Computer Learning Systems - Predicted Maintenance vs. Manufacturers' Recommendations
Computer learning, often called machine learning, refers to the reliance on technology to solve problems and make decisions as a human would – without relying on humans to make it happen by coding it in. In manufacturing, computer learning takes many forms. One of the most useful applications of the concept occurs in the area of equipment maintenance. Most machine operators know that routine maintenance generally increases the lifespan of a piece of equipment and maintains efficiency. However, many factors affect machines, how they operate, and potential issues they may have. Because of this, many operators look to other resources (like computer learning), rather than recommendations from the manufacturer, to plan and schedule preventative maintenance.
Manufacturers’ Recommendations for Maintenance
For nearly every piece of manufacturing equipment, there are manufacturer recommendations for the types of maintenance owners should invest in – and how often. While these recommendations certainly provide valuable guidance on how to keep a machine running efficiently, they are by no means absolute rules. This is mainly due to a couple of factors. First, a large percentage of routine maintenance is unnecessary and likely to cause more issues than it resolves. Also, it’s proven that the majority of equipment failures occur randomly, despite the frequency of maintenance measures.
What’s the consequence of solely relying on manufacturer recommendations for preventative maintenance? There are several. First, performing maintenance on a machine generally means it can’t be used in production during that time. If not properly planned, downtime can result in manufacturers having operators on the floor without machines to operate. This inefficiency can be expensive in terms of labor costs if it happens often.
On the other hand, maintenance recommendations don’t take into account actual machine data. So, even though a machine may be clearly demonstrating malfunction, maintenance plans based solely on the manufacturer’s recommendations may not recommend maintenance until it is too late. When operators ignore a malfunction, it can often lead to machine failure, causing unplanned downtime.
There are a few ramifications of this reactive approach to maintenance. For example, unplanned downtime immediately means lost revenue. Additionally, if delays in the manufacturing process are significant, it can affect customer orders and the shipping timeline. Ultimately, this affects when customers receive their products and can easily result in poor customer satisfaction. Reactive maintenance also tends to be much more expensive. Fixes to major issues generally cost more and take more time than preventative efforts. Because of the unplanned downtime, productivity also suffers, and the overall cost of the machine increases.
So, how does a manufacturer plan preventive maintenance – not solely relying on recommendations from machine manufacturers AND not waiting until machines fail to take preventative steps. Computer learning can help with that.
Predicted Maintenance from a Computer Learning System
What if there existed a system that would alert machine operators when productivity started to decrease, signaling maintenance to be scheduled? Luckily for modern manufacturers, it does. Computer learning helps to identify machine degradation in its early stages, giving ample warning time to operators to investigate potential issues. This information doesn’t come from explicit human programming, but rather from analyzing the history of a machine’s production. Basing decisions on actual data versus blanket operating rules means greater accuracy of those decisions.
The Computer Learning Module of the CIMAG MES
In general, computer learning can significantly help machine operators better care for their equipment and better anticipate maintenance needs. The computer learning module of the CIMAG manufacturing execution system goes even further. This robust module not only gives maintenance warnings very early in a machine’s degradation but also is self-diagnosing. This means that it can often identify a precise root cause of an issue quicker than a human.
Because of the increased lead time the module gives for maintenance, operators can take initiatives to address issues before repair cost increases due to ignoring the issue (or not knowing it existed). The highest repair costs generally occur when an issue is so severe, it leads to equipment failure – and a 0% production rate. Now, not only are repairs likely more expensive than they would have previously been, but the machine is also down for production until repairs are completed. The module uses historical data rather than programmed rules, which is a much more accurate method of analysis.
Relying solely on manufacturers’ recommendations often results in one of two occurrences. Many times, more maintenance is performed than is actually required for a machine, costing the company time and money. In other cases, signals from the equipment that it is degrading are ignored and issues escalate, becoming more severe and more expensive. From a management perspective, machine learning is a great alternative approach to maintenance, because it ensures that operators perform maintenance when it is needed, not too early or too late.
The Computer Learning module of CIMAG predicts equipment failure in the early stages of degradation, helping to better plan repairs and limit needed downtime. Because the module integrates with the rest of the CIMAG software, these plans are shared in real-time with everyone involved – from those on the shop floor operating the machines to those in management making strategic decisions for the company. Having that full visibility integrated with the production schedule, managers can better plan maintenance rather than relying on machine manufacturers telling them when to do it. How machine learning operates can be confusing, but this white paper can help clarify the process. For more information about the Computer Learning module or our MES software, contact us today.