Application of the Bayesian approach in an expert’s system for diagnostic fault in computing engineering

The problem of increasing the efficiency of maintenance and repair of computer equipment is solved by applying the method of probability theory in expert systems. An expert system is a set of programs that accumulates the knowledge of specialists in a specific subject area and replicates this empirical experience for consultations of less qualified specialists. A probabilistic analysis method of equipment operation is used in the expert system under development for diagnosing and preventing computer malfunctions, is used. This article provides detailed description of the expert system, diagnosing the causes of computer malfunctions, provides a structural diagram of the system, rules and facts of work are generated. Based on the well-known Bayesian formula, this article describes a method for using fault information to clarify initial probabilities.


Introduction
Research and analysis of existing strategies shows that the planned maintenance and repair strategy, (the strategy according to the schedule), leads to a significant cost overrun of labor and material resources [1]. Moreover, this strategy does not reduce the probability of after-repair failures. The solution to this problem is the use of more flexible strategies for maintenance and repair (MRO), which will combine to ensure the reliability of computer technology through long-term diagnostics and forecastingprojective and predictive [2]. The application of such maintenance and repair strategies requires the support of information systems, including mobile ones. The purpose of such systems is to provide access for service engineers to the manufacturer's documentation on equipment, credentials for its actual state, databases of typical problems and methods for eliminating them. The application of such maintenance and repair strategies provides troubleshooting until the moment when the system has obvious problems.
A preventative maintenance strategy is effectively and fully implemented if the staff has the knowledge, skills and time necessary to carry out the relevant activities.
A monitoring and troubleshooting method is proposed that uses information technology of engineering knowledge, which is practically implemented as an expert system to overcome the indicated difficulties, The relevance of this theme is the number of failures in the work of computer technology is increasing, and modern methods for their detection do not provide timely diagnostics, it is required to find a new approach for detecting malfunctions. The proposed approach allows you to create a system that clarifies the probability of various failures according to the Bayes formula, with accumulated operating experience. qis the power supply OK ?; 4 q -Is the device case hot?
In the designed ES in the dialog box, the user is provided with questions concerning the specific problem situation that has arisen. Depending on the choice the user can agree with this fact or not, but the EC will provide the following question: = 1 , . . . , .
(2) Products in this system are Many facts and products are collected in a certain system, represented as a diagnostic OR column with terminal vertices q -the router firmware is out of date; 7 qwear of the power supply; 8 qoverheating of the device; 9 q -possibly increased load of the router. Figure 1 shows a fragment of such a graph. In the "OR" column, the orientation of the arcs shows the direction of the output. The natural division of the graph vertices into tiers reflects the depth of the output. However, in practice, there are much more frequent cases where the decision maker does not have complete information about the situation, and an assessment of the reliability of decisions is required. One way to evaluate this reliability is to use the Bayesian probabilistic approach [6] [7] [8].
The knowledge based on a diagnosing ES with this approach contains two types of records:  A key element of this methodology is the calculation of the price of certificates k F by the formula:  The firmware did not help; did the router firmware help? The power supply is not working; Is the power supply operational? The case of the device is not hot; Is the device case hot? A large load on the router; Is there a big load on the router? The router does not work properly; Is the router working properly? Based on this knowledge base and formula (7), it is possible to form an array of initial prices of evidence (table 1), as well as the probability of hypotheses (table 2) at each step of the decisionmaking procedure. Step 1 1 k C Step 2  Table 2. Hypothesis probabilities.

Discussion
According to the initial a priori probabilities, the prices of certificates are calculated and a dialogue with the system begins. The certificates with the highest price are highlighted ( Table 1). As a result of the obtained values in table 1, we see that the probability estimates at each step changes, which indicates a change in the degree of faith in a particular hypothesis Step 1. The certificate has the highest price 6 F : The router is malfunctioning. The question is: is the router working properly? When answered YES, the probabilities of hypotheses and the prices of evidence are recalculated.
Step 2. The certificate has the highest price 2 F : The firmware is outdated. Asked: Is the firmware out of date? When answered YES, the probabilities of hypotheses and the prices of evidence are recalculated.
Step 3. The highest price has a certificate 5 F : A large load on the router. The question is asked: Is the load on the router heavy? When answered YES, the probabilities of hypotheses and the prices of evidence are recalculated.
Step 4. The certificate has the highest price 3 F : The power supply is defective. The question is asked: Is the power supply OK? When I answer, I DO NOT KNOW the probabilities do not change.
Step From this we can conclude that the observed evidence fully confirms the hypothesis, which suggests that the malfunction of the router is in an outdated version of the software, which may cause the device to malfunction.

Conclusion
Bayesian approach can be applied to all areas of activity related to malfunctions of computing equipment.
Calculations carried out on the available experimental data show that as a result of recalculation, the posterior probability of a particular malfunction can either increase or decrease. After several steps, this algorithm leads to the fact that some faults, the posterior probabilities of which have become very small, are discarded (no longer taken into account), and others are suggested to be corrected.
The proposed relatively simple method will significantly accurately assess the reliability of decisions made by an engineer on computer malfunctions, in contrast to other existing methods. Since often the engineer does not have complete information about the situation, and an assessment of the reliability of decisions is required.
Continuous automated recalculation of a priori probabilities based on the incoming information will significantly improve the quality of forecasting, as well as increase the efficiency of maintenance and repair.