The natural gas transmission requires compressors equipments continuously operating. Unexpected equipment out of service changes gas pipeline operating condition. These facts have an economic impact with possible invoicing reductions, the payment of penalties and diminution business profits. The machine failure probability distribution is the base for analysis model. It allows to study the impact on planning, maintenance management and to evaluate different operation states. The present work gives the way to obtain machine failure probability distribution and develops a model to predict operating compressor out of service. As result, we obtained an easy-to-use and powerful tool for operating states simulations.


To determine the failure probability of a machine it is essential to know the out of service probability distribution function. In case of a known distribution as Weibull, Bathtub curve, Infant mortality, Constant Probability or any other that represents the system, we ought to adjust its parameters to calculate the probability out of service. If the error is higher than the acceptable limits, then the alternative is to work with the relative frequencies of the row data.


Figure 1 describes high pressure gas pipeline transmission. It begins with gathering injection and treatment plants, continues with several compressors stations located along the gas pipeline in the end of a section and the beginning of the other. Each section of pipeline has delivery points for the customers. The scope of this work is the study and analysis of the non programmed compressor out of service. It includes all the compressors, "Reciprocating" or "Centrifugal".

Sorting Data

We worked with 29 centrifugal compressors, 31 reciprocating compressors in 18 compressors stations across the 7,000 km gas pipeline. The oldest records date from 1996 and the newest begin in 2001. The running machine time is reported monthly. The data is presented in the follow scheme:

  • Programmed Maintenance. Planned maintenance tasks.

  • Non Programmed Maintenance. Emergency or corrective maintenance tasks.

  • Reserve. The compressor is waiting for start up.

  • Operating Hours. It includes all kinds of operation. The out service frequency distribution is based on Non Programmed Maintenance. To compute a failure it is only need one hour registered as a non programmed maintenance task in the month. The focus is the interruption of the gas transmission. This is the reason to compute one single and simple failure as the same as breakdown machine. Following this criterion and assumptions we could develop a very useful tool for programming, planning and studying maintenance practice, equipment performance, improvements and deal with commercial agreements.

Model Basics

Once the data were collected and sorted, the steps followed to get F functions were:

  1. Data revision.

  2. Sorting the data by plant, equipment and date.

  3. Days between out of service calculation.

  4. List the number of failures by each machine for each time interval.

  5. Sum of total failures by equipment

  6. Relative frequency ft

  7. Calculation of cumulative frequency F. calculation

This content is only available via PDF.
You can access this article if you purchase or spend a download.