds the preset impact threshold, it is marked as a major accident and associated with the potential major accident risk.

[0022] Obtaining incident report information includes: Obtain information about the time period, location area, equipment type involved, and category of personnel involved in the accident, and extract entities and attributes from the text through regular expression pattern matching.

[0023] Analyze the accident occurrence patterns of different characteristics for each dimension and determine the most frequent time periods, areas, equipment and positions for accidents, including: Based on the time and location information extracted from the accident knowledge graph, determine the time periods and areas with the most frequent accidents in different time periods and regions; calculate a score for each fault based on scoring indicators, including the impact range, duration and repair cost of the fault; sort the equipment types according to the number of times the fault occurs, and the higher the ranking, the higher the fault frequency of the equipment type; determine the positions with the most frequent accidents due to direct operation and management errors based on the number of times personnel in different operating positions are involved in the accident; for different feature dimensions, calculate a rating for each accident based on rating indicators, including the degree of casualties, property damage and environmental impact, and determine the number of accidents and the degree of loss; establish a matching relationship with the corresponding impact rating in the accident knowledge graph based on the type of accident or the degree of environmental damage during the process.

[0024] Analyzing the cause of the accident specifically includes the following steps: S1.1. Analyze the time, location, equipment and personnel information of the accident.

[0025] S1.2. Analyze the frequency and areas of accidents.

[0026] S1.3. Analyze the type and position of accident equipment.

[0027] S1.4. Analyze the causes of the accident, including time reasons, location reasons, equipment reasons and personnel reasons.

[0028] Analyze the causes of accidents based on the characteristics of time period, region, equipment and position, and form hypotheses on time, location, equipment and personnel, including: Obtain the time period with the most frequent accidents, the regional scope, the faulty equipment involved, and the positions where accidents were caused by operational errors. If the time period with the most frequent accidents corresponds to the peak season for product delivery, analyze the time reasons for the mismatch between the time period characteristics and the task objectives. If the emergency plan for the area with the most frequent accidents is incomplete, evaluate the location reasons for the expansion of the accidents. If the equipment status detection data of the most frequent accidents is abnormal, associate the corresponding equipment reasons. If the average training score of the post personnel is lower than the preset average score threshold, judge that it is related to the personnel reasons for operational behavior errors.

[0029] S2. Based on the cause hypothesis, the decision tree algorithm is used to obtain preventive countermeasures and form preventive measures targeting time, place, equipment and personnel.

[0030] Using a decision tree involves the following steps: S2.1. Construct corresponding nodes and branches in the decision tree model, classify and predict according to the input attribute values, and generate corresponding preventive measures.

[0031] S2.2. Adjust and optimize preventive measures according to different time periods, locations, equipment and personnel combinations.

[0032] S2.3. Evaluate and adjust preventive measures based on actual results, and continuously optimize prevention work.

[0033] S3. Conduct a loss assessment on the accident results, determine the severity of the accident, and identify major accidents. The severity of the accident includes the number of casualties, the amount of economic and property losses, and the degree of environmental pollution. If the combined score of the number of casualties, the amount of economic and property losses, and the degree of environmental pollution exceeds the preset standard, the accident is a major accident.

[0034] Specifically include: Based on the accident result impact data, obtain quantitative indicators of the number of casualties, the amount of economic and property losses, and the degree of environmental pollution caused by the accident; based on the number of casualties, the amount of economic and property losses, and the degree of environmental pollution, quantify the indicators, assign weights to different indicators, and perform weighted calculations on different indicators to obtain the severity of the accident; if the number of casualties, the amount of economic and property losses, and the degree of environmental pollution exceed the preset standards, it is determined to be a major accident; through the analysis of the accident result data, calculate the frequency of occurrence of different types of accidents, and identify the most frequent types of accidents; based on the loss assessment of the accident result data, generate an accident assessment report and corresponding recommendations, which include the severity of the accident, the identification of major accidents, and the most frequent types of accidents, and the recommendations include emergency measures and rectification measures taken.

[0035] S4. Use the Apriori algorithm to analyze the characteristics of major accidents and obtain the relationship between major accidents and risks. Specifically, the following steps are included: S4.1. Apriori algorithm is used to extract the features of each major accident and convert the features into binary representation.

[0036] S4.2. Set a minimum support, and generate all possible frequent item sets from the data set based on the minimum support.

[0037] S4.3. Generate association rules based on the frequent item sets to screen out several risk factors most relevant to major accidents.

[0038] S4.4. Based on the screened risk factors, further determine the relationship between the characteristics of major accidents and risks.

[0039] S5. Based on the relationship between the major accidents and risks, major accidents are predicted and prevented in advance, and a text report is automatically generated. Specifically, the following steps are included: S5.1. Conduct regular system safety inspections, risk identification and assessments based on the relationship between major accidents and risks.

[0040] S5.2. Develop checklists and scoring criteria for the most significant hazard sources, areas and systems where accidents occur most frequently, and record existing risks and hidden dangers.

[0041] S5.3. Search and match historical cases for major accident types, associate accident characteristics with impact results, and assess the probability of damage caused by major accidents.

[0042] S5.4. Use the risk matrix and index rule to calculate the severity of individual risks, and determine the system's major accident risk value through weighted synthesis.

[0043] S5.5. When the assessed risk value exceeds the established threshold, take appropriate preventive measures to reduce the possibility of major accidents or losses and implement risk management.

[0044] S5.6. Automatically generate a PDF report of the dispatch accident data and visualize the technical solution for accident handling.

[0045] Embodiment 2: include: S101. Obtain the time, location, equipment and personnel information in the accident report, construct an accident knowledge graph, and form the time, location, equipment and personnel dimensions of the accident.

[0046] According to the accident report, the time period, location area, equipment type involved and personnel category information of the accident are obtained. Through regular expression pattern matching, entities and attributes in the text are extracted to build an accident knowledge graph containing time, location, equipment and personnel dimensions. If the number of accidents of a certain category of equipment in the obtained time period exceeds the preset accident number threshold, the association between the time period and the equipment category is established in the accident knowledge graph. According to the location data in the geographic information system, the region to which the accident location belongs is determined, and the environmental conditions and personnel distribution related to the region are associated. If the number of accidents caused by the operation and maintenance errors of a certain category of personnel exceeds the preset accident number threshold, the causal relationship between the personnel category and the corresponding accident is marked in the accident knowledge graph. According to the economic loss, environmental and social impact assessment under different accident levels, a quantitative correspondence between the accident level and the impact result is established. If the environmental or social impact exceeds the preset impact threshold, it is marked as a major accident and associated with the potential major accident risk.

[0047] For example, there is a welding machine accident that occurred in a factory in District B, City A, on the morning of July 5, 2022. Through regular expression pattern matching, we can extract multi-dimensional information from the accident report, including the time of occurrence 2022-07-05 09:35, the location A City B District, the equipment type is welding machine, and the person involved is the operator Li. In the accident knowledge graph, this information is used as entity nodes, and the relationship between them is established, "welding machine accident" - "occurrence time" - "2022-07-05 09:35", "welding machine accident" - "occurrence location" - "A City B District", "welding machine accident" - "involved equipment" - "welding machine", "welding machine accident" - "operator" - "Li". Statistical analysis shows that the number of accidents in this welding machine category in July 2022 is relatively large, so an association relationship and edge are added between the equipment category node and the time node "2022-07". Count the number of accidents in District B of City A in 2022, determine its accident distribution and compare it with other administrative regions, and establish a quantitative correlation with other characteristic information such as environmental pollution and population density. If the accident causes certain economic losses and is assessed to not exceed the threshold of major accidents such as environmental impact, only a correlation with the category of minor accidents and the range of low economic losses will be established. There is another accident whose social impact score exceeds the major accident standard of 60 points, and a qualitative corresponding logical relationship is established with the major accident category and the corresponding risk node in the accident knowledge graph.

[0048] S102. Analyze the accident occurrence patterns with different characteristics in each dimension, and determine the time period, area, equipment and position where accidents occur most frequently.

[0049] According to the time and location information extracted from the accident knowledge graph, determine the time periods and areas with the most frequent accidents in different time periods and different areas. Calculate the score for each fault based on the scoring indicators, including the impact range, duration and repair cost of the fault. Sort the equipment types according to the number of times the fault occurs. The higher the ranking, the higher the fault frequency of the equipment type. According to the number of times personnel in different operating positions are involved in the accident, determine the positions with the most frequent accidents caused by direct operation and management errors. For different feature dimensions, calculate the rating for each accident based on the rating indicators, including the degree of casualties, property damage and environmental impact, and determine the number of accidents and the degree of loss. According to the type of accident or the degree of environmental damage during the process, establish a matching relationship with the corresponding impact rating in the accident knowledge graph.

[0050] For example, based on the time and location information extracted from the accident knowledge graph, the number of accidents that occurred in each month and administrative region can be obtained. In January 2021, 10 accidents occurred, and 15 accidents occurred in February. Based on these data, it can be judged that February is the time period with the most frequent accidents. Similarly, 20 accidents occurred in the city center and 10 accidents occurred in the suburbs. Based on these data, it can be concluded that the city center is the geographical range with the most frequent accidents. For fault score calculation, the impact range of a certain fault is 1,000 people, the duration is 5 days, and the repair cost is 10,000 yuan. According to the given scoring indicators, the impact range, duration and repair cost can be weighted and summed to obtain the score of the fault. The impact range weight is 0.4, the duration weight is 0.3, and the repair cost weight is 0.3. Then the score of the fault is, score = 0.4*1000+0.3*5+0.3*10000=400+1.5+3000=3401.5. According to the number of equipment failures, equipment type A had 50 failures, equipment type B had 30 failures, and equipment type C had 20 failures. According to the failure frequency, equipment type A ranked first, equipment type B ranked second, and equipment type C ranked third. According to the number of times personnel in different operating positions participated in accidents, operating position A participated in 10 accidents, and management position B participated in 5 accidents. According to the number of accidents involved, operating position A ranked first and management position B ranked second. It can be determined that the position with the most frequent accidents caused by operating errors is operating position A. For the accident rating calculation, the degree of casualties in a certain accident is level 3, the degree of property loss is level 2, and the degree of environmental impact is level 4. According to the given rating indicators, the degree of casualties, the degree of property loss, and the degree of environmental impact can be weighted and summed to obtain the rating of the accident. The weight of casualties is 0.4, the weight of property damage is 0.3, and the weight of environmental impact is 0.3. Then the rating of the accident is, rating = 0.4*3 + 0.3*2 + 0.3*4 = 1.2 + 0.6 + 1.2 = 3. According to the environmental damage degree of the accident type or process, a matching relationship is established with the corresponding impact rating. If the accident type is fire and the environmental damage degree rating is level 4, then the rating of the accident can be set to level 4.

[0051] S103. Analyze the causes of accidents based on the characteristics of time period, region, equipment and job position, and form hypotheses on time, location, equipment and personnel causes.

[0052] Obtain the time period with the most frequent accidents, the area scope, the faulty equipment involved, and the positions where the accidents were caused by operational errors. If the time period with the most frequent accidents corresponds to the peak season for product delivery, analyze the time reasons for the mismatch between the time period characteristics and the task objectives. If the emergency plan for the area with the most frequent accidents is incomplete, evaluate the location reasons for the expansion of the accidents. If the equipment status detection data for the most frequent accidents is abnormal, associate the corresponding equipment reasons. If the average training score of the position personnel is lower than the preset average score threshold, it is judged to be related to the personnel reasons for operational behavior errors.

[0053] For example, a factory has a series of accidents, and the accidents need to be counted, evaluated, and analyzed. Get the time period when the accidents are most frequent. There were 100 accidents in the past year, and record the time of each accident. Through statistical analysis, it is found that the most frequent time period is from 8 am to 10 am. This means that the most accidents occurred between 8 am and 10 am. Get the area range where the accidents are most frequent. The factory has three areas, Area A, Area B, and Area C. Record the area where each accident occurred and count the number of accidents in each area. Through statistical analysis, it is found that the most frequent area is Area A, with a total of 60 accidents. Get the faulty equipment involved. The faulty equipment involved in each accident is recorded in the accident report. Through statistical analysis, it is found that the most frequently involved faulty equipment is machine A, with a total of 30 accidents related to this equipment. Determine the position where the accident was caused by operating error, and record the position of the relevant personnel in each accident. By evaluating the operating error rate of each position, it is found that the operating error rate of position B is the highest, which is 20%. This indicates that the operating error of position B may be the main cause of the accident. Analyzing the reasons for the mismatch between the most frequent accident time period and the task objectives, the factory is in the peak season for product delivery from 8 to 10 in the morning, and the goal is to improve production efficiency and speed. However, this period is also the time period when accidents occur most frequently. Through analysis, it is found that the time reason may be due to the increase in fatigue and decrease in concentration of staff from 8 to 10 in the morning, which leads to an increase in operational errors, resulting in an increase in accidents. Evaluating the location reasons for the expansion of the accident, the accident plan for Area A and Area B is relatively complete, but the emergency plan for Area C is not perfect. Through evaluation, it is found that after the accident, Area C lacks effective emergency measures and emergency evacuation plans, resulting in an increased risk of the accident expanding. The status detection data of the corresponding equipment is abnormal, and the abnormal status detection data of machine A is related to the frequent occurrence of accidents. By analyzing the equipment reasons, the temperature sensor data of machine A is abnormal, resulting in excessive temperature when the machine is running, which causes the accident. To determine the personnel reasons related to operational behavior errors, the factory trains the operators and records the training scores of each person. By calculating the average score, it is found that the average score of position B is 6 points, which is lower than the preset average score threshold of 7 points. This indicates that the quality of training for Position B is insufficient, which may be related to operational errors.

[0054] S104. Based on the cause hypothesis, the decision tree algorithm is used to obtain corresponding preventive measures to form preventive measures targeting time, location, equipment and personnel.

[0055] Based on the assumptions of time period, region, equipment and personnel, the decision tree algorithm is used to obtain the corresponding solution strategy. The corresponding nodes and branches are constructed in the decision tree model, and the input attribute values ​​are classified and predicted to generate corresponding preventive measures. The preventive measures are adjusted and optimized for different time periods, locations, equipment and personnel combinations. The preventive measures are evaluated and adjusted according to the actual results, and the preventive work is continuously optimized.

[0056] For example, through preliminary analysis, it is determined that high work intensity is one of the reasons for the high accident period. Then, for this time reason, four optional prevention and control strategies are collected, including adjusting task goals, increasing personnel support, improving equipment level, and extending project cycle. The decision tree model is used to calculate the expected effect scores of each strategy, among which adjusting task goals has the highest score of 88 points. According to the results of the decision tree, the preventive measures for the time period are to appropriately lower the production target to avoid excessive fatigue of personnel. Then, for the regional reasons for repeated equipment failures, three corresponding strategies are collected, including replacing new equipment, improving maintenance plans, and adding condition monitoring sensors. After inputting the decision tree model for evaluation, it is found that improving planned maintenance and repair has the best expected effect, with a score of 76 points. Therefore, the preventive measures for this location are to strengthen the maintenance and protection of the device. If it is determined that the uneven skills of operators are one of the causes of the accident, the decision tree model compares training, assessment, and introduction of personnel. The results show that strengthening training has the best effect, with a score of 92 points. Therefore, this can be used as a preventive measure for the corresponding personnel factors.

[0057] S105. Conduct loss assessment on the accident result data, determine the severity of the accident, and identify major accidents. Major accidents are accidents in which the number of casualties, the amount of economic and property losses, and the degree of environmental pollution exceed preset standards.

[0058] According to the accident result impact data, obtain the quantitative indicators of the number of casualties, the amount of economic and property losses, and the degree of environmental pollution caused by the accident. According to the number of casualties, the amount of economic and property losses, and the degree of environmental pollution, through the quantification of indicators, the weights of different indicators are assigned, and the different indicators are weighted and calculated to obtain the severity of the accident. If the number of casualties, the amount of economic and property losses, and the degree of environmental pollution exceed the preset standards, it is determined to be a major accident. By analyzing the accident result data, calculate the frequency of different types of accidents, and identify the most frequent types of accidents. Based on the loss assessment of the accident result data, generate an accident assessment report and corresponding recommendations. The report includes the severity of the accident, the identification of major accidents, and the most frequent types of accidents. The recommendations include emergency measures and rectification measures taken.

[0059] For example, an accident occurred in a factory, resulting in 5 injuries, economic losses of 5 million yuan, and a moderate degree of environmental pollution. According to the weight distribution of the accident assessment indicators, the number of casualties accounts for 40% of the total assessment indicators, the amount of economic property losses accounts for 40% of the total assessment indicators, and the degree of environmental pollution accounts for 20% of the total assessment indicators. First, the number of casualties, the amount of economic losses, and the degree of environmental pollution can be quantified and converted into numerical values. The number of casualties is quantified as 0-10, the amount of economic losses is quantified as 0-1000, and the degree of environmental pollution is quantified as 0-5. The number of casualties is 5, the amount of economic losses is 5 million yuan, and the degree of environmental pollution is 3. According to the weight distribution and the numerical values ​​of the indicator quantification, a comprehensive calculation can be performed. The score of the number of casualties is 5 / 10*40%=20%, the score of the amount of economic losses is 500 / 1000*40%=20%, and the score of the degree of environmental pollution is 3 / 5*20%=12%. Adding up the scores, the severity of the accident is 20%+20%+12%=52%. According to the preset criteria, if the severity exceeds 50%, it is determined to be a major accident. Therefore, the accident is determined to be a major accident. Next, the frequency of different types of accidents can be calculated through the analysis of the accident result data, and the most frequent accident type can be identified. Analyzing the accident data that occurred in the past year, it was found that there were 10 fire accidents, 5 chemical leaks, and 15 traffic accidents. Based on the frequency, it can be identified that traffic accidents are the most frequent type of accidents. Finally, based on the loss assessment of the accident result data, an accident assessment report and corresponding recommendations are generated. The accident assessment report will include the severity of the accident as a major accident, the identification of a major accident as this accident, and the most frequent accident type as a traffic accident. Recommendations may include emergency measures such as strengthening safety training and setting up emergency rescue teams, as well as corrective measures such as improving traffic facilities and strengthening safety management systems.

[0060] S106. Use the Apriori algorithm to analyze the characteristics of major accidents and obtain the relationship between major accidents and risks.

[0061] The Apriori algorithm is used to extract the features of each major accident and convert the features into binary representation. The minimum support is set and all possible frequent item sets are generated from the data set based on the minimum support. Association rules are generated based on the frequent item sets to screen out several risk factors most relevant to major accidents. Based on the screened risk factors, the relationship between the features of major accidents and risks is further determined.

[0062] For example, 100 leakage accident data were collected, and 45 accidents occurred in the tank system. Then, the Apriori algorithm was used to mine association rules, with the accident type and system category as the analysis fields, and the minimum support was set to 0.3 and the minimum confidence was set to 0.6. Finally, multiple association rules were output, among which the support of "tank system" -> "leakage accident" was 0.45 and the confidence was 0.8. This shows that nearly half of all leakage accidents occurred in the tank system, and 80% of the tank system accidents were leakage accidents. Thus, a strong association rule between the two fields was found, and the intrinsic connection between the accident type and the involved system was established. Then, it was found that the vast majority of major leakage accidents occurred in the Tank-101 area, and the detection instruments in this area were seriously damaged. Therefore, according to the correlation analysis, the detection system failure also has a strong logical relationship with the accident impact results. This provides a basis for discovering the correlation between major accident feature information and risk indicators. In addition, the quantitative correlation between accidents and safety risks in terms of loss impact, environmental damage, etc. was comprehensively evaluated, and the matching rules between accident characteristics and corresponding risks were formed.

[0063] S107. Based on the relationship between major accidents and risks, major accidents can be predicted and prevented in advance, and text reports can be automatically generated.

[0064] Based on the relationship between major accidents and risks, regular system safety inspections, risk identification and assessments are carried out. Several checklists and scoring criteria for the most serious sources of danger, the areas where accidents occur most frequently, and the systems are formulated to record the existing risks and hidden dangers. According to the types of major accidents, historical cases are searched and matched, accident characteristics are associated with impact results, and the probability of damage caused by major accidents is assessed. The severity of individual risks is calculated using the risk matrix and index law, and the major accident risk value of the system is determined through weighted synthesis. When the assessed risk value exceeds the established threshold, corresponding preventive measures are taken to reduce the possibility of major accidents or the impact losses, and risk management and control are implemented. A PDF report of the dispatch accident data is automatically generated to visualize the technical solutions for accident handling.

[0065] For example, through preliminary analysis, the matching rules between the high-incidence characteristics of leakage accidents in the tank area and the risk of major accidents were determined. A relevant safety checklist was established, including risk factors such as the integrity of the sealing device, abnormal pressure indicators, and instrument operation status. Inspectors inspect the tank area every week, record abnormal conditions, and score each inspection item. If the total score exceeds the warning line of 60 points, corresponding controls will be implemented. Then, historical leakage accident cases that occurred in the tank area are searched and retrieved, and their damage range and impact are counted. The risk matrix method is used to determine the probability score of the accident and the impact loss score. If the probability value of the leakage accident damage is 0.8 and the impact loss score is 75, the risk value is equal to 0.8*75=60. The established threshold is 50. When the assessed risk value exceeds 50, corresponding prevention and control measures are implemented, such as replacing the sealing device, increasing the inspection frequency, and formulating emergency plans, etc., to reduce the risk value of major accidents. By establishing an accident risk assessment and prediction system, risk management can be carried out according to the characteristics of major accidents to prevent major accidents from occurring. Finally, a PDF report of the dispatch accident data is generated, and the technical solution for accident handling is visualized to facilitate communication and decision-making.

[0066] The specific embodiments described herein are merely examples of the spirit of the present invention. Those skilled in the art may make various modifications or additions to the specific embodiments described or replace them in similar ways, but they will not deviate from the spirit of the present invention or exceed the scope defined by the appended claims.

Claims

1. A method for automatically generating an analysis report, characterized in that: The following steps are involved: S1. Obtain accident report information, build accident knowledge graph, and analyze the cause of the accident; S2. Based on the cause hypothesis, a preventive measure is obtained using a decision tree algorithm; S3. Evaluate the loss of the accident, determine the severity of the accident, and identify major accidents; S4. Analyze the characteristics of major accidents and obtain the relationship between major accidents and risks; S5. Based on the relationship between the major accidents and risks, major accidents are predicted and prevented in advance, and a text report is automatically generated.

2. The method for automatically generating an analysis report according to claim 1, characterized in that: In step S1, the acquisition of accident report information includes acquiring the time period, location area, equipment type involved and category information of personnel involved in the accident, and extracting entities and attributes in the text through regular expression pattern matching.

3. A method for automatically generating an analysis report according to claim 1 or 2, characterized in that: The analysis of the cause of the accident specifically includes the following steps: S1.

1. Analyze the time, location, equipment and personnel information of the accident; S1.

2. Analyze the frequency and area of ​​accidents; S1.

3. Analyze the type and position of accident equipment; S1.

4. Analyze the causes of the accident, including time reasons, location reasons, equipment reasons and personnel reasons.

4. The method for automatically generating an analysis report according to claim 3, characterized in that: The step S1.4 includes obtaining the time period with the highest frequency of accidents, the area range, the faulty equipment involved, and the positions where the accidents were caused by operational errors. If the most frequent accident period corresponds to the peak season for product delivery, analyze the time reasons why the period characteristics do not match the task objectives; If the emergency response plan in the area where accidents occur most frequently is incomplete, evaluate the location and reasons for the expansion of accidents; If the equipment status detection data with the most frequent accidents is abnormal, associate it with the corresponding equipment cause; If the average training score of the post personnel is lower than the preset average score threshold, it is judged to be related to the personnel reasons for the operational behavior error.

5. The method for automatically generating an analysis report according to claim 1, characterized in that: Described utilizing decision tree comprises the following steps: S2.

1. Construct corresponding nodes and branches in the decision tree model, classify and predict according to the input attribute values, and generate corresponding preventive measures; S2.

2. Adjust and optimize preventive measures for different time periods, locations, equipment and personnel combinations; S2.

3. Evaluate and adjust preventive measures based on actual results, and continuously optimize prevention work.

6. The method for automatically generating an analysis report according to claim 1, characterized in that: The Apriori algorithm is used to analyze the characteristics of major accidents and obtain the relationship between major accidents and risks, which specifically includes the following steps: S4.

1. Extract the features of each major accident and convert the features into binary representation; S4.

2. Set the minimum support and generate all possible frequent item sets from the data set according to the minimum support; S4.3, generating association rules based on the frequent item sets, and screening out several risk factors most relevant to major accidents; S4.

4. Based on the screened risk factors, further determine the relationship between the characteristics of major accidents and risks.

7. A method for automatically generating an analysis report according to claim 1 or 6, characterized in that: In step S4.1, the features of each major accident are extracted using the Apriori algorithm.

8. The method for automatically generating an analysis report according to claim 1, characterized in that: Step S5 specifically includes the following steps: S5.

1. Conduct regular system safety inspections, risk identification and assessments based on the relationship between major accidents and risks; S5.

2. Develop a checklist and scoring criteria for the most significant hazard sources, accident-prone areas and systems, and record existing risk hazards; S5.

3. Search and match historical cases for major accident types, associate accident characteristics with impact results, and assess the probability of damage caused by major accidents; S5.

4. Use the risk matrix and index rule to calculate the severity of individual risks and determine the system's major accident risk value through weighted synthesis; S5.

5. When the assessed risk value exceeds the established threshold, take appropriate preventive measures to reduce the possibility of major accidents or losses and implement risk management and control; S5.

6. Automatically generate a PDF report of the dispatch accident data and visualize the technical solution for accident handling.

9. The method for automatically generating an analysis report according to claim 1, characterized in that: In step S3, the severity of the accident includes the number of casualties, the amount of economic and property losses, and the degree of environmental pollution.

10. The method for automatically generating an analysis report according to claim 9, characterized in that: If the combined score of the number of casualties, the amount of economic and property losses, and the degree of environmental pollution exceeds the preset standards, the accident is a major accident.

Images