In most maintenance organizations today, predictive maintenance is not a foreign concept. The level at which an organization utilizes predictive maintenance can vary depending upon the type of industry the business operates within, the leadership knowledge regarding predictive methods, and the level of severity to the organization when equipment failure occurs. A typical response among organizational leaders in describing the desired state of equipment maintenance is that they want the predictive model of support in their organization. Some companies to not fully embrace the predictive model because they feel that it is an advanced maintenance program that would be expensive, or that their particular group cannot execute. Any organization can find benefit and can easily add at least one predictive tool that would improve the reliability of the machinery.
In a basic sense, predictive maintenance is straightforward. Predictive maintenance relies upon understanding what the standard “good” condition is for measurement and having the ability to detect when a change occurs. A person also needs to understand what that measurement is telling us. One of the most common activities that people associate with predictive maintenance is vibration analysis. In this example, a technician measures the vibration of machine bearings, compares that value to the known reasonable amount and specifications, and detects when the vibration readings go up. When this occurs, the technician investigates the issue and plans the corrective action before having a bearing failure and unplanned downtime on the machine. This same method of taking the measurement, noticing a change, and planning corrective action can be applied to many types of equipment and measure.
For example, at one manufacturing company, there was a piece of equipment that measured the change in motor current over a while as it performed its operation. If the change in current/time were too great, the machine would fault. Sometimes in the manufacturing process, this fault could occur due to material inconsistencies or other reasons not related to the material condition of the equipment, so it would be hard for the maintenance technician to determine if the fault was a machine issue or a process issue. From studying previous equipment failures and correlating how many of these faults were received leading up to the crash, we were able to determine that we were at risk for unplanned downtime if we collect more than ten mistakes per day. We then were able to make a software program that would retrieve the alarm log from the machines, total up the alarms from each machine each shift, and automatically send us an email if any machine had more than ten faults in 24 hours.
Many different opportunities, such as the previous example, exist to take advantage of the data already contained in the organization’s equipment. Motor amperage that can be measured and recorded, the temperature of a gearbox that is either automatically or manually recorded, particles of iron in a gearbox lubricant when measured with an inexpensive meter, are all examples of measurements that can predict the performance of a machine and predict the component life remaining. These simple items can be recorded and used as predictors without expensive analytical equipment or extensive training. A company can then bring in outside experts for the advanced technical aspects (such as advanced vibration analysis) on an as-needed basis, reducing the overall cost for an external resource. The main point to remember is that anything that can be measured can be used to alert the organization when a change occurs, and then is used to predict.