Laboratory safety accident emergency decision-making method and system based on hybrid reasoning mechanism
Through the CBR-RBR hybrid reasoning mechanism and Drools reasoning mechanism, combining knowledge metatheory and case description indexing method, the problem of inscientific decision-making in laboratory safety accidents is solved, and more efficient and accurate emergency decision-making in laboratory safety accidents is achieved, and the improvement of enhanced laboratory safety accidents is enhanced. Campus safety management capabilities.
Patent Information
- Application Number
- CN202410875885.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-02
- Publication Date
- 2024-11-26
AI Technical Summary
It is difficult for the existing technology to clearly describe the situations and tasks in laboratory accident cases and obtain historical case knowledge, which leads to inadequate scientific decision-making in laboratory safety management and it is difficult to conduct training and prevent similar accidents according to time.
Using a method based on CBR-RBR hybrid inference mechanism, the laboratory safety accidents are uniformly represented and analyzed through knowledge metatheory and case description indexing method, and decision-making suggestions are constructed using Drools inference mechanism and rule base, and combined with case similarity calculation to generate comprehensive Decision-making advice.
It improves the efficiency and accuracy of emergency decision-making for laboratory safety accidents, enhances campus safety management capabilities, ensures the safety of teachers and students, and the independence of the rule database and case database is convenient for maintenance and updates.
Smart Images

Figure CN119026933A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of laboratory safety management, and in particular to a laboratory safety accident emergency decision-making method and system based on a CBR-RBR hybrid reasoning mechanism. Background Art
[0002] "Safety" and "danger" are like two sides of a coin, and accidents cannot be completely avoided. Although accidents bring negative impacts such as property losses and casualties, they also provide a lot of on-site information such as improper management and system defects. As the cornerstone of the overall safety of university laboratories, the accident database can assist university laboratory managers and accident investigators to understand past accidents similar to the current accident, so that they can better understand the causes of accidents and formulate preventive measures. At the same time, the experience and knowledge gained from combing through historical accident cases can improve the situation where laboratory management departments rely on human experience to make decisions, which is helpful to improve the decision-making level of laboratory management agencies and achieve scientific decision-making. It is of great significance. The accident case library can not only support the identification of accident causes and hidden dangers, but also provide factual basis for arranging reasonable drills and training to prevent similar accidents. The case library of the cause analysis of university laboratory accidents can play the role of a "storage library" well, effectively enhance the safety management capabilities and levels of university laboratories, and ensure the safety and stability of campuses and the safety of teachers and students.
[0003] In the above context, how to clearly describe the scenarios and tasks in accident cases and obtain historical case knowledge to guide emergency drill scenarios that are in line with the characteristics of the period, so as to achieve training effects that are adapted to the times, has become a key issue that urgently needs to be solved. Summary of the invention
[0004] The purpose of the present invention is to overcome the shortcomings of the prior art and provide a laboratory safety accident emergency decision-making method based on a CBR-RBR hybrid reasoning mechanism. By introducing knowledge meta theory and using the knowledge meta model to uniformly represent scenario elements, the information extraction of historical accident cases is convenient. The association rules between scenario elements are analyzed based on the CBR-RBR hybrid reasoning process and case description indexing method, and the safety problems and defects of university laboratories in a certain period of time are obtained to ensure the safety and stability of the campus and the safety of teachers and students.
[0005] Another object of the present invention is to provide a laboratory safety accident emergency decision-making system based on a CBR-RBR hybrid reasoning mechanism.
[0006] The purpose of the present invention is achieved through the following technical solutions:
[0007] A laboratory safety accident emergency decision-making method based on CBR-RBR hybrid reasoning mechanism, comprising the steps of:
[0008] Write the target accident cases into the laboratory safety accident case database;
[0009] Adopting the Drools reasoning mechanism, the rule standards and expert experience knowledge are expressed in the form of production rules, a rule base is constructed, and RBR reasoning is performed on the target accident cases;
[0010] Select the characteristic attributes of the target accident case, and assign weights to each characteristic attribute;
[0011] Calculate the similarity between the target accident case and the historical case based on the weight of the obtained feature attributes;
[0012] If the similarity is greater than the set threshold, the case with the maximum similarity is screened out, and the case is modified by combining the maximum similarity case and the decision suggestion of RBR reasoning to generate a comprehensive decision suggestion based on CBR-RBR hybrid reasoning; if the similarity is less than or equal to the set threshold, a comprehensive decision suggestion based on CBR-RBR hybrid reasoning is generated according to the target accident case and the decision suggestion of RBR reasoning.
[0013] Furthermore, the target accident case is written into the laboratory safety accident case database. The following specific operations are performed:
[0014] The overall framework of laboratory safety accident cases is represented by {T, S, R}. The attribute settings of laboratory safety accident cases include the following:
[0015] T i , description of the scenario and status of the case, that is, how the case occurred, and the description of the time-space information of the incident, including the characteristic attributes of "project scale, accident quarter, accident time, accident factor, accident level, and other status attributes";
[0016] S i ,Emergency strategy, when an incident occurs, intelligently judge the solution to be taken. The solution mainly includes implementation strategy and corresponding emergency handling mechanism, including "general accident emergency strategy", "major accident emergency strategy", "major accident emergency strategy", "particularly major accident emergency strategy" emergency strategy characteristic attributes;
[0017] R i , accident consequences, that is, the final result of the evolution of the incident after the incident occurred and after taking or not taking countermeasures, including the characteristic attributes of "number of disabled persons", "number of deaths", "amount of loss", "equipment damage", and "other losses";
[0018] After conducting a structured analysis of the cases, the historical safety accident descriptions are transformed from human language descriptions to machine expert knowledge and entered into the laboratory fire accident case library.
[0019] Further, the rule base is constructed, and the swrlb language description is as follows:
[0020] Accident time (T3)^^swrlb:equal( )Accident level (T5)^^swrlb:equal( ), number of disabled persons (R1)^^swrlb:lessThan( ,10), death toll (R2)^^swrlb:lessThan( ,3), loss amount (R3)^^swrlb:lessThan( ,1E7), equipment damaged (R4)^^swrlb:lessThan( ,1E7), delay time (R5)^^swrlb:lessThan( ,1E7),Other losses(R6)^^swrlb:lessThan( ,1E7)→According to the accident level (?x, general accident)^Take countermeasures (?x, start the general accident emergency plan)
[0021] Accident time (T3)^^swrlb:equal( )Accident level (T5)^^swrlb:equal( ), number of disabled persons (R1)^^swrlb:MoreThan( ,10)^^swrlb:lessThan( ,50), death toll (R2)^^swrlb:MoreThan( ,3)^^swrlb:lessThan( ,10),Loss amount(R3)^^swrlb:MoreThan( ,1E7)^^swrlb:lessThan( ,5E7), equipment damaged (R4))^^swrlb:MoreThan( ,1E7)^^swrlb:lessThan( ,5E7), delay time (R5)^^swrlb:MoreThan( ,1E7)^^swrlb:lessThan( ,5E7),Other losses(R6)^^swrlb:MoreThan( ,1E7)^^swrlb:lessThan( ,5E7)→According to the accident level (?x, major accident)^Take countermeasures (?x, activate the emergency plan for major accidents)
[0022] Accident time (T3)^^swrlb:equal( )Accident level (T5)^^swrlb:equal( ), number of disabled persons (R1)^^swrlb:MoreThan( ,50)^^swrlb:lessThan( ,100), death toll (R2)^^swrlb:MoreThan( ,10)^^swrlb:lessThan( ,30), loss amount (R3)^^swrlb:MoreThan( ,5E7)^^swrlb:lessThan( ,10E7),Equipment damaged(R4))^^swrlb:MoreThan( ,5E7)^^swrlb:lessThan( ,10E7), delay time (R5)^^swrlb:MoreThan( ,5E7)^^swrlb:lessThan( ,10E7),Other losses(R6)^^swrlb:MoreThan( ,5E7)^^swrlb:lessThan( ,10E7)→According to the accident level (?x, major accident)^Take countermeasures (?x, initiate the major accident emergency plan)
[0023] Accident time (T3)^^swrlb:equal( )Accident level (T5)^^swrlb:equal( ), number of disabled persons (R1)^^swrlb:MoreThan( ,100), death toll (R2)^^swrlb:MoreThan( ,30), loss amount (R3)^^swrlb:MoreThan( ,10E7),Equipment damaged(R4))^^swrlb:MoreThan( ,10E7), delay time (R5)^^swrlb:MoreThan( ,10E7),Other losses(R6)^^swrlb:MoreThan( ,10E7)→According to the accident level (?x, particularly serious accident)^Take response measures (?x, activate the emergency plan for particularly serious accidents).
[0024] Furthermore, the characteristic attributes of the target accident case are selected, and the weights of each characteristic attribute are assigned and calculated. The following operations are performed specifically:
[0025] Analyze the relationship between the characteristic attributes of the target accident case and establish the original data matrix of the target accident case:
[0026] X=(X ij ) m×n
[0027] Among them, X is the feature attribute matrix; X ij is the attribute information value; m is the number of historical cases; n is the number of characteristic attributes;
[0028] The extreme value method is used to perform dimensionless processing on the characteristic attribute indicators of the target accident case. The calculation formula of the extreme value method is as follows:
[0029]
[0030] Among them, M j and m j are the maximum and minimum values of the j-th attribute respectively;
[0031] Calculate the feature weight or contribution of the i-th case under the j-th feature attribute:
[0032]
[0033] Calculate the entropy value of the jth feature attribute:
[0034]
[0035] Among them, e j ≥0, k>0, k=lnm;
[0036] Calculate the difference coefficient g of the jth feature attribute j :
[0037] g j =1-e j
[0038] Calculate the relative weight of the feature attributes for the target accident case and determine the weight w of each attribute for the target accident case j :
[0039]
[0040] Further, the similarity between the target accident case and the historical case is calculated based on the weight of the obtained feature attributes. The following specific operations are performed:
[0041] Similarity calculation of numerical type feature attributes:
[0042] For numeric type attribute A k , k=1,2,3…m, first find A k The numerical range of [A kmin , A kmax ], then for A k Type Attribute X i With X j The similarity calculation formula is:
[0043]
[0044] Boolean type feature attribute similarity calculation:
[0045] For a Boolean attribute A k , k=1,2, then for A k Attribute X i With X j The similarity calculation formula is:
[0046]
[0047] Hierarchical enumeration type feature attribute similarity calculation:
[0048] For enumeration type attribute A k , k=1,2,3…m, then for A k Attribute X i With X j The similarity calculation formula is:
[0049]
[0050] Among them, Ord represents A k In the sorting number of the entire enumeration set, Card represents the number of elements in the enumeration set;
[0051] Character type feature attribute similarity calculation:
[0052] For any two strings D1 and D2, the vector space composed of all their valid keywords is V = {A1, A2, ..., A n}, where A i is a valid keyword, then the similarity between string attributes D1 and D2 is calculated as:
[0053]
[0054] Where ui refers to the vector V1={u1,u2,…,u n}, the effective keyword A i The number of times it appears in string D1; z i The vector V2 in the string D2 = {z1,z2,…,z n}, the effective keyword A i The number of times it appears in string D2;
[0055] After calculating the local similarity of each attribute, the overall similarity of the case is calculated comprehensively. The calculation formula for the overall similarity Sim(C1, C2) of case C1 and case C2 is:
[0056]
[0057] Another object of the present invention is achieved by the following technical solutions:
[0058] A laboratory safety accident emergency decision-making system based on CBR-RBR hybrid reasoning mechanism, including:
[0059] A writing module, used to write the target accident case into the laboratory safety accident case library;
[0060] The RBR reasoning module uses the Drools reasoning mechanism to express rule standards and expert experience knowledge in the form of production rules, build a rule base, and perform RBR reasoning on target accident cases;
[0061] The weight calculation module is used to select the characteristic attributes of the target accident case and assign weights to each characteristic attribute;
[0062] A similarity calculation module is used to calculate the similarity between the target accident case and the historical case according to the weight of the obtained feature attributes;
[0063] In the decision generation module, if the similarity is greater than the set threshold, the case with the maximum similarity is screened out, and the case is modified by combining the case with the maximum similarity and the decision suggestion of RBR reasoning to generate a comprehensive decision suggestion based on CBR-RBR hybrid reasoning; if the similarity is less than or equal to the set threshold, a comprehensive decision suggestion based on CBR-RBR hybrid reasoning is generated according to the target accident case and the decision suggestion of RBR reasoning.
[0064] Further, the writing module specifically performs the following operations:
[0065] The overall framework of laboratory safety accident cases is represented by {T, S, R}. The attribute settings of laboratory safety accident cases include the following:
[0066] T i, description of the scenario and status of the case, that is, how the case occurred, and the description of the time-space information of the incident, including the characteristic attributes of "project scale, accident quarter, accident time, accident factor, accident level, and other status attributes";
[0067] S i ,Emergency strategy, when an incident occurs, intelligently judge the solution to be taken. The solution mainly includes implementation strategy and corresponding emergency handling mechanism, including "general accident emergency strategy", "major accident emergency strategy", "major accident emergency strategy", "particularly major accident emergency strategy" emergency strategy characteristic attributes;
[0068] R i , accident consequences, that is, the final result of the evolution of the incident after the incident occurred and after taking or not taking countermeasures, including the characteristic attributes of "number of disabled persons", "number of deaths", "amount of loss", "equipment damage", and "other losses";
[0069] After conducting a structured analysis of the cases, the historical safety accident descriptions are transformed from human language descriptions to machine expert knowledge and entered into the laboratory fire accident case library.
[0070] Furthermore, a rule base is constructed in the RBR reasoning module, and the swrlb language description is as follows:
[0071] Accident time (T3)^^swrlb:equal( )Accident level (T5)^^swrlb:equal( ), number of disabled persons (R1)^^swrlb:lessThan( ,10), death toll (R2)^^swrlb:lessThan( ,3), loss amount (R3)^^swrlb:lessThan( ,1E7), equipment damaged (R4)^^swrlb:lessThan( ,1E7), delay time (R5)^^swrlb:lessThan( ,1E7),Other losses(R6)^^swrlb:lessThan( ,1E7)→According to the accident level (?x, general accident)^Take countermeasures (?x, start the general accident emergency plan)
[0072] Accident time (T3)^^swrlb:equal( )Accident level (T5)^^swrlb:equal( ), number of disabled persons (R1)^^swrlb:MoreThan( ,10)^^swrlb:lessThan( ,50), death toll (R2)^^swrlb:MoreThan( ,3)^^swrlb:lessThan( ,10),Loss amount(R3)^^swrlb:MoreThan( ,1E7)^^swrlb:lessThan( ,5E7), equipment damaged (R4))^^swrlb:MoreThan( ,1E7)^^swrlb:lessThan( ,5E7), delay time (R5)^^swrlb:MoreThan( ,1E7)^^swrlb:lessThan( ,5E7),Other losses(R6)^^swrlb:MoreThan( ,1E7)^^swrlb:lessThan( ,5E7)→According to the accident level (?x, major accident)^Take countermeasures (?x, activate the emergency plan for major accidents)
[0073] Accident time (T3)^^swrlb:equal( )Accident level (T5)^^swrlb:equal( ), number of disabled persons (R1)^^swrlb:MoreThan( ,50)^^swrlb:lessThan( ,100), death toll (R2)^^swrlb:MoreThan( ,10)^^swrlb:lessThan( ,30), loss amount (R3)^^swrlb:MoreThan( ,5E7)^^swrlb:lessThan( ,10E7),Equipment damaged(R4))^^swrlb:MoreThan( ,5E7)^^swrlb:lessThan( ,10E7), delay time (R5)^^swrlb:MoreThan( ,5E7)^^swrlb:lessThan( ,10E7),Other losses(R6)^^swrlb:MoreThan( ,5E7)^^swrlb:lessThan( ,10E7)→According to the accident level (?x, major accident)^Take countermeasures (?x, initiate the major accident emergency plan)
[0074] Accident time (T3)^^swrlb:equal( )Accident level (T5)^^swrlb:equal( ), number of disabled persons (R1)^^swrlb:MoreThan( ,100), death toll (R2)^^swrlb:MoreThan( ,30), loss amount (R3)^^swrlb:MoreThan( ,10E7),Equipment damaged(R4))^^swrlb:MoreThan( ,10E7), delay time (R5)^^swrlb:MoreThan( ,10E7),Other losses(R6)^^swrlb:MoreThan( ,10E7)→According to the accident level (?x, particularly serious accident)^Take response measures (?x, activate the emergency plan for particularly serious accidents).
[0075] Furthermore, the weight calculation module specifically performs the following operations:
[0076] Analyze the relationship between the characteristic attributes of the target accident case and establish the original data matrix of the target accident case:
[0077] X=(X ij ) m×n
[0078] Among them, X is the feature attribute matrix; X ij is the attribute information value; m is the number of historical cases; n is the number of characteristic attributes;
[0079] The extreme value method is used to perform dimensionless processing on the characteristic attribute indicators of the target accident case. The calculation formula of the extreme value method is as follows:
[0080]
[0081] Among them, M j and m j are the maximum and minimum values of the j-th attribute respectively;
[0082] Calculate the feature weight or contribution of the i-th case under the j-th feature attribute:
[0083]
[0084] Calculate the entropy value of the jth feature attribute:
[0085]
[0086] Among them, e j ≥0, k>0, k=lnm;
[0087] Calculate the difference coefficient g of the jth feature attribute j :
[0088] g j =1-e j
[0089] Calculate the relative weight of the feature attributes for the target accident case and determine the weight w of each attribute for the target accident case j :
[0090]
[0091] Furthermore, the similarity calculation module specifically performs the following operations:
[0092] Similarity calculation of numerical type feature attributes:
[0093] For numeric type attribute A k , k=1,2,3…m, first find A k The numerical range of [A kmin , A kmax ], then for A k Type Attribute X i With X j The similarity calculation formula is:
[0094]
[0095] Boolean type feature attribute similarity calculation:
[0096] For a Boolean attribute A k , k=1,2, then for A k Attribute X i With X j The similarity calculation formula is:
[0097]
[0098] Hierarchical enumeration type feature attribute similarity calculation:
[0099] For enumeration type attribute Ak , k=1,2,3…m, then for A k Attribute X i With X j The similarity calculation formula is:
[0100]
[0101] Among them, Ord represents A k In the sorting number of the entire enumeration set, Card represents the number of elements in the enumeration set;
[0102] Character type feature attribute similarity calculation:
[0103] For any two strings D1 and D2, the vector space composed of all their valid keywords is V = {A1, A2, ..., A n}, where A i is a valid keyword, then the similarity between string attributes D1 and D2 is calculated as:
[0104]
[0105] Where ui refers to the vector V1={u1,u2,…,u n}, the effective keyword A i The number of times it appears in string D1; z i The vector V2 in the string D2 = {z1,z2,…,z n}, the effective keyword A i The number of times it appears in string D2;
[0106] After calculating the local similarity of each attribute, the overall similarity of the case is calculated comprehensively. The calculation formula for the overall similarity Sim(C1, C2) of case C1 and case C2 is:
[0107]
[0108] Compared with the prior art, the present invention has the following advantages and beneficial effects:
[0109] 1. The CBR-RBR hybrid reasoning model constructed by the present invention combines analogical reasoning with deductive reasoning, and the CBR and RBR systems can work relatively independently, which improves the efficiency and accuracy of reasoning to a certain extent. The independence of the rule base and the case base makes it easy to maintain and update.
[0110] 2. The RBR in the present invention is capable of generating a variety of rule engine architectures for complex rules for avoiding and responding to security risks, so that the rules can achieve the effect of reasoning and can be dynamically configured.
[0111] 3. The present invention can support knowledge expression and rule reasoning, and describe the conceptual model of the security risk management model from a semantic level, which to some extent makes up for the semantic inconsistency in case reasoning and the poor scalability of case representation.
[0112] 4. In the present invention, CBR reasoning can also be assisted by RBR reasoning, and the retrieved cases can be modified on the basis of RBR.
[0113] 5. For the field of laboratory safety management, it has both empirical properties such as accident cases and expert experience, and knowledge properties such as laws, regulations, rules and standards. A single reasoning method cannot fully meet the needs of laboratory safety management. The present invention combines CBR with RBR to form a more refined hybrid reasoning. BRIEF DESCRIPTION OF THE DRAWINGS
[0114] Figure 1 The present invention is a flow chart of the method.
[0115] Figure 2 It is a schematic diagram of the laboratory safety accident case library in the present invention.
[0116] Figure 3 This is a schematic diagram of setting characteristic attributes in the present invention.
[0117] Figure 4 It is a schematic diagram of the working mechanism of the rule engine in the present invention.
[0118] Figure 5 It is a schematic diagram of the combination relationship between SWRL and the ontology in the present invention.
[0119] Figure 6 Schematic diagram of the process of calculating case similarity in the present invention.
[0120] Figure 7 It is a schematic diagram of the RBR reasoning results in the present invention.
[0121] Figure 8 A schematic diagram of assigning case weights in the present invention. DETAILED DESCRIPTION
[0122] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments in the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.
[0123] Embodiment 1:
[0124] like Figures 1 to 6 As shown, this embodiment provides a laboratory safety accident emergency decision-making method based on a CBR-RBR hybrid reasoning mechanism, including the steps of:
[0125] Write the target accident case into the laboratory safety accident case database, and perform the following operations:
[0126] The overall framework of laboratory safety accident cases is represented by {T, S, R}. The attribute settings of laboratory safety accident cases include the following:
[0127] T i , description of the scenario and status of the case, that is, how the case occurred, and the description of the time-space information of the incident, including the characteristic attributes of "project scale, accident quarter, accident time, accident factor, accident level, and other status attributes";
[0128] S i ,Emergency strategy, when an incident occurs, intelligently judge the solution to be taken. The solution mainly includes implementation strategy and corresponding emergency handling mechanism, including "general accident emergency strategy", "major accident emergency strategy", "major accident emergency strategy", "particularly major accident emergency strategy" emergency strategy characteristic attributes;
[0129] R i , accident consequences, that is, the final result of the evolution of the incident after the incident occurred and after taking or not taking countermeasures, including the characteristic attributes of "number of disabled persons", "number of deaths", "amount of loss", "equipment damage", and "other losses";
[0130] After conducting a structured analysis of the cases, the historical safety accident descriptions are transformed from human language descriptions to machine expert knowledge and entered into the laboratory fire accident case library.
[0131] The Drools reasoning mechanism is used to express rule standards and expert experience knowledge in the form of production rules, build a rule base, and perform RBR reasoning on the target accident case. The swrlb language description is as follows:
[0132] Accident time (T3)^^swrlb:equal( )Accident level (T5)^^swrlb:equal( ), number of disabled persons (R1)^^swrlb:lessThan( ,10), death toll (R2)^^swrlb:lessThan( ,3), loss amount (R3)^^swrlb:lessThan( ,1E7), equipment damaged (R4)^^swrlb:lessThan( ,1E7), delay time (R5)^^swrlb:lessThan( ,1E7),Other losses(R6)^^swrlb:lessThan( ,1E7)→According to the accident level (?x, general accident)^Take countermeasures (?x, start the general accident emergency plan)
[0133] Accident time (T3)^^swrlb:equal( )Accident level (T5)^^swrlb:equal( ), number of disabled persons (R1)^^swrlb:MoreThan( ,10)^^swrlb:lessThan( ,50), death toll (R2)^^swrlb:MoreThan( ,3)^^swrlb:lessThan( ,10),Loss amount(R3)^^swrlb:MoreThan( ,1E7)^^swrlb:lessThan( ,5E7), equipment damaged (R4))^^swrlb:MoreThan( ,1E7)^^swrlb:lessThan( ,5E7), delay time (R5)^^swrlb:MoreThan( ,1E7)^^swrlb:lessThan( ,5E7),Other losses(R6)^^swrlb:MoreThan( ,1E7)^^swrlb:lessThan( ,5E7)→According to the accident level (?x, major accident)^Take countermeasures (?x, activate the emergency plan for major accidents)
[0134] Accident time (T3)^^swrlb:equal( )Accident level (T5)^^swrlb:equal( ), number of disabled persons (R1)^^swrlb:MoreThan( ,50)^^swrlb:lessThan( ,100), death toll (R2)^^swrlb:MoreThan( ,10)^^swrlb:lessThan( ,30), loss amount (R3)^^swrlb:MoreThan( ,5E7)^^swrlb:lessThan( ,10E7),Equipment damaged(R4))^^swrlb:MoreThan( ,5E7)^^swrlb:lessThan( ,10E7), delay time (R5)^^swrlb:MoreThan( ,5E7)^^swrlb:lessThan( ,10E7),Other losses(R6)^^swrlb:MoreThan( ,5E7)^^swrlb:lessThan( ,10E7)→According to the accident level (?x, major accident)^Take countermeasures (?x, initiate the major accident emergency plan)
[0135] Accident time (T3)^^swrlb:equal( )Accident level (T5)^^swrlb:equal( ), number of disabled persons (R1)^^swrlb:MoreThan( ,100), death toll (R2)^^swrlb:MoreThan( ,30), loss amount (R3)^^swrlb:MoreThan( ,10E7),Equipment damaged(R4))^^swrlb:MoreThan( ,10E7), delay time (R5)^^swrlb:MoreThan( ,10E7),Other losses(R6)^^swrlb:MoreThan( ,10E7)→According to the accident level (?x, particularly serious accident)^Take response measures (?x, activate the emergency plan for particularly serious accidents).
[0136] Select the characteristic attributes of the target accident case and assign weights to each characteristic attribute. The following operations are performed:
[0137] Analyze the relationship between the characteristic attributes of the target accident case and establish the original data matrix of the target accident case:
[0138] X=(X ij ) m×n
[0139] Among them, X is the feature attribute matrix; X ij is the attribute information value; m is the number of historical cases; n is the number of characteristic attributes;
[0140] The extreme value method is used to perform dimensionless processing on the characteristic attribute indicators of the target accident case. The calculation formula of the extreme value method is as follows:
[0141]
[0142] Among them, M j and m j are the maximum and minimum values of the j-th attribute respectively;
[0143] Calculate the feature weight or contribution of the i-th case under the j-th feature attribute:
[0144]
[0145] Calculate the entropy value of the jth feature attribute:
[0146]
[0147] Among them, e j ≥0, k>0, k=lnm;
[0148] Calculate the difference coefficient g of the jth feature attribute j :
[0149] g j =1-e j
[0150] Calculate the relative weight of the feature attributes for the target accident case and determine the weight w of each attribute for the target accident case j :
[0151]
[0152] The similarity between the target accident case and the historical case is calculated based on the weight of the obtained feature attributes. The following operations are performed specifically:
[0153] Similarity calculation of numerical type feature attributes:
[0154] For numeric type attribute A k , k=1,2,3…m, first find A k The numerical range of [A kmin , A kmax ], then for A k Type Attribute X i With X j The similarity calculation formula is:
[0155]
[0156] Boolean type feature attribute similarity calculation:
[0157] For a Boolean attribute A k , k=1,2, then for A k Attribute X i With X j The similarity calculation formula is:
[0158]
[0159] Hierarchical enumeration type feature attribute similarity calculation:
[0160] For enumeration type attribute A k , k=1,2,3…m, then for A k Attribute X i With X j The similarity calculation formula is:
[0161]
[0162] Among them, Ord represents A k In the sorting number of the entire enumeration set, Card represents the number of elements in the enumeration set;
[0163] Character type feature attribute similarity calculation:
[0164] For any two strings D1 and D2, the vector space composed of all their valid keywords is V = {A1, A2, ..., A n}, where A i is a valid keyword, then the similarity between string attributes D1 and D2 is calculated as:
[0165]
[0166] Where ui refers to the vector V1={u1,u2,…,u n}, the effective keyword A i The number of times it appears in string D1; z i The vector V2 in the string D2 = {z1,z2,…,z n}, the effective keyword A i The number of times it appears in string D2;
[0167] After calculating the local similarity of each attribute, the overall similarity of the case is calculated comprehensively. The calculation formula for the overall similarity Sim(C1, C2) of case C1 and case C2 is:
[0168]
[0169] If the similarity is greater than the set threshold, the case with the maximum similarity is screened out, and the case is modified by combining the maximum similarity case and the decision suggestion of RBR reasoning to generate a comprehensive decision suggestion based on CBR-RBR hybrid reasoning; if the similarity is less than or equal to the set threshold, a comprehensive decision suggestion based on CBR-RBR hybrid reasoning is generated according to the target accident case and the decision suggestion of RBR reasoning.
[0170] The above method is used to construct a laboratory fire accident target accident case and perform CBR-RBR hybrid reasoning on the case to assist safety managers in making decisions.
[0171] For example, the background information of the target accident case is that a fire broke out in the laboratory of Building A at X time on X month X day of XX year, resulting in 0 deaths and 2 injuries. The cause of the accident was that the sparks generated by the failure of electrical equipment ignited the flammable chemicals, and the laboratory stopped working.
[0172] 1. Build an example to obtain decision suggestions for RBR reasoning
[0173] Create a new instance in the domain ontology model and enter the background information of the accident case into the corresponding attributes of the instance. The city where the accident occurred is City A; the time of occurrence is X-month-X-day, X-hour; the line where the accident occurred is located in Laboratory XX, Building A; the cause of the fire is "electrical equipment failure" and "hazardous chemicals and combustibles ignited"; the number of injured people is 89; the number of deaths is 15. At the same time, the economic loss and downtime duration must be entered to infer the accident level. Because the accident level is divided according to the most serious consequences of various losses, the two can be entered as smaller values or estimated values depending on the situation.
[0174] After creating the accident instance and entering the attribute values, perform RBR reasoning to complete the rule reasoning. The reasoning results are presented in the instance interface, such as Figure 7 shown.
[0175] The inference results show that the accident level of this accident is a major accident. The safety management measures that should be taken are: emergency evacuation of laboratory personnel; immediate suspension of the laboratory; opening of the smoke exhaust system; management personnel using fire extinguishers to put out the fire; fire disposal; activation of major accident emergency plans; cutting off the power supply of the subway line and spraying dry powder to put out the fire; and strengthening the cleaning of debris in the laboratory.
[0176] 2. Obtaining the maximum similarity case based on CBR reasoning
[0177] (1) Feature attribute acquisition and weight determination
[0178] According to the collected laboratory fire accidents, the accident information is classified and sorted. The basic information of the accident describes the background information including the time and city of the accident; the accident process information includes the type of accident, the factors causing the accident, and the number of injuries, deaths, economic losses and delays caused by the accident.
[0179] The laboratory fire accident case database contains 12 accident cases in total. Eight characteristic attributes, including accident level, death toll, injured number, electrical equipment failure, cable short circuit, arson, high temperature equipment, and flammable (combustible) fire in the laboratory, are selected as similarity calculation parameters, as shown in Table 1.
[0180]
[0181] Table 1 Feature attributes
[0182] The weights of each feature attribute are calculated as shown in Table 2.
[0183]
[0184]
[0185] Table 2 Feature attribute weights
[0186] (2) Case similarity calculation
[0187] Model the calculation of feature attribute similarity. According to the classification of the types of the eight feature attributes, model the similarity function of the attributes respectively, which will serve as the basis for the subsequent calculation of the local similarity of the attributes. After completing the modeling of the local similarity function of the feature attributes, start to calculate the overall similarity between the two cases. The overall similarity between the cases is the weighted sum of the local similarities. The interface for setting the feature attributes and assigning the weights is as follows: Figure 8 shown.
[0188] Through case similarity calculation, the similarity values of each case in the case library and the new case are obtained and sorted, and similar cases are screened out according to the threshold (based on multiple experimental judgments on the samples of this study and reference to the relevant threshold settings of previous studies, it is concluded that the threshold is set to 0.9 for small samples. It is more reasonable). The attribute values and measures of each similar case are shown in Table 3 below:
[0189]
[0190]
[0191] Table 3 Attribute values and measures of similar cases
[0192] 3. Case-based comprehensive decision-making and improvement of rule chains based on mixed reasoning
[0193] Combining the reasoning results of RBR and CBR, it is found that the CBR results can supplement the RBR results with measures: drive the train to the nearest station; send the injured to the hospital. After correcting the RBR results, it is concluded that the emergency measures that should be taken for this target case are: activate the emergency plan for major accidents; immediately close the laboratory; urgently evacuate the laboratory staff; send the injured to the hospital; open the smoke exhaust system; the staff use fire extinguishers to put out the fire; fire disposal; cut off the power supply of the subway line and spray dry powder to put out the fire. The post-event safety management measures that should be taken are: strengthen the cleaning of debris in the laboratory.
[0194] Embodiment 2:
[0195] This embodiment provides a laboratory safety accident emergency decision-making system based on a CBR-RBR hybrid reasoning mechanism, including:
[0196] The writing module is used to write the target accident case into the laboratory safety accident case library; the following operations are performed specifically:
[0197] The overall framework of laboratory safety accident cases is represented by {T, S, R}. The attribute settings of laboratory safety accident cases include the following:
[0198] T i , description of the scenario and status of the case, that is, how the case occurred, and the description of the time-space information of the incident, including the characteristic attributes of "project scale, accident quarter, accident time, accident factor, accident level, and other status attributes";
[0199] S i ,Emergency strategy, when an incident occurs, intelligently judge the solution to be taken. The solution mainly includes implementation strategy and corresponding emergency handling mechanism, including "general accident emergency strategy", "major accident emergency strategy", "major accident emergency strategy", "particularly major accident emergency strategy" emergency strategy characteristic attributes;
[0200] R i , accident consequences, that is, the final result of the evolution of the incident after the incident occurred and after taking or not taking countermeasures, including the characteristic attributes of "number of disabled persons", "number of deaths", "amount of loss", "equipment damage", and "other losses";
[0201] After conducting a structured analysis of the cases, the historical safety accident descriptions are transformed from human language descriptions to machine expert knowledge and entered into the laboratory fire accident case library.
[0202] The RBR reasoning module is used to use the Drools reasoning mechanism to express rule standards and expert experience knowledge in the form of production rules, build a rule base, and perform RBR reasoning on target accident cases; the swrlb language description is as follows:
[0203] Accident time (T3)^^swrlb:equal( )Accident level (T5)^^swrlb:equal( ), number of disabled persons (R1)^^swrlb:lessThan( ,10), death toll (R2)^^swrlb:lessThan( ,3), loss amount (R3)^^swrlb:lessThan( ,1E7), equipment damaged (R4)^^swrlb:lessThan( ,1E7), delay time (R5)^^swrlb:lessThan( ,1E7),Other losses(R6)^^swrlb:lessThan( ,1E7)→According to the accident level (?x, general accident)^Take countermeasures (?x, start the general accident emergency plan)
[0204] Accident time (T3)^^swrlb:equal( )Accident level (T5)^^swrlb:equal( ), number of disabled persons (R1)^^swrlb:MoreThan( ,10)^^swrlb:lessThan( ,50), death toll (R2)^^swrlb:MoreThan( ,3)^^swrlb:lessThan( ,10),Loss amount(R3)^^swrlb:MoreThan( ,1E7)^^swrlb:lessThan( ,5E7), equipment damaged (R4))^^swrlb:MoreThan( ,1E7)^^swrlb:lessThan( ,5E7), delay time (R5)^^swrlb:MoreThan( ,1E7)^^swrlb:lessThan( ,5E7),Other losses(R6)^^swrlb:MoreThan( ,1E7)^^swrlb:lessThan( ,5E7)→According to the accident level (?x, major accident)^Take countermeasures (?x, activate the emergency plan for major accidents)
[0205] Accident time (T3)^^swrlb:equal( )Accident level (T5)^^swrlb:equal( ), number of disabled persons (R1)^^swrlb:MoreThan( ,50)^^swrlb:lessThan( ,100), death toll (R2)^^swrlb:MoreThan( ,10)^^swrlb:lessThan( ,30), loss amount (R3)^^swrlb:MoreThan( ,5E7)^^swrlb:lessThan( ,10E7),Equipment damaged(R4))^^swrlb:MoreThan( ,5E7)^^swrlb:lessThan( ,10E7), delay time (R5)^^swrlb:MoreThan( ,5E7)^^swrlb:lessThan( ,10E7),Other losses(R6)^^swrlb:MoreThan( ,5E7)^^swrlb:lessThan( ,10E7)→According to the accident level (?x, major accident)^Take countermeasures (?x, initiate the major accident emergency plan)
[0206] Accident time (T3)^^swrlb:equal( )Accident level (T5)^^swrlb:equal( ), number of disabled persons (R1)^^swrlb:MoreThan( ,100), death toll (R2)^^swrlb:MoreThan( ,30), loss amount (R3)^^swrlb:MoreThan( ,10E7),Equipment damaged(R4))^^swrlb:MoreThan( ,10E7), delay time (R5)^^swrlb:MoreThan( ,10E7),Other losses(R6)^^swrlb:MoreThan( ,10E7)→According to the accident level (?x, particularly serious accident)^Take response measures (?x, activate the emergency plan for particularly serious accidents).
[0207] The weight calculation module is used to select the characteristic attributes of the target accident case and assign weights to each characteristic attribute. The following operations are performed specifically:
[0208] Analyze the relationship between the characteristic attributes of the target accident case and establish the original data matrix of the target accident case:
[0209] X=(X ij ) m×n
[0210] Among them, X is the feature attribute matrix; X ij is the attribute information value; m is the number of historical cases; n is the number of characteristic attributes;
[0211] The extreme value method is used to perform dimensionless processing on the characteristic attribute indicators of the target accident case. The calculation formula of the extreme value method is as follows:
[0212]
[0213] Among them, M j and m j are the maximum and minimum values of the j-th attribute respectively;
[0214] Calculate the feature weight or contribution of the i-th case under the j-th feature attribute:
[0215]
[0216] Calculate the entropy value of the jth feature attribute:
[0217]
[0218] Among them, e j ≥0, k>0, k=lnm;
[0219] Calculate the difference coefficient g of the jth feature attribute j :
[0220] g j =1-e j
[0221] Calculate the relative weight of the feature attributes for the target accident case and determine the weight w of each attribute for the target accident case j :
[0222]
[0223] The similarity calculation module is used to calculate the similarity between the target accident case and the historical case according to the weight of the obtained feature attributes; the following operations are performed specifically:
[0224] Similarity calculation of numerical type feature attributes:
[0225] For numeric type attribute A k , k=1,2,3…m, first find A k The numerical range of [A kmin , A kmax ], then for A k Type Attribute X i With X j The similarity calculation formula is:
[0226]
[0227] Boolean type feature attribute similarity calculation:
[0228] For a Boolean attribute A k , k=1,2, then for A k Attribute X i With X j The similarity calculation formula is:
[0229]
[0230] Hierarchical enumeration type feature attribute similarity calculation:
[0231] For enumeration type attribute A k , k=1,2,3…m, then for A k Attribute X i With X j The similarity calculation formula is:
[0232]
[0233] Among them, Ord represents A k In the sorting number of the entire enumeration set, Card represents the number of elements in the enumeration set;
[0234] Character type feature attribute similarity calculation:
[0235] For any two strings D1 and D2, the vector space composed of all their valid keywords is V = {A1, A2, ..., A n}, where A i is a valid keyword, then the similarity between string attributes D1 and D2 is calculated as:
[0236]
[0237] Where ui refers to the vector V1={u1,u2,…,u n}, the effective keyword A i The number of times it appears in string D1; z i The vector V2 in the string D2 = {z1,z2,…,z n}, the effective keyword A i The number of times it appears in string D2;
[0238] After calculating the local similarity of each attribute, the overall similarity of the case is calculated comprehensively. The calculation formula for the overall similarity Sim(C1, C2) of case C1 and case C2 is:
[0239]
[0240] In the decision generation module, if the similarity is greater than the set threshold, the case with the maximum similarity is screened out, and the case is modified by combining the case with the maximum similarity and the decision suggestion of RBR reasoning to generate a comprehensive decision suggestion based on CBR-RBR hybrid reasoning; if the similarity is less than or equal to the set threshold, a comprehensive decision suggestion based on CBR-RBR hybrid reasoning is generated according to the target accident case and the decision suggestion of RBR reasoning.
[0241] The above is only a preferred embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can make equivalent replacements or changes according to the technical solution and invention concept of the present invention within the scope disclosed by the present invention, which shall fall within the protection scope of the present invention.
Claims
1. A laboratory safety accident emergency decision-making method based on CBR-RBR hybrid reasoning mechanism, It is characterized in that Including steps, Write the target accident cases into the laboratory safety accident case database; Adopting the Drools reasoning mechanism, the rule standards and expert experience knowledge are expressed in the form of production rules, a rule base is constructed, and RBR reasoning is performed on the target accident cases; Select the characteristic attributes of the target accident case, and assign weights to each characteristic attribute; Calculate the similarity between the target accident case and the historical case based on the weight of the obtained feature attributes; If the similarity is greater than the set threshold, the case with the maximum similarity is selected, and the case is modified by combining the case with the maximum similarity and the decision suggestion of RBR reasoning to generate a comprehensive decision suggestion based on CBR-RBR hybrid reasoning; If the similarity is less than or equal to the set threshold, a comprehensive decision recommendation based on CBR-RBR hybrid reasoning is generated according to the decision recommendation of the target accident case and RBR reasoning.
2. According to claim 1, the laboratory safety accident emergency decision-making method based on the CBR-RBR hybrid reasoning mechanism, It is characterized in that The target accident case is written into the laboratory safety accident case database. The following specific operations are performed: The overall framework of laboratory safety accident cases is represented by {T, S, R}. The attribute settings of laboratory safety accident cases include the following: T i , description of the scenario and status of the case, that is, how the case occurred, description of the time-space information of the incident, including "project scale, accident quarter, accident time, accident factor, accident level, other status attributes" characteristic attributes; S i , emergency strategy, when an incident occurs, intelligently judge the solution to be taken. The solution mainly includes implementation strategy and corresponding emergency handling mechanism, including "general accident emergency strategy", "major accident emergency strategy", "major accident emergency strategy", "particularly major accident emergency strategy" emergency strategy characteristic attributes; R i , accident consequences, that is, the final result of the evolution of the incident after the incident occurred and after the response measures were taken or not taken, including the characteristic attributes of "number of disabled persons", "number of deaths", "amount of loss", "equipment damage", and "other losses"; After conducting a structured analysis of the cases, the historical safety accident descriptions are transformed from human language descriptions to machine expert knowledge and entered into the laboratory fire accident case library.
3. According to claim 1, the laboratory safety accident emergency decision-making method based on the CBR-RBR hybrid reasoning mechanism, It is characterized in that Build a rule base, the swrlb language description is as follows: Accident time Accident level Number of disabled persons Death toll Amount of loss Equipment damage Delay duration Other losses →According to the accident level (?x, general accident)^take countermeasures (?x, activate the general accident emergency plan) Accident time Accident level Number of disabled persons Death toll Amount of loss Equipment damage Delay duration Other losses →According to the accident level (?x, major accident)^take countermeasures (?x, activate the emergency plan for major accidents) Accident time Accident level Number of disabled persons Death toll Amount of loss Equipment damage Delay duration Other losses →According to the accident level (?x, major accident)^take countermeasures (?x, activate the major accident emergency plan) Accident time Accident level Number of disabled persons Death toll Amount of loss Equipment damage Delay duration Other losses →According to the accident level (?x, particularly serious accident)^ take response measures (?x, activate the emergency plan for particularly serious accidents).
4. According to claim 1, the laboratory safety accident emergency decision-making method based on the CBR-RBR hybrid reasoning mechanism, It is characterized in that Select the characteristic attributes of the target accident case and assign weights to each characteristic attribute. The following operations are performed specifically: Analyze the relationship between the characteristic attributes of the target accident case and establish the original data matrix of the target accident case: X=(X ij ) m×n Among them, X is the feature attribute matrix; X ij is the attribute information value; m is the number of historical cases; n is the number of characteristic attributes; The extreme value method is used to perform dimensionless processing on the characteristic attribute indicators of the target accident case. The calculation formula of the extreme value method is as follows: Among them, M j and m j are the maximum and minimum values of the j-th attribute respectively; Calculate the feature weight or contribution of the i-th case under the j-th feature attribute: Calculate the entropy value of the jth feature attribute: Among them, e j ≥0, k>0, k=lnm; Calculate the difference coefficient g of the jth feature attribute j : g j =1-e j Calculate the relative weight of the feature attributes for the target accident case and determine the weight w of each attribute for the target accident case j :
5. According to claim 1, the laboratory safety accident emergency decision-making method based on the CBR-RBR hybrid reasoning mechanism, It is characterized in that The following operations are performed to calculate the similarity between the target accident case and the historical case based on the weight of the obtained feature attributes: Similarity calculation of numerical type feature attributes: For numeric type attribute A k , k=1,2,3…m, first find A k The numerical range of [A kmin , A kmax ], then for A k Type Attribute X i With X j The similarity calculation formula is: Boolean type feature attribute similarity calculation: For a Boolean attribute A k , k=1,2, then for A k Attribute X i With X j The similarity calculation formula is: Hierarchical enumeration type feature attribute similarity calculation: For enumeration type attribute A k , k=1,2,3…m, then for A k Attribute X i With X j The similarity calculation formula is: Among them, Ord represents A k In the sorting number of the entire enumeration set, Card represents the number of elements in the enumeration set; Character type feature attribute similarity calculation: For any two strings D 1 and D 2 , the vector space formed by all their valid keywords is V = {A 1 ,A 2 ,…,A n }, where A i is a valid keyword, then the string attribute D 1 and D 2 The similarity is calculated as: Among them, u i Point to the string D 1 The vector V 1 = {u 1 ,u 2 ,…,u n }, the effective keyword A i The number of times it appears in string D1; z i Refers to the string D 2 The vector V 2 ={z 1 ,z 2 ,…,z n }, the effective keyword A i In string D 2 The number of times it appears in After calculating the local similarity of each attribute, the overall similarity of the case is calculated comprehensively. 1 And Case C 2 The calculation formula of the overall similarity Sim(C1,C2) is:
6. A laboratory safety accident emergency decision-making system based on CBR-RBR hybrid reasoning mechanism, It is characterized in that include, A writing module, used to write the target accident case into the laboratory safety accident case library; The RBR reasoning module uses the Drools reasoning mechanism to express rule standards and expert experience knowledge in the form of production rules, build a rule base, and perform RBR reasoning on target accident cases; The weight calculation module is used to select the characteristic attributes of the target accident case and assign weights to each characteristic attribute; A similarity calculation module is used to calculate the similarity between the target accident case and the historical case according to the weight of the obtained feature attributes; The decision generation module selects the case with the maximum similarity if the similarity is greater than the set threshold, and modifies the case by combining the case with the maximum similarity and the decision suggestion of RBR reasoning to generate a comprehensive decision suggestion based on CBR-RBR hybrid reasoning; If the similarity is less than or equal to the set threshold, a comprehensive decision recommendation based on CBR-RBR hybrid reasoning is generated according to the decision recommendation of the target accident case and RBR reasoning.
7. The laboratory safety accident emergency decision-making system based on the CBR-RBR hybrid reasoning mechanism according to claim 6, It is characterized in that The writing module specifically performs the following operations: The overall framework of laboratory safety accident cases is represented by {T, S, R}. The attribute settings of laboratory safety accident cases include the following: T i , description of the scenario and status of the case, that is, how the case occurred, description of the time-space information of the incident, including "project scale, accident quarter, accident time, accident factor, accident level, other status attributes" characteristic attributes; S i , emergency strategy, when an incident occurs, intelligently judge the solution to be taken. The solution mainly includes implementation strategy and corresponding emergency handling mechanism, including "general accident emergency strategy", "major accident emergency strategy", "major accident emergency strategy", "particularly major accident emergency strategy" emergency strategy characteristic attributes; R i , accident consequences, that is, the final result of the evolution of the incident after the incident occurred and after the response measures were taken or not taken, including the characteristic attributes of "number of disabled persons", "number of deaths", "amount of loss", "equipment damage", and "other losses"; After conducting a structured analysis of the cases, the historical safety accident descriptions are transformed from human language descriptions to machine expert knowledge and entered into the laboratory fire accident case library.
8. According to claim 6, the laboratory safety accident emergency decision-making system based on the CBR-RBR hybrid reasoning mechanism, It is characterized in that The rule base is constructed in the RBR reasoning module, and the swrlb language description is as follows: Accident time Accident level Number of disabled persons Death toll Amount of loss Equipment damage Delay duration Other losses →According to the accident level (?x, general accident)^take countermeasures (?x, activate the general accident emergency plan) Accident time Accident level Number of disabled persons Death toll Amount of loss Equipment damage Delay duration Other losses →According to the accident level (?x, major accident)^take countermeasures (?x, activate the emergency plan for major accidents) Accident time Accident level Number of disabled persons Death toll Amount of loss Equipment damage Delay duration Other losses →According to the accident level (?x, major accident)^take countermeasures (?x, activate the major accident emergency plan) Accident time Accident level Number of disabled persons Death toll Amount of loss Equipment damage Delay duration Other losses →According to the accident level (?x, particularly serious accident)^ take response measures (?x, activate the emergency plan for particularly serious accidents).
9. The laboratory safety accident emergency decision-making system based on the CBR-RBR hybrid reasoning mechanism according to claim 6, It is characterized in that The weight calculation module specifically performs the following operations: Analyze the relationship between the characteristic attributes of the target accident case and establish the original data matrix of the target accident case: X=(X ij ) m×n Among them, X is the feature attribute matrix; X ij is the attribute information value; m is the number of historical cases; n is the number of characteristic attributes; The extreme value method is used to perform dimensionless processing on the characteristic attribute indicators of the target accident case. The calculation formula of the extreme value method is as follows: Among them, M j and m j are the maximum and minimum values of the j-th attribute respectively; Calculate the feature weight or contribution of the i-th case under the j-th feature attribute: Calculate the entropy value of the jth feature attribute: Among them, e j ≥0, k>0, k=lnm; Calculate the difference coefficient g of the jth feature attribute j : g j =1-e j Calculate the relative weight of the feature attributes for the target accident case and determine the weight w of each attribute for the target accident case j :
10. The laboratory safety accident emergency decision-making system based on the CBR-RBR hybrid reasoning mechanism according to claim 6, It is characterized in that The similarity calculation module specifically performs the following operations: Similarity calculation of numerical type feature attributes: For numeric type attribute A k , k=1,2,3…m, first find A k The numerical range of [A kmin , A kmax ], then for A k Type Attribute X i With X j The similarity calculation formula is: Boolean type feature attribute similarity calculation: For a Boolean attribute A k , k=1,2, then for A k Attribute X i With X j The similarity calculation formula is: Hierarchical enumeration type feature attribute similarity calculation: For enumeration type attribute A k , k=1,2,3…m, then for A k Attribute X i With X j The similarity calculation formula is: Among them, Ord represents A k In the sorting number of the entire enumeration set, Card represents the number of elements in the enumeration set; Character type feature attribute similarity calculation: For any two strings D 1 and D 2 , the vector space formed by all their valid keywords is V = {A 1 ,A 2 ,…,A n }, where A i is a valid keyword, then the string attribute D 1 and D 2 The similarity is calculated as: Among them, u i Point to the string D 1 The vector V 1 = {u 1 ,u 2 ,…,u n }, the effective keyword A i The number of times it appears in string D1; z i Refers to the string D 2 The vector V 2 ={z 1 ,z 2 ,…,z n }, the effective keyword A i In string D 2 The number of times it appears in After calculating the local similarity of each attribute, the overall similarity of the case is calculated comprehensively. 1 And Case C 2 The calculation formula of the overall similarity Sim(C1,C2) is:
