CN115952919A - Intelligent risk prediction method based on process mining - Google Patents
Intelligent risk prediction method based on process mining Download PDFInfo
- Publication number
- CN115952919A CN115952919A CN202310059941.1A CN202310059941A CN115952919A CN 115952919 A CN115952919 A CN 115952919A CN 202310059941 A CN202310059941 A CN 202310059941A CN 115952919 A CN115952919 A CN 115952919A
- Authority
- CN
- China
- Prior art keywords
- type
- monitor
- abnormal
- emergency management
- data
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
- Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
- Y02P90/30—Computing systems specially adapted for manufacturing
Landscapes
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a risk intelligent prediction method based on process mining, which comprises the following steps: reading real-time monitoring data of a sensor, and preprocessing; step two, judging whether the real-time monitoring data is abnormal or not and judging the abnormal grade; step three, judging whether the abnormity is false alarm; generating a flow model candidate set according to the abnormal type; step five, screening the process models in the candidate set according to the abnormity in a grading manner; step six, calculating the probability that the screened emergency management process model is suitable for the current abnormity; and seventhly, sequencing the prediction results and sending out prediction early warning. The method can not only send out early warning by monitoring the abnormal condition of data, but also obtain the possible future events, dangerous situations, sensor value change and the probability of occurrence thereof by using the process model base and the emergency management knowledge base, and even the historical processing mode and the effect of the specific condition as decision support information to be output.
Description
Technical Field
The invention belongs to the technical field of safe production and emergency management, relates to a risk prediction method, and particularly relates to a process mining-based risk intelligent prediction method taking a chemical industry park as a core.
Background
By the end of 2020, 616 families are shared in national key chemical industry parks or industrial parks with petroleum and chemical industry as the leading industries. Taking a traditional park as an example, once an emergency public event occurs in the park, most of work such as making of an overall emergency plan, descending of a management mechanism, personnel management and control and the like is still completed by manpower, rapid management response and data statistics are difficult to realize, timely and accurate collection of information cannot be guaranteed, construction in the aspects of evaluation, reconstruction, capacity recovery and the like of the influence of the overall event is not complete enough, various emergency resources are unreasonably configured, and emergency rescue capability is weak. Although a set of own regulation system is established in China aiming at the safety supervision of dangerous chemicals, and a series of management regulations and standards of the dangerous chemicals are correspondingly established, which plays a positive role in effectively controlling and preventing the harm of the dangerous chemicals, the current safety supervision system of the dangerous chemicals in China cannot adapt to the needs of market economy, and the system construction has the outstanding problems of law lag, system standard contradiction, international deviation and the like. Secondly, the enterprise prevents that consciousness is weak, and early warning and monitored control system is imperfect. Moreover, when dangerous chemicals accidents happen, the dangerous chemicals are still in the face of the warehouse, and unnecessary loss is caused.
Risk refers to the possibility that some loss occurs in a certain environment for a certain period of time. The risk is composed of elements such as risk factors, risk accidents and risk losses. Risk has two definitions: one definition emphasizes risk performance as uncertainty; while another definition emphasizes the uncertainty that the risk represents as a loss. Risks, both risk and loss-causing uncertainties, particularly in chemical parks, for example, in chemical wastewater, excessive organic phosphorus can have serious ecological/economic negative effects on the surrounding environment, but it is uncertain when such overproof occurs and the severity of the hazard. If the risk can be predicted, specific measures to reduce the probability of occurrence of the risk can be implemented for specific situations to reduce the negative effects and seek better safety level and economic benefit.
The risk intelligent prediction is a method for outputting decision support after data processing is carried out by combining advanced technical means such as process mining and the like. According to real-time major emergency incident information and synchronous historical data acquired by monitoring, intelligent analysis and risk identification are carried out, scientific comprehensive study and judgment is carried out on the possible hazard range and disaster derivation with potential risks, meanwhile, more accurate prediction and early warning results are provided through a graphical display function, and the emergency treatment is strived for first. The intelligent prediction method needs to combine modules such as a process model base, an emergency management knowledge base and the like, and finally outputs the modules in the form of decision support information.
The process model library is a data warehouse for storing information such as processing measures, emergency resource configuration conditions, processing effects and the like in the processing process by digging an emergency management plan process model from historical emergency processing event logs of the chemical industry park. Different exception types (e.g., SO) may be retrieved from the flow model library 2 Leaks/excessive pipe temperatures). After the type is determined, a process model conforming to the current anomaly may be further retrieved according to the severity of the anomaly, such as a deviation value of the excess concentration. According to a plurality of screening conditions, one or more possible emergency management processes meeting certain requirements can be finally determined in the process model library. The emergency management knowledge base is a data warehouse for storing knowledge of abnormality judgment, abnormal event correlation and the like in the chemical industry park. Knowledge of emergency management can be obtained from the emergency management knowledge base, for example, how to judge whether the real-time monitored data is abnormal or not and whether the data is abnormal or notThe frequent severity, the probability suitable for an emergency management process aiming at a certain abnormality and the like.
Process mining refers to a technique of acquiring process knowledge from data collected by an information system for actual business execution and extracting a structured process model, which can discover, monitor and improve actual system behavior. The alpha algorithm is a milestone algorithm for process mining. It constructs a process model represented by a Petri net by finding out the activities in the process and mining four basic relations in the log, namely following, concurrent, causal and irrelevant relations. However, since the α algorithm cannot solve the noise problem and the identification anomaly problem, researchers have proposed a series of new algorithms with different angles to solve the problems, such as an inductive mining algorithm, a heuristic mining algorithm, and the like.
The multi-source heterogeneous data fusion analysis refers to that different data from multiple sources such as meteorology, sensors, RFID, cameras and edge servers and attributes/dimensions are involved in emergency management of a chemical industry park.
Although the technologies are widely applied to actual scenes, the application scenes of the process mining technology need to be expanded in consideration of the characteristics and the requirements of the emergency management field, and a new risk intelligent prediction method suitable for emergency management is provided.
Disclosure of Invention
The invention provides a risk intelligent prediction method based on process mining, and aims to solve the problems that at present, chemical industry accidents frequently cause loss of human and property, meanwhile, an enterprise risk prediction means is in a primary stage, emergency treatment management is incomplete, and how to realize risk prediction by using advanced technology and existing domain knowledge. The method is based on the data of a process model base and an emergency knowledge base based on the Internet of things and big data, is used for analyzing the safety risk of the park from real-time data generated by the park, not only can send out early warning by monitoring abnormal conditions of the data, but also can obtain events, dangerous situations, sensor value changes and the occurrence probability thereof which are possibly generated in the future by using the process model base and the emergency management knowledge base, and even the historical processing mode and the effect of the specific conditions are taken as decision support information to be output.
The purpose of the invention is realized by the following technical scheme:
a risk intelligent prediction method based on process mining comprises the following steps:
step one, reading real-time monitoring data of a sensor, and preprocessing
Reading real-time monitoring data from each sensor of the chemical industry park, fusing and analyzing multi-source heterogeneous data of the sensors to obtain standardized and formatted data D (Type) easy to process i ,Monitor j ) Data indicating that the monitoring subject sensor j monitors the abnormality type i;
step two, judging whether the real-time monitoring data is abnormal or not and judging the abnormal grade
Inquiring an emergency management knowledge base EMKB, retrieving abnormal data judgment knowledge with an abnormal Type i and a sensor j, and retrieving D (Type) i ,Monitor j )∈K(Type i ,Monitor j ,Level k ) If yes, the current abnormity is judged to be Level k Abnormal degree, if it is Level 0 If no abnormity exists, returning to the step one for continuous monitoring;
step three, judging whether the abnormity is false alarm
Inquiring an emergency management knowledge base EMKB, retrieving alarm property judgment knowledge K (Type) with the abnormal Type i and the sensor j i ,Monitor j ,Monitor else )＝T(D(Type i ,Monitor j )&&D(Type i ,Monitor elses )|t∈[0,5s]) I.e. within 5 seconds, D (Type) i ,Monitor j ) And D (Type) i ,Monitor elses ) If one of the abnormal degree values is Level0, then K (Type) i ,Monitor j ,Monitor else ) If the signal is false (Flase), judging the signal to be false alarm and returning to the first step for continuous monitoring; otherwise K (Type) i ,Monitor j ,Monitor else ) True (True);
step four, generating a flow model candidate set according to the abnormal type
According to Type i 、Monitor j Monitoring subject with access exception type of iAdding the flow models into a candidate set S according to an emergency pipe flow model EMPM with the sensor being j i ＝{EMPM(Type i ,Monitor j )|EMPM∈PML}；
Step five, screening the process models in the candidate set according to the abnormity in a grading way
Inquiring an emergency management knowledge base EMKB, retrieving abnormal data judgment knowledge with an abnormal Type i and a sensor j, and retrieving D (Type) i ,Monitor j )∈K(Type i ,Monitor j ,Level k ) In the candidate set S i Keep exception Level k The process model of (2) deletes the process models of other abnormal levels to obtain the updated candidate set S i ＝{EMPM(Type i ,Monitor j ,Level k )|EMPM∈PML}；
Step six, calculating the probability that the screened emergency management process model is suitable for the current abnormity
According to the candidate set S i The total number N of process instances in (1), and the number N of process instances of each model EMPMi Respectively calculating an emergency management model EMPM i For abnormal data D (Type) i ,Monitor j ) Probability P of the situation EMPMi ，P EMPMi ＝n EMPMi /N×100％；
Seventhly, sequencing the prediction results and sending out prediction early warning
According to the final candidate set S i Probability P of each process model EMPMi And ordering the prediction from big to small, and sending the prediction as a result.
Compared with the prior art, the invention has the following advantages:
1. the invention combines the requirements of emergency management of a chemical industrial park, researches a multi-source heterogeneous data fusion and analysis technology, constructs a data fusion bus by taking an emergency management business process as a core, and filters, cleans, analyzes and standardizes the acquired data, thereby not only linking a physical world with a computer world, but also connecting hardware, embedded software and middleware at the edge of a network with an enterprise system, so that the data formed by distributed physical events can be uniformly transmitted to an emergency management platform, and lays a foundation for subsequent data utilization.
2. The method takes the potential risk of the chemical industry park as the background, adopts the data of the sensor in the chemical industry park information system to obtain the real-time data, and combines the emergency management knowledge base to judge the abnormal condition of the data and the nature of alarming when the data is abnormal, thereby avoiding the delay and the inaccuracy of the artificial judgment of the risk condition and the artificial selection of the emergency process. Meanwhile, the solution of the emergency management plan for accurately identifying the risks and intelligently predicting and processing the risks in the chemical industry park can be helped, the probability of the emergency management plan recommended and adopted in the face of the current risks and the probability of the emergency management plan adapted to the current situation are described, the response efficiency and the processing effect of the chemical industry park for dealing with the safety risks are improved, and the better response emergency situation in the chemical industry park is helped to reduce the loss.
3. When the data abnormity and the alarm property are judged, the invention provides the false alarm for identifying the condition that some abnormity of the data is not risky, and can help the park reduce unnecessary risks and the safety processing process.
Drawings
FIG. 1 is an architecture diagram of an intelligent risk prediction method based on process mining according to the present invention;
FIG. 2 is a schematic view of step (3);
FIG. 3 is a schematic diagram of steps (4) and (5).
Detailed Description
The technical solution of the present invention is further described below with reference to the accompanying drawings, but not limited thereto, and any modification or equivalent replacement of the technical solution of the present invention without departing from the spirit and scope of the technical solution of the present invention shall be covered by the protection scope of the present invention.
In order to show the specific implementation of the present invention more clearly, some related concepts are described first:
a large number of Event logs Log are collected in an emergency management platform based on the Internet of things, wherein the Event logs include executed emergency processing flow instances Trace, and the Trace is composed of a series of Event events which occur according to time sequence.
The event E is the basic element that makes up the Trace,is represented by E s /E c = (eName, t, actor, attrVal), wherein: 1) E s And E c Representing different life cycle states of the same event E, each event having a start and an end state, E s Indicating the start state of event E, E c Indicating the end state of event E, E s And E c eName, t, actor, attrVal of (a) are all the same; 2) eName is the name of event E, which can be represented by E (eName); 3) t is the timestamp of event E, indicating the time at which the event occurred; 4) an operator is a person, organization, or automated resource that participates in event E. The operator may be denoted as operator = { service provider, partner, customer, automation resource }, and the automation resource may be an external entity such as a sensor, software system, etc. 5) attrVal { (attr) 1 ,val 1 ),...,(attr n ,val n ) Is the set of attributes and values of event E, attr i Is the ith attribute, val, of event E i Is attribute attr i The value of (c).
An emergency process flow instance T consists of a series of chronologically ordered events, which can be denoted as T (tName) =<E(1) (s/c) ,E(2) (s/c) ,...,E(m) (s/c) >Wherein: 1) the tName is the unique identifier of the process instance T; 2) E (i) is the ith event in process instance T.
A value arc represents a flow of value between an activity and a participant. When an activity is performed by a participant, there is a value arc between them, denoted va = (a, ar, v) in ,v out ) Wherein: 1) a is an activity; 2) ar is actor to perform a; 3) v. of in Is the value or cost entered into ar when engaged in activity a; 4) v. of out Is the output value obtained after ar completes activity a.
An emergency management process model EMPM in a process model library PML is a value-oriented business process BPMN model, and is defined as EMPM = (a, actor, VA, edge, GXOR, GAND, start, end), where: 1) A is the set of service activities; 2) Edge = { (head, tail) } is a set of sequential streams, where (head, tail) represents a sequential stream from head to tail, head e { a, GXOR, GAND, start }, tail e { a, GXOR,GAND, end }; 3) GXOR and GAND represent exclusive and parallel gateways, respectively; 4) start and end represent the start and end events of the flow, respectively. 5) Actor is the set of participants in the service flow model; 6) VA is the set of merit arcs between a and Actor. Emergency management processes for storing different abnormal types of types in process model library PML in a partitioned manner, e.g. SO processing 2 All emergency management processes of the leakage are stored in a set S Sulfur dioxide In (1), PML = { S = { [ S ] typei I typei is a distinct type }. And an abnormal type process model is classified and stored according to the abnormal degree, for example, the concentration of sulfur dioxide leakage within 10%, 50% and 100% of the standard value corresponds to different emergency management process models, which are expressed as S Sulfur dioxide ＝{EMPM(SO 2 ,[i,j]) And | i, j is the lower limit and the upper limit of the sulfur dioxide concentration }.
The emergency management knowledge base EMKB is the existing domain knowledge, and the knowledge in the EMKB is divided into abnormal data judgment knowledge and alarm property judgment knowledge. The abnormal data determination knowledge is an abnormal range of data (description of state, numerical value, and the like) of each event in the flow model, and is represented by K (Type) i ,Monitor j ,Level k )＝{t|t∈S i In which S is i Is the decision range of the monitoring subject sensor j when the monitoring type i is at the severity level K, for example, K (SO) 2 ,M 1 ,III)＝[300,+∞]Indicates if the sensor M is 1 Judging the three-stage abnormality when the monitored sulfur dioxide concentration is more than or equal to 300 units; the alarm nature determination knowledge is determination knowledge for determining whether some abnormality of the sensor data is represented as an error, a false alarm behavior such as an error, or a true alarm, and is represented as K (Type) i ,Monitor j ) = (Boolean expression), e.g. K (SO) 2 ,M 1 ) = (75% of observed concentration values in 1 minute are greater than upper normal limit)&&Adjacent two observations differ by no more than 100%).
Data D of real-time monitoring Ai The data are divided into two types of human participation activity signals and unmanned sensor data, and represent A i Real-time monitoring data of activity events. The data value range of the real-time monitoring of the activity signals participated by people is { true, false }, and the data range of the unmanned sensor isE.g. D A2 ＝0.9mol/L。
In the invention, the intelligent risk prediction method based on process mining comprises the following steps:
(1) And reading the real-time monitoring data of the sensor, and preprocessing. Real-time monitoring data are read from all sensors in the chemical industry park, and the data types of all the sensors and the abnormal types of monitoring are different. The existing data fusion analysis technology is utilized to analyze the multisource heterogeneous data of the sensor to obtain the standardized and formatted data D (Type) which is easy to process i ,Monitor j ) And data indicating that the monitoring subject sensor j monitors the abnormality type i.
(2) And judging whether the real-time monitoring data is abnormal or not and judging the abnormal grade. Accessing an Emergency Management Knowledge Base (EMKB), searching abnormal data judgment knowledge with an abnormal Type i and a sensor j, and searching D (Type) i ,Monitor j )∈K(Type i ,Monitor j ,Level k ) If yes, the current abnormity is judged to be Level k Abnormality of degree. If it is Level 0 If no abnormity exists, the method returns to (1) to continuously monitor.
(3) And judging whether the abnormity is false alarm or not. Inquiring an emergency management knowledge base EMKB, retrieving alarm property judgment knowledge K (Type) with an abnormal Type of i and a sensor of j i ,Monitor j ,Monitor else )＝T(D(Type i ,Monitor j )&&D(Type i ,Monitor elses )|t∈[0,5s]) I.e. within 5 seconds, D (Type) i ,Monitor j ) And D (Type) i ,Monitor elses ) If one of the abnormal degree values is Level0, K (Type) i ,Monitor j ,Monitor else ) If the result is false (Flase), judging the result to be false alarm and returning to the first step for continuous monitoring; otherwise K (Type) i ,Monitor j ,Monitor else ) True (True).
(4) And generating a flow model candidate set according to the exception type. According to Type i 、Monitor j Accessing an emergency pipe flow model EMPM with an abnormal type of i and a monitoring subject sensor of j, and adding the flow models into a candidate set S i ＝{EMPM(Type i ,Monitor j )|EMPM∈PML}。
(5) And screening the process models in the candidate set according to the abnormity in a grading way. Inquiring an emergency management knowledge base EMKB, retrieving abnormal data judgment knowledge with an abnormal Type i and a sensor j, and retrieving D (Type) i ,Monitor j )∈K(Type i ,Monitor j ,Level k ). In the candidate set S i Keep exception Level k The process model of (2) deletes the process models of other abnormal levels to obtain the updated candidate set S i ＝{EMPM(Type i ,Monitor j ,Level k )|EMPM∈PML}。
(6) And calculating the probability that the screened emergency management process model is suitable for the current abnormity. According to the candidate set S i The total number N of process instances in (1), and the number N of process instances of each model EMPMi Respectively calculating an emergency management model EMPM i Applicable to abnormal data D (Type) i ,Monitor j ) Probability P of the situation EMPMi ＝n EMPMi /N×100％。
(7) And sequencing the prediction results and sending out prediction early warning. According to the final candidate set S i Probability P of each process model EMPMi And ordering the prediction from big to small, and sending the prediction as a result.
In the invention, the complete intelligent risk prediction steps are shown in fig. 1, and the specific description of the algorithm is shown in table 1.
TABLE 1
The invention provides a modeling method of an emergency management process model EMPM, which is characterized in that a chemical industry park is used as a background, various event log data collected by various devices and terminal systems of an Internet of things emergency management platform are used as a basis, process examples are analyzed and counted and then abstracted into process variants, the process variants are combined into a process model, value information is extracted from the data, an emergency management process model is further built, and a process example set of each classified main body is repeated so as to build a process model base. The specific implementation steps are as follows:
step 1, data preprocessing:
cleaning, filtering and classifying event data stored in an emergency management platform to obtain an event log set log;
taking a process example set of a classification subject S as a sample, analyzing and counting a process example T in the sample to obtain a set { (T) 1 ,n 1 ),(T 2 ,n 2 ),...,(T k ,n k ) Where T is i Is a specific example of a procedure, n i The number of the process examples is as follows:
step 21, taking a process instance set of a classification subject S as a sample, mapping events of a process instance T into corresponding activities a with the same name, and mining an activity sequence relation and a parallel relation according to a time sequence relation so as to determine the activities a and the relation thereof contained in the process instance T;
step 22, merging the included activities a and the process instances T with the same relationship, and counting the number of the same instances to obtain a set { (T) 1 ,n 1 ),(T 2 ,n 2 ),...,(T k ,n k )}；
Step 3, abstracting the process example into a process variant:
determining a set of process instances { T } 1 ,T 2 ,...,T k Merging the process instances T with the same cyclic relationship, and abstracting the process instance set into a process variant set { VI ] 1 ,VI 2 ,...,VI p }；
And 4, merging the process variants into a process model PM:
for variant set { VI 1 ,VI 2 ,...,VI p Combining every two process variants in sequence to determine the selection relation of the variants so as to obtain a complete process model PM, wherein the specific steps are as follows:
step 41, set of variants { VI 1 ,VI 2 ,...,VI p The flow in (1) is changedCombining every two variants in sequence, integrating and reserving the same part in the two variants, and merging the process variants with the selection relation, wherein the different parts exist in the form of the selection relation;
step 42, extracting all unrepeated events, activities and gateways in the original flow variant, and identifying the gateways according to the connected elements;
step 43, traversing each edge (head, tail) in all the original process variants, adding the edge into the new process model without repetition, wherein different parts exist among the variants, and the new process model may have a situation that a non-gateway element has a plurality of edges, a selection gateway needs to be added for the element, the head in all the edges of the element is replaced by the selection gateway, all the edges of the element are deleted, and finally, an edge pointing to the selection gateway by the element is added, so as to obtain a complete process model PM;
extracting from the corresponding event E, based on each activity in the set of activities obtained in step 2, a respective participant ar and an input value v in And an output value v out Thereby obtaining a value arc set VA, which comprises the following steps:
step 51, based on the activity a obtained in the step two, extracting corresponding participant ar from the corresponding event E, and extracting the input value v of the ar participating in the activity a from the set attrVal of the attributes and values of the event E in And an output value v out Thereby determining the value arc va = (a, ar, v) between them in ,v out )；
Step 52, executing the operation of the step five to each activity, and obtaining a value arc set VA;
step 6, constructing an emergency management process model EMPM:
marking each value arc in the value arc set VA to a corresponding activity a in the process model PM i In the above step, an emergency management process model EMPM is established, and the specific steps are as follows:
step 61, if the value arc (a) i ,ar j ,v in ,v out ) Middle activity a i And the flow ofActivity in model PM a i If the value arc is the same, then label the value arc to the corresponding activity a in the process model PM i The above step (1);
step 62, repeating step 61, and establishing an emergency management process model EMPM = (a, actor, VA, edge, GXOR, GAND, start, end), wherein: a is the set of service activities; edge = { (head, tail) } is a set of sequential streams, head ∈ { a, gxxor, GAND, start }, tail ∈ { a, gxxor, GAND, end }; GXOR and GAND represent exclusive and parallel gateways, respectively; start and end represent the start and end events of the flow, respectively; actor is the set of participants in the service flow model; VA is the set of value arcs between a and Actor;
step 7, forming a process model library:
adding the emergency management process model EMPM of the obtained classification subject S into a process model library; and repeating the step 2 to the step 6 for the process example set of each classification main body, and adding the obtained emergency management process model corresponding to each classification entity into the process model library so as to obtain a complete process model library.
In the above steps, step 3, step 4 and step 5 are parallel execution relations, and the execution process algorithm of the whole method is specifically described as shown in table 2.
The invention provides a definition and determination process of a candidate set of an emergency management process model, which comprises the following steps: the candidate set is a set of emergency management processes which are possibly adapted to a certain risk type and is marked as S i ＝{EMPM(Type i ,Monitor j ) I EMPM belongs to PML }. The determination process comprises the steps of screening abnormal types in a flow model library, then screening a sensor main body, and finally screening abnormal grades.
In the invention, the finally output risk prediction emergency plan has the following characteristics: firstly, aiming at the risk condition, matching the current risk abnormality; secondly, the probability of being adopted is the maximum when the abnormality of the same condition is processed in history; thirdly, when the abnormity of the same condition is processed in history, the risk abnormity can be effectively solved.
Example (b):
in this embodiment, monitoring data of an emergency management platform in a chemical industry park is used as a sample, tens of millions of data volumes are stored in the platform, and a process model library and an emergency management knowledge base are not the key points of the present invention, and subsequent steps of the present invention are verified mainly by taking a data set of an operation process of primary intelligent risk prediction as an example:
the emergency management platform of the chemical industry park is provided with a process model library. The process model library orderly stores the emergency management processes used historically. The first-class classification index is of an abnormal type, for example, the sulfur dioxide leakage and the nitric oxide leakage are different abnormal types, and an emergency management process for treating the sulfur dioxide leakage and an emergency management process for treating the nitric oxide leakage are stored separately; the second-level classification indexes are sensors, for example, an emergency management process adopted by data abnormity of a sulfur dioxide concentration sensor in a warehouse in a chemical industry park is different from an emergency management process adopted by data abnormity of a sulfur dioxide concentration sensor in a factory, and a process model library is used for separately storing process models corresponding to different sensors under the same abnormity type; the third-level classification index is an abnormal grade, for example, the data abnormality degree of the same sulfur dioxide concentration sensor in a warehouse in a chemical industry park is different (for example, two times exceeding the standard is compared with ten times exceeding the standard), the emergency management processes adopted in the history are different, and the emergency management processes are also stored separately.
An emergency management knowledge base is arranged in an emergency management platform of the chemical industry park. Emergency management knowledge required by intelligent risk prediction is orderly stored in the emergency management knowledge base. The emergency management knowledge comprises anomaly judgment knowledge and alarm property judgment knowledge of each sensor. The anomaly determination knowledge refers to the data anomaly range of each risk prediction sensor in the chemical industry park. The alarm property judgment knowledge is used for judging whether the real-time data of the sensor represents the risk or not by combining the characteristics of the sensor such as precision, stability and reliability and integrating the occurrence of the risk. And the abnormity judgment knowledge is searched according to three parameters of the abnormity type, the sensor and the abnormity grade, and whether the alarm is effective or not is judged after required real-time input.
As shown in table 3, the detailed procedure is as follows:
the first step is as follows: inputting sensor real-time monitoring data, analyzing and fusing to obtain standard formatted data D 1 (SO 2 Sensor 1) =0.1mol/L.
The second step is that: accessing the emergency management repository, retrieving D (SO) 2 Sensor 1) abnormality determination knowledge (abnormal data range and level), D 1 (SO 2 Sensor 1) is at [0.05,0.15]mol/L of Level 2 It is a secondary anomaly.
The third step: accessing the emergency management repository, retrieving D (SO) 2 Knowledge of the nature of the alarm of sensor 1), K 1 (Sulfur dioxide, sensor 1, sensor 2) and inputting data of sensor 1 and sensor 2 into K 1 The data of the two sensors are both at Level within 5 seconds 2 Within the range, abnormal data D of the sensor 1 judged as an effective alarm is obtained 1 There is a risk. The specific process is shown in fig. 2.
The fourth step: accessing a process model library, and performing process model matching according to the abnormal type of sulfur dioxide leakage and the sensor of the sensor 1 to obtain a process model candidate set S 1 ＝{EMPM(SO 2 Leakage, sensor 1) | EMPM ∈ PML }.
The fifth step: the candidate set obtained in the last step is Level according to the abnormal Level 2 Screening out the final candidate set S meeting the abnormal grade 1 ＝{EMPM(SO 2 Leakage, sensor 1,level 2 ) I EMPM belongs to PML }. As shown in fig. 3, it is the implementation process of the fourth step and the fifth step.
And a sixth step: to S 1 And (4) carrying out appropriate probability calculation on each flow model. And calculating the frequency of each emergency management flow as the probability of the suitable emergency management flow adopted to process the current abnormal risk according to the frequency of the flow instance of each flow model in the history and the frequency of each flow model. P is EMPM1 ＝70％，P EMPM2 ＝20％，P EMPM3 ＝10％。
The seventh step: sequencing the probabilities of all the emergency management processes in the sixth step from large to small, P EMPM1 And maximally, as a solution for intelligently predicting risks and an emergency management solution, a prediction result is given in a decision support suggestion form.
TABLE 3
The invention provides a risk intelligent prediction method based on flow mining according to the characteristics of flow data and existing structural functions in an emergency management platform of a chemical park. By the method, early warning can be given out by monitoring abnormal conditions of data, events, dangerous cases, sensor value changes and the probability of the changes can be obtained by using the process model base and the emergency management knowledge base, and even the historical processing mode and the effect of the specific conditions can be used as decision support information to be output. The intelligent risk prediction method is a comprehensive application of an emergency management platform, a process model library and an emergency management knowledge base, and comprises risk identification and judgment and emergency scheme prediction, and the perfect process model library is favorable for helping enterprises to better cope with emergency situations.
Claims (8)
1. A risk intelligent prediction method based on process mining is characterized by comprising the following steps:
step one, reading real-time monitoring data of a sensor and preprocessing the data
Reading real-time monitoring data from each sensor in the chemical industry park, analyzing multi-source heterogeneous data of the sensors to obtain standardized and formatted data D (Type) which is easy to process i ,Monitor j ) Data indicating that the monitoring subject sensor j monitors the abnormality type i;
step two, judging whether the real-time monitoring data is abnormal or not and judging the abnormal grade
Inquiring an emergency management knowledge base EMKB, retrieving abnormal data judgment knowledge with an abnormal Type i and a sensor j, and retrieving D (Type) i ,Monitor j )∈K(Type i ,Monitor j ,Level k ) If yes, the current abnormity is judged to be Level k Abnormal degree, if it is Level 0 If no abnormity exists, returning to the first step for continuous monitoring;
step three, judging whether the abnormity is false alarm
Inquiring an emergency management knowledge base EMKB, retrieving alarm property judgment knowledge K (Type) with an abnormal Type of i and a sensor of j i ,Monitor j ,Monitor else )＝T(D(Type i ,Monitor j )&&D(Type i ,Monitor elses )|t∈[0,5s]) I.e. within 5 seconds, D (Type) i ,Monitor j ) And D (Type) i ,Monitor elses ) If one of the abnormal degree values is Level0, K (Type) i ,Monitor j ,Monitor else ) If the result is false, judging the result to be a false alarm and returning to the step I for continuous monitoring; otherwise K (Type) i ,Monitor j ,Monitor else ) Is true;
step four, generating a flow model candidate set according to the abnormal type
According to Type i 、Monitor j Accessing an emergency pipe flow model EMPM with an abnormal type i and a monitored subject sensor j, and adding the flow models into a candidate set S i ＝{EMPM(Type i ,Monitor j )|EMPM∈PML}；
Step five, screening the process models in the candidate set according to the abnormity in a grading way
Inquiring an emergency management knowledge base EMKB, retrieving abnormal data judgment knowledge with an abnormal Type i and a sensor j, and retrieving D (Type) i ,Monitor j )∈K(Type i ,Monitor j ,Level k ) In the candidate set S i Keep exception Level k The process model of (2) deletes the process models of other abnormal levels to obtain the updated candidate set S i ＝{EMPM(Type i ,Monitor j ,Level k )|EMPM∈PML}；
Step six, calculating the probability that the screened emergency management process model is suitable for the current abnormity
According to the candidate set S i The total number N of process instances in (1), and the number N of process instances of each model EMPMi Respectively calculating an emergency management model EMPM i For abnormal data D (Type) i ,Monitor j ) Probability P of the situation EMPMi ；
Seventhly, sequencing the prediction results and sending out prediction early warning
According to the final candidate set S i Probability P of each process model EMPMi The predictions are issued as a result, sorted from large to small.
2. The intelligent risk prediction method based on process mining according to claim 1, wherein the emergency management process model EMPM = (a, actor, VA, edge, GXOR, GAND, start, end), wherein: a is the set of service activities; edge = { (head, tail) } is a set of sequential streams, (head, tail) denotes a sequential stream from head to tail, head ∈ { a, gxxor, GAND, start }, tail ∈ { a, gxxor, GAND, end }; GXOR and GAND represent exclusive and parallel gateways, respectively; start and end represent the start and end events of the flow, respectively; actor is the set of participants in the service flow model; 6) VA is the set of merit arcs between a and Actor.
3. The intelligent risk prediction method based on process mining according to claim 2, characterized in that the emergency management process model EMPM is modeled according to the following method:
step 1, data preprocessing:
cleaning, filtering and classifying event data stored in an emergency management platform to obtain an event log set log;
step 2, analyzing and counting process examples:
taking a process example set of a classification subject S as a sample, analyzing and counting process examples T in the sample to obtain a set { (T) 1 ,n 1 ),(T 2 ,n 2 ),...,(T k ,n k ) Where T is i Is a specific example of a procedure, n i Is the number of instances of the procedure;
step 3, abstracting the process example into a process variant:
determining a set of process instances { T } 1 ,T 2 ,...,T k Merging the process instances T with the same cyclic relationship, and abstracting the process instance set into a process variant set { VI ] 1 ,VI 2 ,...,VI p }；
And 4, combining the process variants into a process model PM:
for variant set { VI 1 ,VI 2 ,...,VI p Combining the process variants in the method pairwise in sequence to determine the selection relation of the variants so as to obtain a complete process model PM;
step 5, extracting a value arc set VA:
based on each activity in the set of activities obtained in step 2, extracting from the corresponding event E the corresponding participant ar and the input value v in And an output value v out Thus obtaining a value arc set VA;
step 6, constructing an emergency management process model EMPM:
marking each value arc in the value arc set VA to a corresponding activity a in the process model PM i Firstly, establishing an emergency management process model EMPM;
step 7, forming a process model library:
adding the emergency management process model EMPM of the obtained classification subject S into a process model library; and repeating the step 2 to the step 6 for the process example set of each classification main body, and adding the obtained emergency management process model corresponding to each classification entity into the process model library so as to obtain a complete process model library.
4. The intelligent risk prediction method based on process mining according to claim 3, wherein the specific steps of the step 2 are as follows:
step 21, taking a process instance set of a classification subject S as a sample, mapping events of a process instance T into corresponding activities a with the same name, and mining an activity sequence relation and a parallel relation according to a time sequence relation so as to determine the activities a and the relation thereof contained in the process instance T;
step 22, merging the included activities a and the process instances T with the same relationship, and counting the number of the same instances to obtain a set { (T) 1 ,n 1 ),(T 2 ,n 2 ),...,(T k ,n k )}。
5. The intelligent risk prediction method based on process mining according to claim 3, wherein the specific steps of the step 4 are as follows:
step 41, set of variants { VI 1 ,VI 2 ,...,VI p Combining every two process variants in sequence, integrating and reserving the same part in the two variants, and merging the process variants with the selection relation, wherein the different parts exist in the form of the selection relation;
step 42, extracting all unrepeated events, activities and gateways in the original flow variant, and identifying the gateways according to the connected elements;
and 43, traversing each edge (head, tail) in all the original process variants, adding the edges into the new process model without repetition, wherein due to different parts existing among the variants, the situation that a non-gateway element has a plurality of outgoing edges may exist in the new process model, a selection gateway needs to be added to the element, the head in all the outgoing edges of the element is replaced by the selection gateway, all the outgoing edges of the element are deleted, and finally, an edge pointing to the selection gateway by the element is added, so that a complete process model PM is obtained.
6. The intelligent risk prediction method based on process mining according to claim 3, wherein the specific steps of the step 5 are as follows:
step 51, based on the activity a obtained in the step two, extracting corresponding participant ar from the corresponding event E, and extracting the input value v of the ar participating in the activity a from the set attrVal of the attributes and values of the event E in And an output value v out Thereby determining the value arc va = (a, ar, v) therebetween in ,v out )；
And step 52, executing the operation of the step five to each activity to obtain the value arc set VA.
7. The intelligent risk prediction method based on process mining according to claim 3, wherein the specific steps of the step 6 are as follows:
step 61, if the value arc (a) i ,ar j ,v in ,v out ) Middle activities a i And activity a in the Process model PM i If the value arc is the same, then label the value arc to the corresponding activity a in the process model PM i The above step (1);
step 62, repeating step 61, and establishing an emergency management process model EMPM = (a, actor, VA, edge, GXOR, GAND, start, end), wherein: a is the set of service activities; edge = { (head, tail) } is a set of sequential streams, head ∈ { a, gxxor, GAND, start }, tail ∈ { a, gxxor, GAND, end }; GXOR and GAND represent exclusive and parallel gateways, respectively; start and end represent the start event and end event of the flow, respectively; actor is the set of participants in the service flow model; VA is the set of merit arcs between a and Actor.
8. The intelligent risk prediction method based on process mining as claimed in claim 1, wherein P is EMPMi The calculation formula of (c) is: p EMPMi ＝n EMPMi /N×100％。
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310059941.1A CN115952919A (en) | 2023-01-16 | 2023-01-16 | Intelligent risk prediction method based on process mining |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310059941.1A CN115952919A (en) | 2023-01-16 | 2023-01-16 | Intelligent risk prediction method based on process mining |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115952919A true CN115952919A (en) | 2023-04-11 |
Family
ID=87287840
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310059941.1A Pending CN115952919A (en) | 2023-01-16 | 2023-01-16 | Intelligent risk prediction method based on process mining |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115952919A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117194083A (en) * | 2023-06-19 | 2023-12-08 | 山东理工大学 | Causal inference-based method and causal inference-based system for tracing and analyzing abnormal root cause of process time |
| CN117406972A (en) * | 2023-12-14 | 2024-01-16 | 安徽思高智能科技有限公司 | RPA high-value flow instance discovery method and system based on fitness analysis |
-
2023
- 2023-01-16 CN CN202310059941.1A patent/CN115952919A/en active Pending
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117194083A (en) * | 2023-06-19 | 2023-12-08 | 山东理工大学 | Causal inference-based method and causal inference-based system for tracing and analyzing abnormal root cause of process time |
| CN117194083B (en) * | 2023-06-19 | 2024-03-29 | 山东理工大学 | Causal inference-based method and causal inference-based system for tracing and analyzing abnormal root cause of process time |
| CN117406972A (en) * | 2023-12-14 | 2024-01-16 | 安徽思高智能科技有限公司 | RPA high-value flow instance discovery method and system based on fitness analysis |
| CN117406972B (en) * | 2023-12-14 | 2024-02-13 | 安徽思高智能科技有限公司 | RPA high-value flow instance discovery method and system based on fitness analysis |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN115952919A (en) | Intelligent risk prediction method based on process mining | |
| CN107886238B (en) | Business process management system and method based on mass data analysis | |
| CN112817280A (en) | Implementation method for intelligent monitoring alarm system of thermal power plant | |
| CN109255523A (en) | Analysis indexes computing platform based on KKS coding rule and big data framework | |
| CN117575332A (en) | Road construction safety monitoring method and system | |
| Seeger et al. | Literature review on using data mining in production planning and scheduling within the context of cyber physical systems | |
| CN114726654B (en) | Data analysis method and server for coping with cloud computing network attack | |
| Glukhikh et al. | Case Based Reasoning for managing urban infrastructure complex technological objects | |
| CN116882823A (en) | Multi-level engineering quality assessment management platform with self-correction function | |
| Atzmueller et al. | Anomaly detection and structural analysis in industrial production environments | |
| CN114281046A (en) | Safety evaluation model, and visual monitoring system and method for initiating explosive device area | |
| Lyu et al. | A data-driven approach for identifying possible manufacturing processes and production parameters that cause product defects: A thin-film filter company case study | |
| Hu et al. | E‐maintenance platform design for public infrastructure maintenance based on IFC ontology and Semantic Web services | |
| CN108981785A (en) | A kind of intelligent Detection of coal breaker equipment safety | |
| CN113869589A (en) | Power transmission line accident prediction method based on knowledge graph and inspection system | |
| Talamo et al. | Industry 4.0 for failure information management within proactive maintenance | |
| Kaupp et al. | „From Raw Data to Smart Documentation: Introducing a Semantic Fusion Process for Cyber-Physical Systems.“ | |
| CN114168409A (en) | Service system running state monitoring and early warning method and system | |
| Lin et al. | Design and implementation of a CPS‐based predictive maintenance and automated management platform | |
| CN113489602A (en) | Communication fault positioning method and system based on data mining | |
| Sun et al. | An artificial intelligence-based real-time monitoring framework for time series | |
| dos Santos et al. | Alarm management according to isa sp 18.2 standard based on process mining and multi-criteria decision making methods | |
| Ma et al. | Data management of salt cavern gas storage based on data model | |
| CN117853078B (en) | Unmanned aerial vehicle-based dynamic safety alarm system and method for construction site of building engineering | |
| CN116227902A (en) | Emergency management flow modeling method based on big data |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |