Data analysis precedes reliability analysis

Modern diesel engines are reliable. Nevertheless internal components do fail occasionally. More often failures occur in ancillary components such as injectors, coolant pump, fuel pump, lubricant pump, and other components.

When performing a reliability analysis we don’t prejudge the outcome of the analysis. Our objective is to discover relationships between failure probability and observed data that may have predictive content.

In other words we don’t extract (from the maintenance management database) only the failure data that we believe is related to oil and glycol fluids analysis or other data that we happen to be monitoring. We extract all the failure and repair data including the preventive repair and replacement events. We need the date, operational hours,  and failure mode (i.e. usually the part or component that failed) for each failure event as well as the date and operational hours of each preventive renewal of a failure mode.

Next extract any other data that can be relevant, that is, that you think may be predictive. This could include, fuel and oil consumption, motor-current data, sensor data, control data, even weather records if they are considered influential by maintenance and operational personnel.

Then, using this large data sample, we begin our analysis. First we look for problems in the data itself. For example the basic problem described in this video

is usually the main data issue. Then we try to determine correlation between each failure mode and monitored data.

New installations

Let’s assume you are involved in a plant or operational expansion and you will be adding new equipment to the maintenance program. You would like to make sure that any new data systems or procedures that will be implemented for these new equipment will apply the logic as discussed in this article. You would arrange that:

1. The maintenance engineers and technicians identify the failure modes that they need to manage. (Often this is done with an initial RCM analysis.)
2. They list those factors from possible monitored sources that can influence (predict) the probability of the failure modes identified.
3. They set up procedures to ensure that, on future work orders, the failure mode will be correctly referenced. Additionally, the work order documentation procedure must indicate the event that ended the life of a failure mode, one of either:
   1. The failure mode did not fail but was renewed preventively (called a “suspension”), and
   2. The failure mode failed (or was close to failure, called a “potential failure”) and was renewed.

The comments in this article are central to a Living RCM program operating in the maintenance organization.

## Post Tree Navigation

LRCM

RCM

Reliability Analysis

CBM