Home >> Industry Knowledge >> Nitrogen Generator (Nitrogen System) real-time fault diagnosis
Preface
With the rapid development of science and technology, modern cryogenic systems Yue Lai
Vietnam becoming a more complex, automation and intelligent, the traditional fault diagnosis
Technology, appears to be insufficient. In recent years, artificial intelligence techniques, particularly
It is the fault diagnosis expert system in a wide range of fields of research and .
A low-temperature real-time fault diagnosis expert system using delphi
Powerful data processing capability and comes with a database of the characteristics of their making
For the development of tools, knowledge base and inference engine in the structure of mutual separation,
Knowledge emphasis on the various operating parameters as the logical relationship between the value of
To determine the normal operation of low-temperature system is based on inference engine by
Knowledge Base sheet automatic search function to achieve knowledge base, knowledge base changes,
Update does not affect the function of the realization of the inference engine, so knowledge base
Updates, management is very convenient.
A real-time fault diagnosis expert system structure
The system, through the process of computer system running a variety of data
Import the information into the database as a basis for fault diagnosis, once the system out of
Is failure, one or some of the data will deviate from the normal operation of the system
. In the expert system established, pre-determined special different fault
Point and the system runs the logical relationship between the data and reasoning based on
Between the various operating parameters is the value of logic.
Cryogenic system consists of the whole process of a computer monitor, process meter
Computer system is a low-temperature system, critical systems are process control, the number of
According to the management of the core system. Real-time fault diagnosis expert system in the PC
Machine to run. According to low-temperature system architecture and real-time fault diagnosis of
Requirements, PC machines and the process of exchanging data between computers 5s intervals 1
Times. If there is failure, real-time fault diagnosis expert system, or tell it is
Barrier occurred in the cause and location, or generate automated troubleshooting instructions (some of
Failure), through the PC, and communication interfaces to pass the process of the computer,
Automatically exclude low-temperature system equipment failure. Cryogenic System and real-time and therefore
Impaired diagnostic system structure shown in Figure 1.
Figure 1 low-temperature systems, and real-time fault diagnosis system structure diagram
Real-time fault diagnosis expert system for state monitoring of the cryogenic system
Monitoring and fault diagnosis, mainly by the knowledge base, diagnostic reasoning module, the letter
Interest Query Module, and display print module to complete. Real-time fault diagnosis
The basic structure of expert system shown in Figure 2. Knowledge base and diagnostic push
Reasonable fault diagnosis expert system module is a core component of its function is to
Based on continuous real-time database is written in response to the state of cryogenic systems
On-site real-time data on the low-temperature systems, condition monitoring and fault diagnosis
Off. If faults occur, promptly inform the user through the man-machine interface, it is
Barrier occurs because the location, severity and troubleshoot programs
And other information.
Figure 2 Real-time fault diagnosis expert system basic structure of the diagram
2 Knowledge Base and Inference Engine
211 of knowledge extraction and knowledge base to establish
Knowledge is mainly used to store industry experts to provide specialized knowledge
Knowledge, for the inference engine to provide the necessary knowledge to solve the problem. Building knowledge
Library, must address the issue of how to store of knowledge, this is the so-called
Knowledge representation, the exact tell us how it can be stored in a computer-shaped
Style to express the knowledge.
Knowledge base is divided into two kinds of numeric and text-based to value-based
Knowledge-based, text-based knowledge base as an effective supplement. Value
Type repository tables in the form of Table 1, text-based form of knowledge base, see Table
Table 2.
Table 1 numeric Knowledge Table
T1 lower limit P1 lower T1 minimum P2 limit P1 maximum P2?? Fault troubleshooting
t11 t11 'p11 - 1 - 1 p21'?? causes an exclusion method 1
t12 - 1 p12 p12 '- 1 - 1 because two exclusion method 2
Diagnostic reasoning process through the program will form part of the database
Knowledge combination of the knowledge base of a simple form of a database search
Faso, the design of reasoning, and further realization of the complexity of artificial neural networks
Reasoning capabilities.
212 inference
Real-time fault diagnosis expert system is compatible with forward reasoning capabilities
Based on the realization of the reverse reasoning and hybrid reasoning of the two kinds of inference engine
System failure for different types of flexibility.
21211 Forward Reasoning
Forward reasoning is the conclusion from the facts the direction of reasoning, but also
Known as the fact-driven reasoning, forward reasoning, knowledge base of the form shown in Table 1.
Forward reasoning Example: reduce the swelling volume (flow decrease) ∩ temperature
Drop decreases (ΔT decrease) ∩ gap pressure rise] work wheel blades
Block] approach: first, an appropriate increase in the amount of blowing smoke addition, if not
OK, only stop blowing heat addition.
Although the forward reasoning approach to solve the basic types of reasoning failures
Requirements, but for complex reasoning functions to achieve the efficiency of low,
And accuracy is not high.
21212 backward reasoning
Backward reasoning to a fault has occurred or will occur to trigger fault parameters
Number changes as a starting point, and further evidence of the failure to find support for
Match, until the failure occurred in the Knowledge Base to find the causes and
Approach so far. Backward reasoning process shown in Figure 3.
Table 2 backward reasoning database tables
O
Gas
Production
Volume
Low
Low-pressure distillation
Connecting pipes, dryer is gas leakage
Scrubber whether the leak valve
In addition to whether the blow off valve or a leak seriously Yan
Operating pressure high temperature gas, cryogenic liquid leakage
Normal pressure distillation column
Nitrogen, oxygen purity of the normal amount of distillate out too much, it should be under the premise of ensuring the purity of the product as much as possible to reduce the amount of distillate out
Purity nitrogen to reduce
Low purity liquid nitrogen, liquid nitrogen through the large throttle opening, should be suitably adjusted
Purity nitrogen gas is much lower than that of liquid nitrogen, liquid nitrogen through the throttle open a small amount of adjustment should be large
Up on the tower of steam increased, the reflux ratio decreases, the amount of oxygen removed is too small, it should be the premise of ensuring the purity of oxygen, under
Open the oxygen valve as much as possible out
Nitrogen, oxygen purity were decreased
Tower body tilt, the vertical degree of calibration tower
Plug tray should be cleaned heating plate
Produced Flooding
Backward reasoning flow chart of Figure 3
Backward reasoning procedure through a database search to find to
Implementation, examples in Table 2.
21213 Hybrid Reasoning
Hybrid reasoning is forward reasoning and backward reasoning and flexible combination
Together, will produce a corresponding failure of reasoning by looking to gather together
Information on the composition of the new database, and then forward to the database
Reasoning, forward reasoning methods mentioned above. Produce hybrid reasoning collection
The database instance, such as Table 1.
21214 artificial neural network function
Simulation of artificial neural networks of people committed to the realization of the right brain functions,
Is a real person on the physiological structure and function of neural networks in order to
And a number of basic characteristics of a theoretical abstraction, simplification and simulation of the structure
Into an information processing system.
In the actual system constructed, the input layer enter the cryogenic system operating
All real-time parameter values, real-time parameters of every 5s updated 1 times,
Artificial neural network diagnostic system at intervals of 5s on the course of computer input
Real-time data for a second diagnosis. When you enter a group of fault parameters
, The output numbers for different types of failure probability values, under a non -
Failure probability of the same type of numbers to determine the size of the possibility of the occurrence of the failure
The size of the order in accordance with the size of the possibility of guiding operator
Members of troubleshooting, Table 3.
Table 3 may be introduced by the possibility of failure followed by failure cause
2, fault 3. The reasoning process is also in line with the real experts, not reasoning
Characteristics of uncertainty and ambiguity. Further guarantee trouble-shooting, increase
Added a diagnostic system of logic and rigor.
3 real-time fault diagnosis expert system operation
Real-time fault diagnosis expert system is running, the operator needs a
Solution system is running a variety of information, including a variety of historical and real-time-like
State information, troubleshooting information and troubleshooting information, and other processes Letter
Interest, therefore, to build a good information retrieval and display mechanism for non -
Often important.
311 the system queries the current state of information
The current state of information inquiry system, allowing users to understand the system when the
5s period of time before the working section of some important parameters, variable values, etc.
Process information. System provides real-time display of various signals, the signal
Is divided into analog and numerical quantity. Numerical quantity through the screen at intervals of 5s
Refresh one time, each value within 5min by check curve
The amount of graphic simulation, clearly see the value of change in trend is super -
And lead a normal range.
312 alarm information query
Real-time fault diagnosis expert system, the alarm information in the screen
Central alarm pop-up window shows the current fault information (including the resolution of
Programs). In addition, the system also provides alarm information query interface:
The current fault information table and the historical failure information log on the table. Current Fault
Information table chronological record of the current has not yet ruled out the occurrence of faults
Time, the name. When a fault is resolved, the fault from the current
Failure in the table disappears. The historical fault information visit the table is also a time-shun
Sequence records the last 1 week when the system is running all the fault occurs
, The name and the failure to exclude the time.
4 Conclusion
Real-time fault diagnosis expert system can accurately to the entire low-temperature system
Commission to condition monitoring and fault diagnosis, diagnostic reasoning module to enable fault
Diagnostic reasoning, clarity, different types of failure to make corresponding
Treatment, guide the user to troubleshoot, greatly improving the efficiency of troubleshooting and quasi -
Indeed the rate. Information query feature allows users to easily transport query system
Line of historical and real-time operation. Show print function correctly
Shows the history of production systems and real-time operating status information, fault report
Police and diagnostic information, support interfaces display information in print, the user can
Easily through the printer to print out the necessary historical and real-time information,
For staff analysis and study.
References:
[1] WANG Shi-tong. AI Tutorial [M]. Beijing: Electronic Industry Publishing
Society, 2001.
[2] Chang will, Shao crane. Real-time fault diagnosis expert system operating mechanism
[J]. Automation Instrumentation, 2000, 21 (4): 8210.
[3] Li T, Jin Ning, Zhu Yan. Artificial neural network used in air-conditioning system
Study on fault diagnosis [J]. Refrigeration & Air Conditioning, 2005, 5 (1): 502
53.
[4] Luo Xiaoping. Delphi Essentials [M]. Beijing: Electronic Industry Press,
2004.
[5] Zhang Hua. Electrical and mechanical equipment condition monitoring and fault diagnosis technology [M]. West
AN: Northwest Industry University Press, 1995.
[6] Yuan served. Artificial neural network and its application [M]. Beijing: Tsinghua
University Press, 1999.
[7] Wen-Jun Cao. Knowledgebase system theory and its applications [M]. Shanghai: Fudan University
University Press, 1995.
[8] WU C H. Building knowledge structures for online
instructionalPlearning systems via knowledge elements
interrelations [J]. Expert Systems with Application, 2004
(26): 3,112,319.
[9] ANGELI CHR. An online expert system for fault diagnosis in
hydraulic systems [J]. Expert Systems, 1999, 16 (2):
4782481.
[10] MARZI H. Real 2time fault detection and isolation in
industrial machines using learning vector quantization [J].
Proquest Science Journals, 2004, 218 (B8): 949.