CN113420147A - Special equipment accident reason identification method and system - Google Patents

Abstract

The invention provides a special equipment accident cause identification method and system. Based on the accident cause database, the invention adopts the fuzzy analytic hierarchy process to calculate the initial importance corresponding to each accident cause in each historical case; According to the accident scenario elements in the accident scenario element system, historical cases similar to the current case are selected as similar accident cases from each historical case; The initial importance corresponding to each similar accident cause determines multiple final accident causes corresponding to each layer in the current case; builds an accident cause tree corresponding to the current case based on the multiple final accident causes corresponding to each layer in the current case, and uses the corresponding The accident cause tree is used to identify the cause of the accident in the current case, thereby improving the accuracy of the identification of the cause of the accident.

Description

A kind of special equipment accident cause identification method and system

technical field

The invention relates to the technical field of accident cause identification, in particular to a special equipment accident cause identification method and system.

Background technique

As of the end of 2019, the total number of special equipment in the country reached 15.2547 million units, an increase of 9.4% over the end of 2018 and an increase of 17.08% over the end of 2017, basically showing a steady upward trend. However, because special equipment operates at high temperature, high pressure or high speed, and usually contains flammable, explosive, toxic media or a large number of people, it is very easy to cause mass death and serious property damage in the event of an accident. For this reason, it is particularly important to do a good job in the accident prevention and control of special equipment. Rapid and accurate identification of accident causes is the key to effectively preventing and controlling special equipment accidents. At present, the existing research on accident cause identification mainly focuses on the field of traffic accidents, but special equipment accidents are different from traffic accidents after all. At the same time, the limited research on the causes of special equipment accidents mainly focuses on two aspects: 1) Statistical analysis of the causes of a large number of accidents, but the results are very general and cannot accurately identify the specific causes of specific accidents; 2) For a certain accident, use Accident investigation technology analyzes the cause, but it requires professional techniques such as non-destructive testing, which is a long process and requires high requirements, and cannot quickly identify the cause of the accident. In addition, the above-mentioned research is more oriented to the post-event, and cannot realize the identification of the accident cause in other situations such as the pre-event and the event. Obviously, it is necessary to carry out in-depth research on the rapid and accurate identification method of special equipment accident causes.

Forming a comprehensive and systematic cognition of the causes of accidents is the premise to identify the causes of special equipment accidents quickly and accurately. At present, there are accident cause models such as the Heinrich model and the 2-4 model, but the emphases of each model are different, resulting in incomplete representation of the accident causes of special equipment. For example, the Heinrich model does not have specific management. factors, the 2-4 model does not consider triggering reasons, etc. In addition, scholars often use fault trees to analyze the causal relationship between accident causes, but the accident cause trees built by different scholars vary widely. To sum up, the identification of accident causes based on the above methods has the problem of low accuracy.

SUMMARY OF THE INVENTION

The purpose of the present invention is to provide a special equipment accident cause identification method and system to improve the accuracy of accident cause identification.

In order to achieve the above purpose, the present invention provides a method for identifying the cause of a special equipment accident, the method comprising:

Use the accident cause model and fault tree to establish the accident cause characterization model of special equipment;

Based on the accident cause representation model, the accident causes of each historical case are extracted to form an accident cause database;

Based on the accident cause database, the fuzzy analytic hierarchy process is used to calculate the initial importance corresponding to each accident cause in each historical case; each historical case includes multiple layers, and each layer includes multiple accident causes;

Use the grounded-factor analysis method to analyze the accident scene elements of each historical case to form the accident scene element system of special equipment;

According to each accident scenario element in the accident scenario element system, historical cases similar to the current case are selected from each historical case as similar accident cases; each similar accident case includes multiple layers, and each layer includes multiple similar accident causes;

Determine multiple final accident causes corresponding to each layer in the current case according to the initial importance corresponding to each similar accident cause in each similar accident case;

Based on the multiple final accident causes corresponding to each layer in the current case, construct the accident cause tree corresponding to the current case according to the structure diagram of the house of quality and the accident cause tree of similar accident cases, and use the accident cause tree corresponding to the current case to carry out the accident analysis of the current case. Cause identification.

Optionally, based on the accident cause database, the fuzzy analytic hierarchy process is used to calculate the initial importance corresponding to each accident cause in each historical case, specifically including:

Based on the accident cause database, a fuzzy complementary judgment matrix corresponding to each historical case is established by using the fuzzy analytic hierarchy process;

Based on the fuzzy complementary judgment matrix corresponding to each historical case, the initial importance corresponding to each accident cause in each historical case is calculated.

Optionally, according to each accident scenario element in the accident scenario element system, a historical case similar to the current case is selected from each historical case as a similar accident case, specifically including:

Use ICTCLAS and stop vocabulary to perform word segmentation and de-stop word processing on the descriptive text of each accident scene element in the current case to obtain the current effective feature word set; use ICTCLAS and stop vocabulary to describe the descriptive text of each accident scene element in the historical case The text is subjected to word segmentation and de-stop word processing to obtain a set of historically effective feature words;

Using the current effective feature word set and the historical effective feature word set to determine the similarity of the current case and each historical case with respect to the scene elements corresponding to each accident scene element;

Integrate information on the similarity of the scenario elements corresponding to each accident scenario element under different scenarios to obtain the scenario similarity of the current case and each historical case under different scenarios;

Determine the scenario similarity threshold SQ;

Determine whether Sim _*d is greater than or equal to SQ; if Sim _*d is greater than or equal to SQ, take the historical case Z _d as a similar accident case of the current case, and take the accident causes of each layer of the historical case Z _d as the similar accident case at each layer. Similar accident causes; where Sim _*d is the situation similarity between the current case and the historical case Z _d under different scenarios, d=1, 2, ..., D, D represents the total number of historical cases.

Optionally, the determining of multiple final accident causes corresponding to each layer in the current case according to the initial importance corresponding to each similar accident cause in each similar accident case specifically includes:

Calculate the similarity importance corresponding to each similar accident cause in each similar accident case based on the initial importance corresponding to each similar accident cause in each similar accident case, the current case and the scenario similarity of each similar accident case;

Combine multiple similar accident causes on each layer of each similar accident case to remove duplicates, and obtain the final set of similar accident causes corresponding to each layer of the current case;

The initial importance corresponding to the multiple identical similar accident causes on each layer of each similar accident case is added and processed, and the final importance vector set corresponding to each layer of the current case is obtained;

Determine the final importance threshold CQ _h ;

是否大于或等于CQ_h；如果

大于或等于CQ_h，则当前案例第h层第t个最终重要度对应的最终相似事故原因为当前案例中第h层对应的最终事故原因，

表示当前案例第h层第t个最终重要度，t＝1，2，...，T_h，T_h为当前案例第h层最终重要度的总个数，h＝1，2，...，7。judge

is greater than or equal to CQ _h ; if

is greater than or equal to CQ _h , then the final similar accident cause corresponding to the t-th final importance degree of the h-th layer of the current case is the final accident cause corresponding to the h-th layer in the current case,

Indicates the t-th final importance degree of the h-th layer of the current case, t=1, 2, ..., T _h , _Th is the total number of the h-th layer final importance degrees of the current case, h=1, 2, ... ., 7.

Optionally, the accident cause tree corresponding to the current case is constructed according to the structure diagram of the house of quality and the accident cause tree of each similar accident case based on the multiple final accident causes corresponding to each layer in the current case, and the accident cause corresponding to the current case is used. The cause tree identifies the cause of the accident in the current case, including:

According to the structure diagram of the house of quality and the accident cause tree of each similar accident case, establish the correlation matrix of each final accident cause in the two adjacent layers of the current case;

Establish the autocorrelation matrix of multiple final accident causes in the hth layer in the current case, h=1, , 2, ..., 7;

Build the auto-correlation matrix of multiple final accident causes in the h-th layer of the current case;

Build the accident cause tree corresponding to the current case according to the correlation matrix, autocorrelation matrix and autocorrelation matrix;

Identify the accident cause of the current case according to the accident cause tree corresponding to the current case.

The present invention also provides a special equipment accident cause identification system, the system includes:

The accident cause characterization model building module is used to establish the accident cause characterization model of special equipment by using the accident cause model and fault tree;

The accident cause library building module is used to extract the accident cause of each historical case based on the accident cause representation model to form an accident cause library;

The initial importance degree calculation module is used to calculate the initial importance degree corresponding to each accident cause in each historical case based on the accident cause database using the fuzzy analytic hierarchy process; each historical case includes multiple layers, and each layer includes multiple accident causes;

The accident scenario element system building module is used to analyze the accident scenario elements of each historical case by using the grounded-factor analysis method to form the accident scenario element system of special equipment;

A similar accident case determination module is used to select historical cases similar to the current case from each historical case as a similar accident case according to each accident scenario element in the accident scenario element system; each similar accident case includes multiple layers, each layer Include multiple similar accident causes;

The final accident cause determination module is used to determine multiple final accident causes corresponding to each layer in the current case according to the initial importance corresponding to each similar accident cause in each similar accident case;

The accident cause identification module is used to construct an accident cause tree corresponding to the current case based on the multiple final accident causes corresponding to each layer in the current case, according to the structure diagram of the house of quality and the accident cause tree of similar accident cases, and use the accidents corresponding to the current case. The cause tree identifies the cause of the accident in the current case.

Optionally, the initial importance calculation module specifically includes:

a fuzzy complementary judgment matrix determination unit, used for establishing a fuzzy complementary judgment matrix corresponding to each historical case by adopting the fuzzy analytic hierarchy process based on the accident cause database;

The initial importance calculation unit is used to calculate the initial importance corresponding to each accident cause in each historical case based on the fuzzy complementary judgment matrix corresponding to each historical case.

Optionally, the similar accident case determination module specifically includes:

The valid feature word set determination unit is used to use ICTCLAS and the stop word list to perform word segmentation and de-stop word processing on the descriptive text of each accident scenario element in the current case, and obtain the current effective feature word set; use ICTCLAS and stop word list to compare historical The descriptive text of each accident scene element in the case is subjected to word segmentation and de-stop word processing to obtain a set of historically effective feature words;

A situation element similarity calculation unit, which is used to determine the situation element similarity corresponding to each accident situation element in the current case and each historical case by using the current effective feature word set and the historical effective feature word set;

Scenario similarity calculation unit, used for information fusion of the similarity of the scenario elements corresponding to each accident scenario element under different scenarios, to obtain the scenario similarity of the current case and each historical case under different scenarios;

a context similarity threshold determination unit, used to determine the context similarity threshold SQ;

The first judgment unit is used to judge whether Sim _*d is greater than or equal to SQ; if Sim _*d is greater than or equal to SQ, the historical case Z _d is regarded as a similar accident case of the current case, and the accident causes of each layer of the historical case Z _d are As the similar accident causes at each level of similar accident cases; where Sim _*d is the situation similarity between the current case and the historical case Z _d under different scenarios, d=1, 2, ..., D, D represents the total number of historical cases number.

Optionally, the final accident cause determination module specifically includes:

The similarity importance calculation unit is used to calculate the similarity importance corresponding to each similar accident cause in each similar accident case based on the initial importance corresponding to each similar accident cause in each similar accident case, the current case and the scenario similarity of each similar accident case ;

The combined deduplication unit is used to merge and deduplicate multiple similar accident causes on each layer of each similar accident case, and obtain the final set of similar accident causes corresponding to each layer of the current case;

The summation processing unit is used to add and process the initial importance degrees corresponding to multiple identical similar accident causes on each layer of each similar accident case, and obtain the final importance vector set corresponding to each layer of the current case;

a final importance threshold determination unit, used for determining the final importance threshold CQ _h ;

是否大于或等于CQ_h；如果

大于或等于CQ_h，则当前案例第h层第t个最终重要度对应的最终相似事故原因为当前案例中第h层对应的最终事故原因，

表示当前案例第h层第t个最终重要度，t＝1，2，...，T_h，T_h为当前案例第h层最终重要度的总个数，h＝1，2，...，7。The second judging unit is used for judging

is greater than or equal to CQ _h ; if

is greater than or equal to CQ _h , then the final similar accident cause corresponding to the t-th final importance degree of the h-th layer of the current case is the final accident cause corresponding to the h-th layer in the current case,

Indicates the t-th final importance degree of the h-th layer of the current case, t=1, 2, ..., T _h , _Th is the total number of the h-th layer final importance degrees of the current case, h=1, 2, ... ., 7.

Optionally, the accident cause identification module specifically includes:

The correlation matrix establishment unit is used to establish the correlation matrix of each final accident cause in the two adjacent layers of the current case according to the structure diagram of the house of quality and the accident cause tree of each similar accident case;

The autocorrelation matrix establishment unit is used to establish the autocorrelation matrix of multiple final accident causes of the hth layer in the current case, h=1,,2,...,7;

The self-or correlation matrix establishment unit is used to establish the self-or correlation matrix of multiple final accident causes of the h-th layer in the current case;

The accident cause tree establishment unit is used to construct the accident cause tree corresponding to the current case by sorting according to the correlation matrix, the autocorrelation matrix and the autocorrelation degree;

The accident cause identification unit is configured to identify the accident cause of the current case according to the accident cause tree corresponding to the current case.

According to the specific embodiments provided by the present invention, the present invention discloses the following technical effects:

According to the method disclosed in the present invention, a fault tree and various accident causation theories are firstly used to establish an accident cause representation model to form an accident cause database at each level, and then the fuzzy analytic hierarchy process, scenario similarity, etc. are used to analyze the accident causes at each level in different scenarios respectively. Finally, the house of quality is used to analyze the relationship between the accident causes at each level, thereby establishing the accident cause tree of the current case, realizing the hierarchical relationship identification between accident causes in different scenarios, and improving the accuracy of accident cause identification.

Description of drawings

In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the accompanying drawings required in the embodiments will be briefly introduced below. Obviously, the drawings in the following description are only some of the present invention. In the embodiments, for those of ordinary skill in the art, other drawings can also be obtained according to these drawings without creative labor.

Fig. 1 is the flow chart of the method for identifying the cause of special equipment accident of the present invention;

Fig. 2 is the schematic diagram of the characterization model of the accident cause of the special equipment of the present invention;

Fig. 3 is the flow chart of determining similar accident cases according to the present invention;

Fig. 4 is the flow chart of determining the final accident cause according to the present invention;

FIG. 5 is a structural diagram of the accident cause identification system for special equipment according to the present invention.

Detailed ways

The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

The purpose of the present invention is to provide a special equipment accident cause identification method and system to improve the accuracy of accident cause identification.

In order to make the above objects, features and advantages of the present invention more clearly understood, the present invention will be described in further detail below with reference to the accompanying drawings and specific embodiments.

The present invention proposes a cause characterization model for special equipment accidents by combining fault trees and various accident cause models, and using the accident cause characterization model is conducive to comprehensively and standardized representation of special equipment accident causes. In addition, the present invention utilizes the situation analysis method combined with historical cases to quickly and accurately identify the cause of the accident.

As shown in FIG. 1, the present invention discloses a method for identifying the cause of special equipment accident, characterized in that the method includes:

S1: Use the accident cause model and fault tree to establish the accident cause representation model of special equipment.

S2: Based on the accident cause representation model, the accident causes of each historical case are extracted to form an accident cause database.

S3: Based on the accident cause database, the fuzzy analytic hierarchy process is used to calculate the initial importance corresponding to each accident cause in each historical case; each historical case includes multiple layers, and each layer includes multiple accident causes.

S4: Use the grounded-factor analysis method to analyze the accident scenario elements of each historical case to form an accident scenario element system for special equipment.

S5: According to each accident scenario element in the accident scenario element system, select a historical case similar to the current case from each historical case as a similar accident case; each similar accident case includes multiple layers, and each layer includes multiple similar accident causes .

S6: Determine a plurality of final accident causes corresponding to each layer in the current case according to the initial importance corresponding to each similar accident cause in each similar accident case.

S7: Based on the multiple final accident causes corresponding to each layer in the current case, construct the accident cause tree corresponding to the current case according to the structure diagram of the house of quality and the accident cause tree of similar accident cases, and use the accident cause tree corresponding to the current case to analyze the current case. Identify the cause of the accident.

Each step is discussed in detail below:

S1: Use the accident cause model and the fault tree to establish an accident cause characterization model for special equipment; the accident cause model includes the Heinrich model and the 2-4 model.

Before identifying the cause of the accident, it is necessary to have a comprehensive and systematic understanding of the type and causal relationship of the cause of the accident.

A characterization model of accident causes for special equipment as shown in Figure 2 is established. The accident cause representation model is mainly composed of seven layers of accident causes, namely two-level trigger causes, two-level direct causes, indirect causes, root causes and root causes, and the causal relationship between the causes at each layer is represented by the AND/or relationship of the fault tree. . The two-level triggering reasons are the accidental release of energy (including mechanical energy, electrical energy, etc.) of the equipment, its medium, and surrounding objects, and the failure or destruction of the control or shielding measures that constrain and limit the energy. The two-level direct causes are mainly the unsafe behavior of people (operators, commanders, etc.), the unsafe state of objects (equipment, surrounding objects, etc.), and the unsafe conditions of the environment (climate and other surrounding natural environments and media and other equipment operating environments) These three aspects, and according to the causal relationship, divide the person/object/environment into two layers: the infringer and the cause. Indirect reasons are habitual behaviors within the organization such as poor safety physiology, and the quality of supplier products and services outside the organization. The fundamental reason is the lack of safety operation procedures and other internal safety management systems, such as political, economic and other external social environments and human genetics. The root causes are the lack of safety culture within the organization such as safety importance, the supervision outside the organization, and the influence of other organizations.

S2: Based on the accident cause representation model, the accident causes of each historical case are extracted to form an accident cause database.

For the 490 special equipment accident cases in the "2005-2013 Special Equipment Typical Accident Cases" (including 90 boilers, 116 pressure vessels, 26 pressure pipelines, 97 hoisting machinery, 43 on-site motor vehicles, elevators 86 passenger ropeways, 4 passenger ropeways, and 28 large-scale amusement facilities), analyze the accident causes and causal relationships of each historical case; secondly, determine the classification and level of the accident causes of each historical case based on the accident cause representation model in Figure 1 , form an accident cause form, and establish an accident cause tree for each historical case; thirdly, according to the names and frequency ratios of various accident causes, according to the principle of simple majority, that is, the name of the cause with a high frequency ratio and a simple and clear reason is the priority item, and the The accident cause name is normalized; finally, according to the normalized accident cause name, the accident cause table and the accident cause tree of each historical case are updated, and the accident cause database of special equipment is established.

S3: Based on the accident cause database, the fuzzy analytic hierarchy process is used to calculate the initial importance corresponding to each accident cause in each historical case. Each historical case includes multiple layers, and each layer includes multiple accident causes.

S3: Based on the accident cause database, the fuzzy analytic hierarchy process is used to calculate the initial importance corresponding to each accident cause in each historical case, specifically including:

S31: Based on the accident cause database, a fuzzy complementary judgment matrix corresponding to each historical case is established by using the fuzzy analytic hierarchy process.

为依据表1中的互补型九标度判断标准，则基于所述事故原因库建立模糊互补判断矩阵，具体公式为：Assumption

In order to base on the complementary nine-scale judgment criteria in Table 1, a fuzzy complementary judgment matrix is established based on the accident cause database, and the specific formula is:

为历史案例Z_d(d＝1，2，...，D)第h(h＝1，2，...，7)层第i(i＝1，2，...，N_dh)个事故原因和第j(j＝1，2，...，N_dh)个事故原因的相对重要度比较结果，

D为历史案例的总个数，N_dh为第h层包括事故原因的总个数。Among them, R _d is the fuzzy complementary judgment matrix corresponding to the historical case Z _d (d=1, 2, ..., D),

is the historical case Z _d (d= ₁ , 2, . The comparison result of the relative importance of the accident cause and the jth (j=1, 2, ..., N _dh ) accident cause,

D is the total number of historical cases, and N _dh is the total number of accident causes in the h-th layer.

Table 1 Nine-scale judgment scale table

S32: Calculate the initial importance corresponding to each accident cause in each layer of each historical case based on the fuzzy complementary judgment matrix corresponding to each historical case, specifically including:

S321: Consistency conversion is performed on the fuzzy complementary judgment matrix corresponding to each historical case to obtain the fuzzy consistency matrix corresponding to each historical case. The specific formula is:

1≤i，j≤N_dh，

为历史案例Z_d第h层第i个事故原因和第j个事故原因的相对重要度比较结果，B_d表示历史案例Z_d对应的模糊一致性矩阵，

为历史案例Z_d(d＝1，2，...，D)第h层第i(i＝1，2，...，N_dh)个事故原因相对于第j(j＝1，2，...，N_dh)个事故原因的模糊一致性值。in,

1≤i, j≤N _dh ,

is the comparison result of the relative importance of the i-th accident cause and the j-th accident cause in the h-th layer of the historical case Z _d , B _d represents the fuzzy consistency matrix corresponding to the historical case Z _d ,

For the historical case Z _d (d=1, 2, ..., D) the i-th (i=1, 2, ..., N _dh ) accident cause of the h-th layer relative to the j-th (j=1, 2 , ..., N _dh ) fuzzy consistency values for accident causes.

S322: Normalize and solve each row in the fuzzy consistency matrix corresponding to each historical case, and obtain the initial importance corresponding to each accident cause at each level of each historical case. The specific formula is:

表示历史案例Z_d(d＝1，2，...，D)第h(h＝1，2，...，7)层第i(i＝1，2，...，N_dh)个事故原因对应的初始重要度，

为历史案例Z_d(d＝1，2，...，D)第h层第i(i＝1，2，…，N_dh)个事故原因相对于第j(j＝1，2，...，N_dh)个事故原因的模糊一致性值。in,

Represents the history case Z _d (d=1, 2, . . . , _D ) h (h=1, 2, . The initial importance corresponding to each accident cause,

For the historical case Z _d (d=1, 2, ..., D) the i-th (i=1, 2, ..., N _dh ) accident cause of the h-th layer is relative to the j-th (j=1, 2, . .., N _dh ) fuzzy consistency values for accident causes.

S4: Use the grounded-factor analysis method to analyze the accident scenario elements of each historical case to form an accident scenario element system for special equipment. In the present invention, grounded theory and factor analysis are referred to as grounded factor analysis.

Before calculating the similarity of accident scenarios, it is necessary to clarify the types, quantities and dimensions of accident scenario elements. Therefore, the present invention provides that the grounded-factor analysis method is used to analyze the accident scenario elements of each historical case, and the specific process of forming a special equipment accident scenario element library is as follows:

(1) Determining the category of scene elements of each special equipment accident: For 490 historical cases, the open decoding of grounded theory is used to extract the concept and category of the descriptive text of the scene elements of historical cases, thus forming 28 special equipment The category of accident scenario elements of the accident.

(2) Determine the main category of accident scenario elements according to the category of scenario elements. The specific steps are as follows:

First, based on the accident scene element categories of 28 special equipment accidents, the Likert scale is used to obtain the influence degree of each category on the special equipment accident scene, and then the factor analysis is carried out using formula (6) to obtain the classification results of each accident scene element category.

Secondly, according to the classification results of each accident scene element category, using the main axis decoding of grounded theory, the main categories corresponding to various accident scene element categories are summarized.

Finally, the relationship between the main category and each accident scenario element category is adjusted appropriately until the meaning set represented by the main category can completely cover the category of accident scenario elements, thus obtaining the main categories of 8 accident scenario elements.

为因子系数。In the formula, X _l is the main category of the l-th initial scenario element, A _o is the o-th scenario element category,

is the factor coefficient.

(3) Determine the core category of accident scenario elements according to the main category of accident scenario elements. The specific steps are as follows:

First, according to the main categories of 8 accident scene elements, the Likert scale is used to analyze the degree of its influence on the accident scene of special equipment, and formula (6) is used for factor analysis to obtain the classification of each main category.

Secondly, using the selective decoding of grounded theory, the core categories corresponding to various main categories are summarized.

Finally, the relationship between the core category and the main category is adjusted appropriately until the meaning set represented by the core category can completely cover the main category, and three core categories of accident scenario elements are obtained. Finally, the special equipment accident scenario element system shown in Table 3 is formed, which is mainly divided into three levels: core category, main category and accident scenario element category. At the same time, according to the accident scenario elements of each historical case, an accident scenario element system table of special equipment is formed, as shown in Table 2.

Table 2 Accident scenario element system table

S5: According to each accident scenario element in the accident scenario element system, select a historical case similar to the current case from each historical case as a similar accident case.

For prior cause identification, the scenario similarity calculation of accident cases is mainly based on the prior scenario elements in Table 2. For in-event cause identification, it is based on the two types of situational elements in Table 2, before and in the event, while for post-event cause identification, it is based on all the situational elements in Table 2, that is, before, during and after the event. Scenario-like elements.

As shown in Figure 3, the specific steps of S5 include:

为当前有效特征词集合中第E_*q个有效特征词，

为历史有效特征词集合中第F_dq个有效特征词，D表示历史案例的总个数。T_*q表示当前案例Z^*对应的有效特征词集合，简称当前有效特征词集，T_dq表示历史案例Z_d对应的有效特征词集合，简称历史有效特征词集合。本实施例中，汉语词法分析系统(Institute ofComputing Technology,Chinese LexicalAnalysisSystem，简称ICTCLAS)。S51: Use ICTCLAS and the stop word list to perform word segmentation and de-stop word processing on the descriptive text of each accident scenario element in the current case Z ^* , and obtain the current effective feature word set T _*q ; use ICTCLAS and stop word list to analyze the historical case Z The descriptive text of each accident scenario element in _d (d=1, 2, ..., D) is subjected to word segmentation and de-stop word processing to obtain a set of historically valid feature words T _dq , where,

is the E _*qth valid feature word in the current valid feature word set,

is the _Fdqth effective feature word in the historical valid feature word set, and D represents the total number of historical cases. T _*q represents the valid feature word set corresponding to the current case Z ^* , referred to as the current valid feature word set, and T _dq represents the valid feature word set corresponding to the historical case Z _d , referred to as the historical valid feature word set. In this embodiment, the Chinese Lexical Analysis System (Institute of Computing Technology, Chinese Lexical Analysis System, ICTCLAS for short).

S52: Use the current valid feature word set T _*q and the historical valid feature word set T _dq to determine the similarity of the current case Z ^* and the historical case Z _d with respect to the scenario elements corresponding to each accident scenario element, and the specific formula is:

表示当前案例Z^*和历史案例Z_d关于事故情形要素A_q的情景要素相似度，l2n()为特征词的总长度，T_*q表示当前有效特征词集，T_dq表示历史有效特征词集合，ρ表示调整系数，可参考经验而定，一般取ρ＝0洠4～0洠6，q＝1，2，...，28，d＝1，2，...，D。in,

Represents the similarity of the situation elements of the current case Z ^* and the historical case Z _d with respect to the accident situation element A _q , l2n() is the total length of the feature words, T _*q represents the current valid feature word set, and T _dq represents the historical valid feature word set , ρ represents the adjustment coefficient, which can be determined with reference to experience. Generally, ρ=0×4～0×6, q=1, 2,..., 28, d=1, 2,...,D.

S53: Integrate information on the similarity of the scenario elements corresponding to each accident scenario element under different scenarios to obtain the scenario similarity Sim _*d of the current case Z ^* and the historical case Z _d under different scenarios, the specific formula is:

表示当前案例Z^*和历史案例Z_d关于事故情景要素A_q的情景要素相似度。in,

Represents the similarity of the current case Z ^* and the historical case Z _d with respect to the accident scenario element A _q .

S54: Determine the scenario similarity threshold, and the specific formula is:

SQ=δ×max{Sim _*d |d=1,2,...,D}(9);

Among them, δ represents the percentage of the maximum similarity between the current case Z ^* and the historical case Z _d , 0<δ≤1, the value of δ is determined according to historical data and experience, D represents the total number of historical cases, Sim _*d represents Scenario similarity of current case Z ^* and historical case _Zd under different scenarios, SQ represents scenario similarity threshold.

S55: Determine whether Sim _*d is greater than SQ; if Sim _*d ≥SQ, take the historical case Z _d as the similar accident case Z _*g of the current case Z ^* , g=1, 2, ..., G, G≤ D. The accident causes of each layer of the historical case Z _d are regarded as the similar accident causes of each layer of the similar accident case Z _{* g} .

当前案例Z^*和相似事故案例Z_*g的情景相似度为Sim_*g。Assuming that there are G (G≤D) similar accident cases in the current case Z ^* , Z _*g is the g (g=1, 2, ..., G) similar accident case of the current case Z ^* , then the similar accident case The initial importance corresponding to the _i - _th (i=1, 2, .

The situational similarity between the current case Z ^* and the similar accident case Z _*g is Sim _*g .

Combined with the initial importance corresponding to each similar accident cause at each level of similar accident cases, and the scenario similarity of similar accident cases, the final importance of the accident cause is calculated, and combined with the importance threshold, the accident cause of each level of the current case is determined. Then combined with the accident cause hierarchy relationship of historical cases, analyze the accident cause hierarchy relationship of the current case, and establish the accident cause tree of the current case, thereby realizing the identification of accident causes at all levels of the current case. As shown in Figure 4, the specific process is as follows:

S6: Determine multiple final accident causes corresponding to each layer in the current case according to the initial importance corresponding to each similar accident cause in each similar accident case, specifically including:

S61: Calculate the similarity importance corresponding to each similar accident cause in each similar accident case based on the initial importance corresponding to each similar accident cause in each similar accident case, the current case and the scenario similarity of each similar accident case, and the specific formula is:

表示相似事故案例Z_*g第h层第i个相似事故原因对应的相似重要度，Sim_*g表示当前案例Z^*和相似事故案例Z_*g的情景相似度，

表示相似事故案例Z_*g中第h层的第i个相似事故原因对应的初始重要度，G表示当前案例Z^*共有相似事故案例的总个数，h＝1，2，...，7，i＝1，2，...，N_gh，N_gh表示第h层有相似事故原因总个数。in,

Represents the similarity importance corresponding to the i-th similar accident cause in the h-th layer of the similar accident case Z _*g , Sim _*g represents the situation similarity between the current case Z ^* and the similar accident case Z _*g ,

Represents the initial importance corresponding to the i-th similar accident cause of the h-th layer in the similar accident case Z _*g , G represents the total number of similar accident cases in the current case Z ^* , h=1, 2, ..., 7 , i=1, 2, ..., N _gh , N _gh represents the total number of similar accident causes in the h-th layer.

For different similar accident cases, the similar accident causes may be repeated. For example, the 6th layer of similar accident cases Z _*1 and Z _*2 have the similar accident cause of "lack of emergency plan". The similar accident causes of similar accident cases, that is, the initial similar accident causes of the current case Z ^* , are merged, de-duplicated and summed. The specific process is as follows:

First, it is necessary to summarize multiple similar accident causes in the h (h=1, 2, ..., 7) layer of all similar accident cases; then merge and deduplicate the same accident causes. There are 4 "deficiencies in emergency plans" in the sixth layer of the accident case, which can be combined into one "deficiency in emergency plans"; finally, the initial importance of these same similar accident causes is summed up, for example, the above 4 The final importance of each "lack of emergency plan" is 0.1, 0.1, 0.15, and 0.12, then through the addition process, the final importance of the combined "lack of emergency plan" is 0.1+0.1+0.15+0.2 =0.55. From this, the h-th layer of the current case Z ^* is finally similar to the accident cause and the corresponding final importance.

其中，

表示当前案例Z^*的第h层第T_h个最终相似事故原因，

S62: Combine multiple similar accident causes on the h-th layer of similar accident cases Z _*g to remove duplicates, and obtain the final set of similar accident causes corresponding to the current case Z ^* -h-th layer

in,

represents the _h -th final similar accident cause of the h-th layer of the current case Z ^* ,

其中，

表示当前案例Z^*的第h层第T_h个最终相似事故原因对应的最终重要度。S63: Add up the initial importance degrees corresponding to multiple identical similar accident causes in the h-th layer of the similar accident cases Z _*g , and obtain the final importance vector set corresponding to the current case Z ^* h-th layer

in,

Represents the final importance corresponding to the T _hth final similar accident cause in the hth layer of the current case Z ^* .

S64: Determine the final importance threshold, the specific formula is:

表示当前案例Z^*第h层第t个最终重要度，t＝1，2，...，T_h。Among them, CQ _h represents the final importance threshold, θ represents the maximum final importance percentage, 0<θ≤1,

Represents the current case Z ^* the t-th final importance of the h-th layer, t=1, 2, . . . , T _h .

是否大于或等于CQ_h；如果

大于或等于CQ_h，则当前案例第h层第t个最终重要度对应的最终相似事故原因为当前案例Z^*中第h层对应的最终事故原因，(t＝1，2，...，T_h)。S65: Judgment

is greater than or equal to CQ _h ; if

is greater than or equal to CQ _h , then the final similar accident cause corresponding to the t-th final importance degree of the h-th layer of the current case is the final accident cause corresponding to the h-th layer in the current case Z ^* , (t=1, 2,..., _Th ).

S7: Based on the multiple final accident causes corresponding to each layer in the current case Z ^* , construct the accident cause tree corresponding to the current case Z ^* according to the structure diagram of the house of quality and the accident cause tree of the similar accident case Z _*g , and use the current case Z* ^* The corresponding accident cause tree identifies the cause of the accident for the current case Z ^* , including:

S71: According to the structure diagram of the house of quality and the accident cause tree of similar accident cases Z _*g , establish a correlation matrix of multiple final accident causes in the current case Z ^* h layer and multiple final accident causes in the h-1 layer, specifically as shown in Table 3. It is assumed in Table 3 that there are U _h final accident causes in the h-th layer and U _h-1 final accident causes in the h-1 layer in the current case Z ^* .

S72: Establish an autocorrelation matrix of multiple final accident causes in the h-th layer in the current case Z ^* , as shown in Table 4. In Tables 4 and 5, it is assumed that there are Uh final accident causes in the _hth layer in the current case Z ^* .

S73: Establish the auto-correlation matrix of multiple final accident causes in the h-th layer in the current case Z ^* , as shown in Table 5.

Specifically, formula (12) is used to calculate the correlation between the final accident cause hu and (h _- 1) _v in Table 3; formula (13) is used to calculate the correlation between the final accident cause _{hu and h o in} Table 4 Auto-correlation degree, use formula (14) to calculate the auto-or correlation degree between the final accident causes _{hu and h o in} Table 5.

Among them, Sim _*g represents the situation similarity between the current case Z ^* and the similar accident case Z _*g . If the final accident cause hu and (h _- 1) _v are related in the similar accident case Z _*g , then η _g =1 , otherwise η _g =0. If the final accident causes _hu and ho are the same in the similar accident case Z _*g , then ξ _g =1, otherwise ξ _g = ₀ . If the final accident causes _hu and _ho are OR in the similar accident case Z _*g , then λ _g =1, otherwise λ _g =0.

S74: Build an accident cause tree corresponding to the current case Z ^* according to the correlation matrix, the self-correlation matrix and the self-OR correlation degree.

S75: Identify the accident cause of the current case Z ^* according to the accident cause tree corresponding to the current case Z ^* .

When there are many accident causes, the correlation threshold can be set with reference to formulas (9) and (11) to further screen the key accident causes and their correlations in the current case Z ^* .

Table 3. Correlation matrix of final accident causes for current case Z ^* h-th layer and h-1 layer

Table 4 The autocorrelation matrix of the h-th tier final accident causes for the current case Z ^*

Table 5 The auto-correlation matrix of the h-th tier final accident causes for the current case Z ^*

Compared with the existing technology:

1. It can quickly and accurately realize the identification of accident causes in different scenarios

Existing cause analysis of special equipment is more for post-event scenarios, and the proposed method can quickly reuse the accident causes and their correlations in corresponding historical cases through the calculation of scenario similarity in three scenarios: pre-event, in-event and post-event. In order to realize the identification of the accident cause of the current case in three different scenarios respectively. Moreover, the proposed method can analyze specific accidents, without accident investigation technology, and has low requirements for users, which is convenient for popularization and application.

2. Can identify the hierarchical relationship between the causes of accidents

Existing research on accident causes of special equipment is mostly oriented to the analysis of accident causes at the same level, and cannot form a hierarchical accident cause identification similar to a fault tree, especially in the pre-event and in-event scenarios. The proposed method firstly uses fault tree and various accident causation theories to establish an accident cause representation model and form an accident cause database at each level. Finally, the house of quality is used to analyze the relationship between the accident causes at each layer, thereby establishing the accident cause tree of the current case, and realizing the hierarchical relationship identification between the accident causes in different scenarios.

As shown in Figure 5, the present invention discloses a special equipment accident cause identification system, the system includes:

The accident cause characterization model building module 501 is used to establish an accident cause characterization model of the special equipment by using the accident cause model and the fault tree.

The accident cause database building module 502 is used for extracting the accident causes of each historical case based on the accident cause representation model to form an accident cause database.

The initial importance degree calculation module 503 is used to calculate the initial importance degree corresponding to each accident cause in each historical case based on the accident cause database using the fuzzy analytic hierarchy process; each historical case includes multiple layers, and each layer includes multiple accident causes .

The accident scenario element system building module 504 is used to analyze the accident scenario elements of each historical case by using the grounded-factor analysis method to form an accident scenario element system of the special equipment.

A similar accident case determination module 505 is configured to select historical cases similar to the current case from each historical case as similar accident cases according to each accident scenario element in the accident scenario element system; each similar accident case includes multiple layers, each Layers include multiple similar accident causes.

The final accident cause determination module 506 is configured to determine a plurality of final accident causes corresponding to each layer in the current case according to the initial importance corresponding to each similar accident cause in each similar accident case.

The accident cause identification module 507 is used to construct an accident cause tree corresponding to the current case based on the multiple final accident causes corresponding to each layer in the current case, according to the structure diagram of the house of quality and the accident cause tree of similar accident cases, and use the corresponding accident cause tree of the current case. The accident cause tree identifies the cause of the accident in the current case.

As an optional implementation manner, the initial importance calculation module 503 of the present invention specifically includes:

The fuzzy complementary judgment matrix determination unit is used for establishing the fuzzy complementary judgment matrix corresponding to each historical case by adopting the fuzzy analytic hierarchy process based on the accident cause database.

The initial importance calculation unit is used to calculate the initial importance corresponding to each accident cause in each historical case based on the fuzzy complementary judgment matrix corresponding to each historical case.

As an optional implementation manner, the similar accident case determination module 505 of the present invention specifically includes:

The valid feature word set determination unit is used to use ICTCLAS and the stop word list to perform word segmentation and de-stop word processing on the descriptive text of each accident scenario element in the current case, and obtain the current effective feature word set; use ICTCLAS and stop word list to compare historical The descriptive text of each accident scene element in the case is subjected to word segmentation and de-stop word processing to obtain a set of historically effective feature words.

The situation element similarity calculation unit is used to determine the situation element similarity corresponding to each accident situation element in the current case and each historical case by using the current effective feature word set and the historical effective feature word set.

The scenario similarity calculation unit is used for information fusion of the similarity of the scenario elements corresponding to each accident scenario element under different scenarios to obtain the scenario similarity of the current case and each historical case under different scenarios.

The context similarity threshold determination unit is used to determine the context similarity threshold SQ.

The first judgment unit is used to judge whether Sim _*d is greater than or equal to SQ; if Sim _*d is greater than or equal to SQ, the historical case Z _d is regarded as a similar accident case of the current case, and the accident causes of each layer of the historical case Z _d are As the similar accident causes at each level of similar accident cases; where Sim _*d is the situation similarity between the current case and the historical case Z _d under different scenarios, d=1, 2, ..., D, D represents the total number of historical cases number.

As an optional implementation manner, the final accident cause determination module 506 of the present invention specifically includes:

The similarity importance calculation unit is used to calculate the similarity importance corresponding to each similar accident cause in each similar accident case based on the initial importance corresponding to each similar accident cause in each similar accident case, the current case and the scenario similarity of each similar accident case .

The merging deduplication unit is used to merge and deduplicate multiple similar accident causes on each layer of each similar accident case, and obtain the final set of similar accident causes corresponding to each layer of the current case.

The summation processing unit is used to add and process the initial importance degrees corresponding to multiple identical similar accident causes on each layer of each similar accident case, and obtain the final importance vector set corresponding to each layer of the current case.

The final importance threshold determination unit is used to determine the final importance threshold CQ _h .

是否大于或等于CQ_h；如果

大于或等于CQ_h，则当前案例第h层第t个最终重要度对应的最终相似事故原因为当前案例中第h层对应的最终事故原因，

表示当前案例第h层第t个最终重要度，t＝1，2，...，T_h，T_h为当前案例第h层最终重要度的总个数，h＝1，2，...，7。The second judging unit is used for judging

is greater than or equal to CQ _h ; if

is greater than or equal to CQ _h , then the final similar accident cause corresponding to the t-th final importance degree of the h-th layer of the current case is the final accident cause corresponding to the h-th layer in the current case,

Indicates the t-th final importance of the h-th layer of the current case, t=1, 2, ..., _Th , Th is the total number of the _h -th layer of the current case's final importance, h=1, 2, ... ., 7.

As an optional implementation manner, the accident cause identification module 507 of the present invention specifically includes:

The correlation matrix establishment unit is used to establish the correlation matrix of each final accident cause in the two adjacent layers of the current case according to the structure diagram of the house of quality and the accident cause tree of each similar accident case.

The autocorrelation matrix establishing unit is used to establish the autocorrelation matrix of multiple final accident causes of the hth layer in the current case, h=1, , 2, . . . , 7.

The auto-correlation matrix establishment unit is used to establish the auto-correlation matrix of multiple final accident causes in the h-th layer in the current case.

The accident cause tree establishment unit is used to construct an accident cause tree corresponding to the current case by sorting according to the correlation matrix, the autocorrelation matrix and the autocorrelation degree.

The accident cause identification unit is configured to identify the accident cause of the current case according to the accident cause tree corresponding to the current case.

The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments, and the same and similar parts between the various embodiments can be referred to each other. For the system disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant part can be referred to the description of the method.

In this paper, specific examples are used to illustrate the principles and implementations of the present invention. The descriptions of the above embodiments are only used to help understand the methods and core ideas of the present invention; meanwhile, for those skilled in the art, according to the present invention There will be changes in the specific implementation and application scope. In conclusion, the contents of this specification should not be construed as limiting the present invention.

Claims (10)

1. a special equipment accident cause identification method, is characterized in that, described method comprises: Use the accident cause model and fault tree to establish the accident cause characterization model of special equipment; Based on the accident cause representation model, the accident causes of each historical case are extracted to form an accident cause database; Based on the accident cause database, the fuzzy analytic hierarchy process is used to calculate the initial importance corresponding to each accident cause in each historical case; each historical case includes multiple layers, and each layer includes multiple accident causes; Use the grounded-factor analysis method to analyze the accident scene elements of each historical case to form the accident scene element system of special equipment; According to each accident scenario element in the accident scenario element system, historical cases similar to the current case are selected from each historical case as similar accident cases; each similar accident case includes multiple layers, and each layer includes multiple similar accident causes; Determine multiple final accident causes corresponding to each layer in the current case according to the initial importance corresponding to each similar accident cause in each similar accident case; Based on the multiple final accident causes corresponding to each layer in the current case, construct the accident cause tree corresponding to the current case according to the structure diagram of the house of quality and the accident cause tree of similar accident cases, and use the accident cause tree corresponding to the current case to carry out the accident analysis of the current case. Cause identification. 2. The special equipment accident cause identification method according to claim 1, characterized in that, based on the accident cause database, the fuzzy analytic hierarchy process is used to calculate the initial importance corresponding to each accident cause in each historical case, specifically comprising: : Based on the accident cause database, a fuzzy complementary judgment matrix corresponding to each historical case is established by using the fuzzy analytic hierarchy process; Based on the fuzzy complementary judgment matrix corresponding to each historical case, the initial importance corresponding to each accident cause in each historical case is calculated. 3. The method for identifying the cause of special equipment accident according to claim 1, wherein, according to each accident scenario element in the accident scenario element system, a historical case similar to the current case is selected from each historical case as the similarity. Accident cases, including: Use ICTCLAS and stop vocabulary to perform word segmentation and de-stop word processing on the descriptive text of each accident scene element in the current case to obtain the current effective feature word set; use ICTCLAS and stop vocabulary to describe the descriptive text of each accident scene element in the historical case The text is subjected to word segmentation and de-stop word processing to obtain a set of historically effective feature words; Using the current effective feature word set and the historical effective feature word set to determine the similarity of the current case and each historical case with respect to the scene elements corresponding to each accident scene element; Integrate information on the similarity of the scenario elements corresponding to each accident scenario element under different scenarios to obtain the scenario similarity of the current case and each historical case under different scenarios; Determine the scenario similarity threshold SQ; Determine whether Sim _*d is greater than or equal to SQ; if Sim _*d is greater than or equal to SQ, take the historical case Z _d as a similar accident case of the current case, and take the accident causes of each layer of the historical case Z _d as the similar accident case at each layer. Similar accident causes; where Sim _*d is the situation similarity between the current case and the historical case Z _d under different scenarios, d=1, 2, ..., D, D represents the total number of historical cases. 4. The method for identifying accident causes of special equipment according to claim 1, characterized in that, according to the initial importance corresponding to each similar accident cause in each similar accident case, a plurality of final accident causes corresponding to each layer in the current case is determined , including: Calculate the similarity importance corresponding to each similar accident cause in each similar accident case based on the initial importance corresponding to each similar accident cause in each similar accident case, the current case and the scenario similarity of each similar accident case; Combine multiple similar accident causes on each layer of each similar accident case to remove duplicates, and obtain the final set of similar accident causes corresponding to each layer of the current case; The initial importance corresponding to the multiple identical similar accident causes on each layer of each similar accident case is added and processed, and the final importance vector set corresponding to each layer of the current case is obtained; Determine the final importance threshold CQ _h ; judge

is greater than or equal to CQ _h ; if

is greater than or equal to CQ _h , then take the final similar accident cause corresponding to the t-th final importance degree of the h-th layer of the current case as the final accident cause corresponding to the h-th layer in the current case,

k represents the t-th final importance of the h-th layer of the current case, t=1, 2, ..., T _h , T _h is the total number of the h-th layer final importance of the current case, h=1, 2, . .., 7. 5. The method for identifying accident causes of special equipment according to claim 1, characterized in that, based on the multiple final accident causes corresponding to each layer in the current case, according to the structure diagram of the House of Quality and the accident cause tree of each similar accident case Build the accident cause tree corresponding to the current case, and use the accident cause tree corresponding to the current case to identify the accident cause of the current case, including: According to the structure diagram of the house of quality and the accident cause tree of each similar accident case, establish the correlation matrix of each final accident cause in the two adjacent layers of the current case; Establish the autocorrelation matrix of multiple final accident causes in the hth layer in the current case, h=1, , 2, ..., 7; Build the auto-correlation matrix of multiple final accident causes in the h-th layer of the current case; Build the accident cause tree corresponding to the current case according to the correlation matrix, autocorrelation matrix and autocorrelation matrix; Identify the accident cause of the current case according to the accident cause tree corresponding to the current case. 6. A special equipment accident cause identification system, characterized in that the system comprises: The accident cause characterization model building module is used to establish the accident cause characterization model of special equipment by using the accident cause model and fault tree; The accident cause library building module is used to extract the accident cause of each historical case based on the accident cause representation model to form an accident cause library; The initial importance degree calculation module is used to calculate the initial importance degree corresponding to each accident cause in each historical case based on the accident cause database using the fuzzy analytic hierarchy process; each historical case includes multiple layers, and each layer includes multiple accident causes; The accident scenario element system building module is used to analyze the accident scenario elements of each historical case by using the grounded-factor analysis method to form the accident scenario element system of special equipment; A similar accident case determination module is used to select historical cases similar to the current case from each historical case as a similar accident case according to each accident scenario element in the accident scenario element system; each similar accident case includes multiple layers, each layer Include multiple similar accident causes; The final accident cause determination module is used to determine multiple final accident causes corresponding to each layer in the current case according to the initial importance corresponding to each similar accident cause in each similar accident case; The accident cause identification module is used to construct an accident cause tree corresponding to the current case based on the multiple final accident causes corresponding to each layer in the current case, according to the structure diagram of the house of quality and the accident cause tree of similar accident cases, and use the accidents corresponding to the current case. The cause tree identifies the cause of the accident in the current case. 7. The special equipment accident cause identification system according to claim 6, wherein the initial importance calculation module specifically comprises: a fuzzy complementary judgment matrix determination unit, used for establishing a fuzzy complementary judgment matrix corresponding to each historical case by adopting the fuzzy analytic hierarchy process based on the accident cause database; The initial importance calculation unit is used to calculate the initial importance corresponding to each accident cause in each historical case based on the fuzzy complementary judgment matrix corresponding to each historical case. 8. The special equipment accident cause identification system according to claim 6, wherein the similar accident case determination module specifically comprises: The valid feature word set determination unit is used to use ICTCLAS and the stop word list to perform word segmentation and de-stop word processing on the descriptive text of each accident scenario element in the current case, and obtain the current effective feature word set; use ICTCLAS and stop word list to compare historical The descriptive text of each accident scene element in the case is subjected to word segmentation and de-stop word processing to obtain a set of historically effective feature words; A situation element similarity calculation unit, which is used to determine the situation element similarity corresponding to each accident situation element in the current case and each historical case by using the current effective feature word set and the historical effective feature word set; Scenario similarity calculation unit, used for information fusion of the similarity of the scenario elements corresponding to each accident scenario element under different scenarios, to obtain the scenario similarity of the current case and each historical case under different scenarios; a context similarity threshold determination unit, used to determine the context similarity threshold SQ; The first judgment unit is used to judge whether Sim _*d is greater than SQ; if Sim _*d is greater than or equal to SQ, the historical case Z _d is regarded as a similar accident case of the current case, and the accident causes of each layer of the historical case Z _d are regarded as similar accident cases. Similar accident causes at each level of accident cases; where Sim _*d is the situation similarity between the current case and the historical case Z _d under different scenarios, d=1, 2, ..., D, D represents the total number of historical cases . 9. The special equipment accident cause identification system according to claim 6, wherein the final accident cause determination module specifically comprises: The similarity importance calculation unit is used to calculate the similarity importance corresponding to each similar accident cause in each similar accident case based on the initial importance corresponding to each similar accident cause in each similar accident case, the current case and the scenario similarity of each similar accident case ; The combined deduplication unit is used to merge and deduplicate multiple similar accident causes on each layer of each similar accident case, and obtain the final set of similar accident causes corresponding to each layer of the current case; The summation processing unit is used to add and process the initial importance degrees corresponding to multiple identical similar accident causes on each layer of each similar accident case, and obtain the final importance vector set corresponding to each layer of the current case; a final importance threshold determination unit, used for determining the final importance threshold CQ _h ; The second judging unit is used for judging

is greater than or equal to CQ _h ; if

is greater than or equal to CQ _h , then the final similar accident cause corresponding to the t-th final importance degree of the h-th layer of the current case is the final accident cause corresponding to the h-th layer in the current case,

Indicates the t-th final importance degree of the h-th layer of the current case, t=1, 2, ..., T _h , _Th is the total number of the h-th layer final importance degrees of the current case, h=1, 2, ... ., 7. 10. The special equipment accident cause identification system according to claim 6, wherein the accident cause identification module specifically comprises: The correlation matrix establishment unit is used to establish the correlation matrix of each final accident cause in the two adjacent layers of the current case according to the structure diagram of the house of quality and the accident cause tree of each similar accident case; The autocorrelation matrix establishment unit is used to establish the autocorrelation matrix of multiple final accident causes of the hth layer in the current case, h=1,,2,...,7; The self-or correlation matrix establishment unit is used to establish the self-or correlation matrix of multiple final accident causes of the h-th layer in the current case; The accident cause tree establishment unit is used to construct the accident cause tree corresponding to the current case by sorting according to the correlation matrix, the autocorrelation matrix and the autocorrelation degree; The accident cause identification unit is configured to identify the accident cause of the current case according to the accident cause tree corresponding to the current case.

Images