دانلود مقاله ISI انگلیسی شماره 29002
ترجمه فارسی عنوان مقاله

استفاده از مدل سازی شبکه های بیزی برای برنامه ریزی تعمیر و نگهداری در صنعت ساخت

عنوان انگلیسی
The use of Bayesian network modelling for maintenance planning in a manufacturing industry
کد مقاله سال انتشار تعداد صفحات مقاله انگلیسی
29002 2010 11 صفحه PDF
منبع

Publisher : Elsevier - Science Direct (الزویر - ساینس دایرکت)

Journal : Reliability Engineering & System Safety, Volume 95, Issue 3, March 2010, Pages 267–277

ترجمه کلمات کلیدی
تعمیر و نگهداری - تعمیر و نگهداری بازرسی - محیط زیست - مدل سازی شبکه های بیزی - تجزیه و تحلیل تاخیر زمان -
کلمات کلیدی انگلیسی
Maintenance, Inspection maintenance, Environment, Bayesian network modelling, Delay-time analysis,
پیش نمایش مقاله
پیش نمایش مقاله      استفاده از مدل سازی شبکه های بیزی برای برنامه ریزی تعمیر و نگهداری در صنعت ساخت

چکیده انگلیسی

This paper has been written in order to apply Bayesian network modelling to a maintenance and inspection department. The primary aim of this paper is to establish and model the various parameters responsible for the failure rate of a system, using Bayesian network modelling, in order to apply it to a delay-time analysis study. The use of Bayesian network modelling allows certain influencing events to be considered which can affect parameters relating to the failure rate of a system. Bayesian network modelling also allows these influencing events to change and update depending on the influencing data available at any given time, thus changing the failure rate or probability of failure. A methodology has been developed and applied to a case study in order to demonstrate the process involved.

مقدمه انگلیسی

When using delay-time analysis to develop a maintenance or inspection model, the need for both relevant and accurate data is vital to the success of the task [1]. The information required in order to carry out such a modelling exercise is gathered from historical data and/or from expert judgement. This information is used to calculate the variables needed to apply delay-time analysis. The variables include: • average downtime due to inspection, d; • average downtime for a breakdown repair, db; • arrival rate of defects per unit time, kf; • failure rate λ (1/MTBF); • inspection period, T. Downtime due to inspection, d is the amount of time, on average, an inspection will take to complete and return the equipment to production. The average downtime due to a breakdown and subsequent repair of the equipment db is the time it takes on average to return the equipment to production. The units of both downtime inspection and breakdown repair downtime must be identical but can be measured in hours, days or months depending on the equipment under investigation. The arrival rate of a defect, kf is the average time a defect arises over a period of time, calculated by the number of defects divided by the total operating time of the equipment under investigation. Failure rate λ is the reciprocal of mean time between failure (MTBF) where MTBF is the mean operating time between failures of a component or piece of equipment. MTBF, however, should not be confused with the delay-time h of a component or piece of equipment. The delay-time h is the time from an initial telltale sign of failure to actual failure, both being dependant on the inspection interval, T. Given the above information, an expected downtime per unit time function D(T) can be obtained as follows [2] and [3]: equation(1) View the MathML sourceD(T)={d+kfT[(1/T)∫0T(T−h)λe−λhdh]dbT+d} Turn MathJax on Looking at the variable failure rate λ (1/MTBF), the information required to populate this is based on statistical averages. For example, if a machine or piece of equipment has experienced 10 breakdowns over a period of 1 year this would result in a failure rate of 0.027 failures/day (MTBF 37 days). To further expand on this example, suppose 70% of the breakdowns occurred during the first 3 months of operation, with only 1 breakdown experienced during the last 2 months of operation. Calculating these figures into failure rates highlights the inadequacy of relying on averages when gathering data of this type [4]. Specifically, the failure rate for the first 3 months is that of 0.077 failures/day (MTBF 13 days) but the failure rate for the last 2 months is 0.016 failures/day (MTBF 63 days). Although the average failure rate of 0.027 failures/day (MTBF 37 days) is correct for average breakdowns it may not be adequate to portray the actual situation. Continuing with this example, there may be a number of influencing factors that have been responsible for the varying failure rates over the 12-month period. For example, poor reliability of equipment may be encountered due to incorrect installation. Conversely, improvements in the design of the equipment may improve the reliability of the equipment. Other typical influencing factors for this example might include [3]: • poor initial implementation of equipment; • improvements in inspection procedure; • improvements in maintenance personnel training; • renewal of key components; • changes to inspection intervals. A modelling technique capable of appreciating the differing influencing factors, which could affect an event or variable is that of Bayesian network modelling. Bayesian network modelling is a simple mathematical formula for calculating conditional and marginal probabilities of a random event. Conditional probability is the probability of an event given the occurrence of an influencing event, whereas marginal probability is the unconditional probability of an event. Bayesian network modelling can also deal with subjective probability, which may represent the degree of belief from an expert, and apply it in a precise and relevant manner.

نتیجه گیری انگلیسی

In Jones et al. [3], a case study was carried out to establish an optimal inspection interval using delay-time analysis. The aim of this case study was to minimise maintenance and inspection costs by reducing downtime whilst optimising inspection intervals. Delay-time analysis relies on several parameters in order for it to be effective. One of the parameters, failure rate λ, relies on the manipulation of historical data using statistical averages. This model has served to give a better understanding and confidence to the parameter failure rate λ. It has not only given an opportunity to increase the accuracy in a modular way but also has given an insight into the likely causes should a failure take place. This paper has demonstrated the use of applying Bayesian network modelling to provide an improved and accurate method of establishing the parameter failure rate λ. Although the inspection interval has reduced, greater confidence can now be given to the results of this study given the inclusion of several consequential factors relating to failure.