5 Predictive Maintenance Program Implementation Pitfalls
For industrial companies, downtime can be exceptionally expensive, while maintenance itself can be time-consuming and resource intensive. It’s no wonder then that the concept of predictive maintenance (PdM) has become one of central promises of the industrial Internet of Things in recent years. But the fact is, predictive maintenance programs often fail to deliver the expected results. This article will consider why that is, pointing to common causes of PdM failure. Let’s briefly review how R. Keith Mobley defined a predictive maintenance program in the 2002 book “An Introduction to Predictive Maintenance,” as:
[A] philosophy or attitude that, simply stated, uses the actual operating condition of plant equipment and systems to optimize total plant operation. A comprehensive predictive maintenance management program uses the most cost-effective tools (e.g., vibration monitoring, thermography, tribology) to obtain the actual operating condition of critical plant systems and based on this actual data schedules all maintenance activities on an as-needed basis.
The Internet of Things has significantly reduced the cost of carrying out many of the data acquisition tasks of a well-developed predictive maintenance program. These devices provide data, that when analyzed, trigger actions to take place. The online monitoring of the condition of industrial systems also significantly reduces the incident of maintenance-induced failure resulting from destructive processes such as breaking into soundly operating equipment to inspect internal components.
With the background set, let’s discuss PdM implementation pitfalls.
Trying to Make Everything a Predictive Task
When the value of PdM technologies is discovered, it could be easy to decide to use a technology for every monitoring point in the plant, thus inundating you with data and increasing your cost of on-line monitoring. PdM should be a part of an overall equipment strategy that begins with the most critical processes and systems.
The most effective way to determine where a predictive technology can be used is through the facilitation of a Reliability Centered Maintenance (RCM) analysis. This process will identify the potential failures and allow for the selection of the best strategy and technology you can implement to minimize the chance of one of those failures actually happening.
Collecting but Not Analyzing the Data
Implementing technology and not effectively using it is more common than it should be. This is partially because as you look to the technology, you neglect the human factors associated with the change to process. When you implement PdM technology, whether that be in the form of IoT devices dedicated to collecting data or you upgrade your preventive maintenance program to become predictive, you have to prepare the team for the changes that come with it.
The collected data needs to be analyzed to provide the information necessary to make correct and timely decisions. It is true some decisions can be made automatically based on rules setup in the computerized maintenance management system (CMMS) or other monitoring application, but there are only so many decision that can be automated this way.
Let’s say you’re only looking at the data from the incoming sensor to see the current condition of your asset. You’re not trying to use that data to look into the future. In this case, you will be basically doing condition-based maintenance and not predictive maintenance.
To get the most out of your data, you will often need to invest into predictive analytics software or a data scientist that will help you early on. As the time goes, if you have any extra manpower, you can look to train one of the maintenance managers or tradesmen to take this assignment over.
Expecting Immediate Results
With any data acquisition scenario, there is a tendency to expect that once the data is online and available, decisions can be immediately made.
Like when you implemented your CMMS software, you have to allow for history to accumulate before you can begin to see trends in equipment performance. Some anomalies will only be captured during seasonal changes in weather, others will be seen due to raw material input.
Don’t expect the data to immediately identify all possible problems. The more data you have, however, the easier it will be to develop accurate predictive models.
Not Integrating Collected Data
You could spend a lot of time implementing PdM tools and techniques only to find you are missing half the story. Equipment run time data is good in and of itself, however, it becomes powerful when coupled with process data. This combination allows you to determine if a part of the process causes certain mechanical or electrical problems or whether the condition of the equipment is causing process problems.
Not Properly Training Technicians to Use the Tools
Buying the latest toys like thermography imaging tools or ultrasonic lubricators is great, but if the technicians do not know how to properly use the tools or how to effectively analyze the data from them, there is a possibility they perform unnecessary and damaging work. While you might be inclined to reduce the overall implementation costs by putting training aside or doing some quick crash courses, that is something that could bite you in the long run.
As already mentioned, your failure predictions are only as good as your data. Technicians who don’t know how to use the new equipment or how to adapt to the new workflow have more chances to cause physical damage and provide you with inaccurate data.
All of these pitfalls can be avoided through the development of a holistic approach to the predictive program at your facility.
Bryan Christiansen is the founder and CEO of Limble CMMS, which produces CMMS software that aims to takes the stress and chaos out of maintenance by helping managers organize, automate and streamline their maintenance operations.