LU503601B1

LU503601B1 - Method for diagnosing alarm root cause in chemical process based on time-dependent abductive reasoning

Info

Publication number: LU503601B1
Application number: LU503601A
Authority: LU
Inventors: Dongfeng Zhao; Xiaomiao Song
Original assignee: Univ China Petroleum East China
Priority date: 2023-03-09
Filing date: 2023-03-09
Publication date: 2023-09-12

Abstract

The invention relates to a method for diagnosing alarm root cause in chemical process based on time-dependent abductive reasoning, which includes alarm data real-time collection, alarm data pre-processing, chemical process topology model establishment, time constraint network formation, alarm root cause diagnosis based on abductive reasoning, fault path display and display output of auxiliary decision-making interface, based on time-dependent abductive reasoning. The method for diagnosing alarm root cause in chemical process based on time-dependent abductive reasoning, comprises alarm data real-time collection, alarm data pre-processing, chemical process topology model establishment, time constraint network formation, alarm root cause diagnosis based on abductive reasoning, fault path display and display output of auxiliary decision-making interface; based on time-dependent abductive reasoning, it can combine the correlation and time in the chemical process, accurately obtain the root cause of alarm and fault path, also, it can use auxiliary decision-making to provide help information for operators.

Description

DESCRIPTION LU503601

METHOD FOR DIAGNOSING ALARM ROOT CAUSE IN CHEMICAL PROCESS

BASED ON TIME-DEPENDENT ABDUCTIVE REASONING

TECHNICAL FIELD

The invention relates to a method for diagnosing alarm root cause in chemical process based on time-dependent abductive reasoning.

BACKGROUND

The alarm system is the core element to ensure the process safety of the process industry. The performance of the alarm system is directly related to the process safety, product quality, production cost and even casualties, and is an important factor affecting the national economy and the people's livelihood. According to Abnormal situation management (ASM) survey, every year, industrial enterprises lose hundreds of millions or even billions of dollars due to unexpected production stoppage, operational errors and other reasons. For example, in the Three Mile Island Accident in 1979, due to the external interference of the system, at least one alarm occurred in 2 to 3 seconds on average. In a large number of alarm messages, the operator failed to identify the valve abnormality and did not take effective measures, which led to the accident. Therefore, a scientific and reasonable alarm system is of great significance to ensure the safety, reliability and economy of industrial processes. However, due to poor alarm management, the industrial alarm systems currently in use have low performance and the most common problem is "alarm flooding", through investigation, it can be seen that the correlation characteristics of chemical process are complex and the abnormal propagation range is wide. In this case, it is difficult to locate the initial alarm quickly, so that the operator can't take scientific and effective measures, which is the main reason for the alarm flooding. How to correlate the relevant characteristics in the chemical process industry, find the alarm source more accurately and determine the fault path is an urgent problem to be solved. LU503601

In view of the above problems, experts, scholars and technicians have also explored the root cause diagnosis to a certain extent, mainly using mechanism modeling methods, knowledge driven methods, data driven methods, information fusion methods, etc. to establish a topology model, and then using backtracking, reasoning, hypothesis testing methods to locate the root cause of the alarm and identify the propagation path. However, in the actual chemical process industry, the above methods are vulnerable to various interferences and their own complex characteristics, which make the location of the alarm source ambiguous. And under the condition that sufficient historical alarm data cannot be guaranteed, the online diagnosis of the alarm mode cannot be realized, the in-depth analysis on the dynamics and hysteresis of the chemical process is lacked, the multi-variable time sequence characteristics are not fully mined, the abductive reasoning method forms the evolution process of events by reasoning and interpreting the observed phenomena and results. It has a good application prospect in the field of fault diagnosis and alarm. At the same time, the HAZOP analysis method based on the structural analysis of the danger and operability of process parameter deviation can comprehensively and systematically study all the elements in the system, which is very convenient for finding the related characteristics in the chemical process, and chemical enterprises basically require full coverage application. Moreover, in the existing automatic control systems of chemical enterprises, single display alarm points are in the majority, there is no correlation indication between alarms, the auxiliary decision-making function is weak, and scientific and effective prompts cannot be given to operators, in the enterprise investigation, the operator hopes that the alarm system can effectively identify the root cause of the alarm when multiple alarms occur, eliminate the initial alarm and provide auxiliary decision-making guidance.

SUMMARY LU503601 (I) Technical problems to be solved

In view of the shortcomings of the prior art, the invention provides a method which can combine the correlation and time in the chemical process, accurately obtain the alarm source and fault path, and at the same time can provide help information for operators by using auxiliary decision-making, and solves the problems raised by the background technology. (I) Technical scheme

In order to realize the above purpose of combining the correlation and time in the chemical process, accurately obtaining the alarm source and fault path, and at the same time providing help information for operators by using auxiliary decision-making, the invention provides the following technical scheme: a method for diagnosing alarm root cause in chemical process based on time-dependent abductive reasoning, which includes alarm data real-time collection, alarm data pre-processing, chemical process topology model establishment, time constraint network formation, alarm root cause diagnosis based on abductive reasoning, fault path display and display output of auxiliary decision-making interface; based on time-dependent abductive reasoning method, the invention can combine the correlation and time in chemical process, and accurately obtain the alarm root cause and fault path, at the same time, use auxiliary decision-making to provide help information for operators.

Preferably, the alarm data real-time collection is realized by Python programming according to the data collection interface of the enterprise. The OPC based data collection system collects the data of BPCS, SIS and other systems in real time, including but not limited to the alarm bit number, alarm value, alarm threshold, alarm time, etc., and stores the alarm data in the distributed storage system HBASE.

Preferably, in the alarm data pre-processing, Python is used to classify the collected original alarm data and perform z-core standardization processing. For missing values and abnormal values, appropriate treatment methods are adopted.

Preferably, in the chemical process topology model establishment, after the HAZOP report is processed through programming by using the input PID diagram and the improved HAZOP analysis method, the process topology relationship that can be used [N 503601 computer processing is established through the causal association between the tag numbers of alarm variables, and the associated topology diagram can be displayed as the reasoning rule of abductive reasoning, and the action control after the input logic judgment in BPCS system is also regarded as the reasoning rule.

Preferably, in the time constraint network formation, based on the above reasoning rules, the time constraint network is produced by using the time constraint between rules.

The time constraint network is to take the vertex as the time of event occurrence, connect the vertices with directed edges, describe the time constraint information between events, and then solve its time consistency problem and minimum network problem.

Preferably, in the alarm root cause diagnosis based on abductive reasoning, judging the alarm root cause by combining reasoning rules and time constrained network, and its algorithm is recursive.

Preferably, the fault path display is based on the chemical process topology model, and the fault path is displayed in the interface through the root cause diagnosis of the alarm derived from the abductive reasoning, which is conducive to the operator's more intuitive and clear observation of the fault path.

Preferably, the display output of the auxiliary decision-making interface is to expand the database of its protection measures and relevant historical data when the improved

HAZOP report is input. When the alarm root cause is diagnosed, firstly, the solution is put forward for the root cause, and the operator can click according to the fault path display, and then click to display the auxiliary operation scheme at this point. (II!) Beneficial effects

Compared with the prior art, the invention has the following beneficial effects: 1. According to the method for diagnosing alarm root cause in chemical process based on time-dependent abductive reasoning, the time-dependent abductive reasoning can effectively combine time information, HAZOP analysis tables, PID diagrams and other existing data to scientifically and effectively determine the root cause of the alarm, and can search and display the fault path and auxiliary decisions according to the root cause. Compared with the existing alarm root cause diagnosis system, the method for diagnosing alarm root cause in chemical process based on time-dependent abductive 503601 reasoning has the advantages of strong scientificity, high accuracy, rapid diagnosis process, and can provide auxiliary decision-making and fault path, and has good application prospects. 2. The method for diagnosing alarm root cause in chemical process based on time-dependent abductive reasoning, comprises alarm data real-time collection, alarm data pre-processing, chemical process topology model establishment, time constraint network formation, alarm root cause diagnosis based on abductive reasoning, fault path display and display output of auxiliary decision-making interface; based on time-dependent abductive reasoning, it can combine the correlation and time in the chemical process, accurately obtain the root cause of alarm and fault path, and at the same time, it can use auxiliary decision-making to provide help information for operators.

BRIEF DESCRIPTION OF THE FIGURES

FIG. 1 is the schematic diagram of an alarm root cause diagnosis system based on chemical process knowledge automation of the invention;

FIG.2 is the flow diagram of HAZOP analysis table conversion of the invention;

FIG.3 is the topological diagram of the chemical process of the invention;

FIG.4 is a schematic diagram of the formation of the time constraint network of the invention;

FIG.5 is the schematic diagram of the fault path and auxiliary decision-making display interface of the invention;

FIG.6 is the schematic diagram of HAZOP analysis after processing of the invention;

FIG.7 shows an example of the alarm result of the invention.

DESCRIPTION OF THE INVENTION LU503601

In the following, the technical scheme in the embodiment of the invention will be clearly and completely described with reference to the attached drawings. Obviously, the described embodiment is only a part of the embodiment of the invention, but not the whole embodiment. Based on the embodiment of the invention, all other embodiments obtained by ordinary technicians in the field without creative labor belong to the scope of protection of the invention.

As shown in FIG. 1- FIG. 7, The invention provides a technical solution, a method for diagnosing alarm root cause in chemical process based on time-dependent abductive reasoning, including alarm data real-time collection, alarm data pre-processing, chemical process topology model establishment, time constraint network formation, alarm root cause diagnosis based on abductive reasoning, fault path display and display output of auxiliary decision-making interface.

Construction of real-time data acquisition platform: based on the enterprise data acquisition interface, using Python programming to realize the server framework that meets the OPC communication protocol. The BPCS, SIS and other systems of the real-time enterprise operation device collect alarm related data (including but not limited to alarm bit number, alarm value, alarm threshold, alarm time, etc.). The alarm data acquisition frequency is as follows: once every 30 seconds before the alarm, once every seconds after the alarm, the collected alarm data is stored by the distributed storage system HBase and extracted to the MYSQL database.

Data pre-processing: using the z-score method for data standardization, the method is based on the mean and standard deviation of the original data to standardize the data, that is, standardizing the original value x of A tox, by using z-score method. The conversion function is as follows:

Transforming the sequence A 1 UV HE , heté °05601 release [LS wy n=" n=-17"

For the processing of missing values, the method of directly using the features containing missing values, deleting the features containing missing values and completing the missing values can be selected according to the data situation. When there is an abnormal value that significantly deviates from the average value of the observation value, the record containing the abnormal value can be deleted according to the data situation, and the abnormal value can be treated as the missing value, and the average value can be used to correct and not treat.

Example 2

Establishment of chemical process topology model: first, optimizing the HAZOP analysis table, and the specific steps are as follows. According to the selected research object, obtaining the latest version of the HAZOP analysis table of the object, analyzing and processing the deviation one by one. The main process of processing includes three steps. The first step is to confirm whether the "recommended measures" corresponding to the current deviation have been implemented. The main significance of this process is that the recommended measures may include some alarm settings. If they have been implemented, the existence of the alarm shall be considered during analysis; the second step is to determine the alarm tag numbers corresponding to the deviation, cause and consequence in combination with the existing protective measures, the proposed measures implemented and the process flow control. In this step, it should be noted that due to the non-uniform standards and other reasons, some deviation causes and consequences in the original HAZOP analysis table are not represented in the form of deviation, but actually there are corresponding alarm tag numbers, it shall be identified as far as possible; the third step is to add one column after the process deviation column, the cause column and the consequence column respectively, and write out the alarm tag number (standard format) corresponding to the deviation. If there is no corresponding alarm tag number for the current deviation cause or the consequence of the deviation, it is unnecessary to fill in. Through the implementation of the above process, the HAZOR 503601 analysis table shown in FIG. 6 can be obtained. After the HAZOP analysis table is processed, using Python programming to identify the input PID diagrams together to form chemical process topological network diagram. Among them, alarm points are identified as nodes, their associated nodes are also displayed, and their causal relationships are connected by arrows, which are displayed as a tree structure.

In FIG. 4, after the high alarm of the bottom liquid level of the vacuum tower C104 in an atmospheric and vacuum distillation unit, according to the input optimized HAZOP analysis table and PID diagram, there are four reason variables for this variable: the overcurrent fault of P-116A/B pump, the decrease of vacuum degree of C104 vacuum tower, the sudden drop of outlet temperature of vacuum furnace and the failure of

FIC-11305 washing oil flow control valve in the third vacuum line; there are two reasons for the decrease of vacuum degree of variable pressure tower C104: the outlet temperature of the pressure reducing furnace is higher, and the top temperature of the pressure reducing tower C104 is higher; there are two reasons for the high outlet temperature of the vacuum furnace: coking of the furnace tube of the vacuum furnace and small steam injection volume of the furnace tube of the vacuum furnace; there are four reasons for the high top temperature of the pressure reducing tower: high outlet temperature of the pressure reducing furnace, fault closing of the circulating control valve of the pressure reducing tower, poor cooling effect of the circulating cooler of the pressure reducing tower, and low return flow after pressure reduction interruption. Thus, a clear and intuitive causal tree structure diagram of variables is obtained.

Example 3 LU503601

By analyzing the operation data of the actual device, the following alarm results can be obtained. The data source is the operating data of the atmospheric and vacuum distillation unit of a refining and chemical enterprise from August 9 to August 16, 2020.

The sampling frequency is 30s/time. The first alarm bit number in FIG. 7 is "LIC-11304".

This variable generates a high alarm at this moment and records its alarm time.

According to the method introduced in this patent, the tree structure diagram of this variable is first obtained as the reasoning rule based on the time related traceability reasoning, and then all alarm times related to the tree diagram are searched. According to the time constraint network in FIG. 4, the variables that meet their time consistency are P-116A/B, EJ-120, FI-11202, TIC-11301, FIC-11308. The time constraint network uses vertices to represent the time variables to be calculated. The connection arc between vertices represents the constraint relationship between time variables. In practice, each variable can represent the start time point and end time point of each alarm. In the alarm, the time relationship between variables can be judged according to whether the time consistency problem is solved. According to the results, that is to say, there is a causal relationship between these variables, and the root cause is the interruption of reflux pump damage.

Example 4 LU503601

Decision interface output: the auxiliary decision information intelligently retrieved by the real-time search system is displayed on the left side of the initial node, then the root cause of the alarm is taken as the starting point, and its cause and effect transmission path is represented in the form of a symbolic directed graph. The alarm root is the initial node, and the cause and effect transmission path is presented in the form of a symbolic directed graph. The alarm tag with the highest alarm priority is marked with emphasis.

When the high liquid level alarm at the bottom of the decompression tower, the overcurrent alarm of the P-116A/B pump, and the low vacuum alarm of the decompression tower occur, the associated alarm is automatically identified based on the time based abductive reasoning, and the root cause can be identified as the failure of the decompression interruption reflux pump. At this time, the auxiliary decision-making information is provided to remind the operator to go to the site as soon as possible to check the interruption reflux pump and determine the root cause of the alarm.

When working, first selecting the research object, selecting the latest version of

HAZOP report, selecting any analysis node, then selecting any deviation, determining the implementation of the proposed measures, determining the deviation, and filling in the alarm signals corresponding to the causes and consequences in combination with the existing protective measures, the implemented recommended measures, and the process flow control (add the EXCEL column), after combing all the deviations according to the score, the next step is to complete the node combing. Otherwise, going back to select any deviation to continue running. If all the nodes are combed, it will end.

Otherwise, going back to select any analysis node to continue running.

Although embodiments of the invention have been shown and described, it can be understood for those skilled in the art that a variety of changes, modifications, substitutions and modifications can be made to these embodiments without departing from the principles and spirit of the invention, and the scope of the invention is defined by the appended claims and their equivalents.

Claims

CLAIMS LU503601

1.A method for diagnosing alarm root cause in chemical process based on time-dependent abductive reasoning, characterized in that: the method includes alarm data real-time collection, alarm data pre-processing, chemical process topology model establishment, time constraint network formation, alarm root cause diagnosis based on abductive reasoning, fault path display and display output of auxiliary decision-making interface; based on time-dependent abductive reasoning method, the invention can combine the correlation and time in chemical process, and accurately obtain the alarm root cause and fault path, at the same time, use auxiliary decision-making to provide help information for operators.

2. The method for diagnosing alarm root cause in chemical process based on time-dependent abductive reasoning according to claim 1, characterized in that: the alarm data real-time collection is realized by Python programming according to the data collection interface of the enterprise; the OPC based data collection system collects the data of BPCS, SIS and other systems in real time, including but not limited to the alarm bit number, alarm value, alarm threshold, alarm time, etc., and stores the alarm data in the distributed storage system HBASE.

3. The method for diagnosing alarm root cause in chemical process based on time-dependent abductive reasoning according to claim 1, characterized in that: in the alarm data pre-processing, Python is used to classify the collected original alarm data and perform z-core standardization processing; for missing values and abnormal values, appropriate treatment methods are adopted.

4. The method for diagnosing alarm root cause in chemical process based on time-dependent abductive reasoning according to claim 1, characterized in that: in the chemical process topology model establishment, after the HAZOP report is processed through programming by using the input PID diagram and the improved HAZOP analysis method, the process topology relationship that can be used in computer processing is established through the causal association between the tag numbers of alarm variables 503601 and the associated topology diagram can be displayed as the reasoning rule of abductive reasoning, and the action control after the input logic judgment in BPCS system is also regarded as the reasoning rule.

5. The method for diagnosing alarm root cause in chemical process based on time-dependent abductive reasoning according to claim 1, characterized in that: in the time constraint network formation, based on the above reasoning rules, the time constraint network is produced by using the time constraint between rules; the time constraint network is to take the vertex as the time of event occurrence, connect the vertices with directed edges, describe the time constraint information between events, and then solve its time consistency problem and minimum network problem.

6. The method for diagnosing alarm root cause in chemical process based on time-dependent abductive reasoning according to claim 1, characterized in that: in the alarm root cause diagnosis based on abductive reasoning, judging the alarm root cause by combining reasoning rules and time constrained network, and its algorithm is recursive.

7. The method for diagnosing alarm root cause in chemical process based on time-dependent abductive reasoning according to claim 1, characterized in that: the fault path display is based on the chemical process topology model, and the fault path is displayed in the interface through the root cause diagnosis of the alarm derived from the abductive reasoning, which is beneficial for operators to observe the fault path more intuitively and clearly.

8. The method for diagnosing alarm root cause in chemical process based 503601 time-dependent abductive reasoning according to claim 1, characterized in that: the display output of the auxiliary decision-making interface is to expand the database of its protection measures and relevant historical data when the improved HAZOP report is input; when the alarm root cause is diagnosed, firstly, the solution is put forward for the root cause, and the operator can click according to the fault path display, and then click to display the auxiliary operation scheme at this point.