The beauty of just in time predictive maintenance
Nothing lasts forever. And every machine eventually breaks down if not being properly maintained. Predictive maintenance is a trend that everyone has probably already heard about couple of times. What is the status in power drive systems / variable speed drive systems? Where does the industry stand? And what is the beauty of just in time predictive maintenance? This post presents few thoughts on that topic.
Types of maintenance
Let’s recap shortly what types of maintenance are commonly used:
1. Reactive maintenance
With this approach, the equipment is used to its limit and repairs are performed only after the machine fails or component breaks down. For inexpensive systems such as e.g. light bulb, the reactive approach might make sense. However, for complex systems with expensive parts and especially consequential damages it is not really suitable. Typical example in VFD world is a blast furnace blower in a metallurgical plant. Unexpected downtime of such a drive system does not just mean loss of production. If the molten metal cools down inside the furnance the whole infrastructure is likely to experience severe damages.
Reactive maintenance may not be an option if the consequences have significant safety relevance or a severe business impact.
2. Preventive maintenance
Preventive maintenance is probably the most common maintenance strategy across the industries. The aim is to prevent failure before it occurs by performing regular maintenance checks. The recommended maintenance schedule is typically provided by the corresponding equipment manufacturer. The intervals are defined conservatively as there are several factors that are not in detail known in advance (specific ambient conditions, operation mode, utilization, duty cycles etc). The drawback of conservative maintenance schedule is the increased cost and reduced availability due to scheduled downtimes.
3. Predictive maintenance
Predictive maintenance shall eliminate the disadvantages of previous two approaches. A suitable algorithm let us estimate time-to-failure. It allows to optimize maintenance interventions to be as effective as possible. Predictive maintenace does not only predict a failure. it helps to discover issues inside the system. In VFD world predictive approach helps to detect non-optimal operation and to increase energy efficiency.
Beauty of just in time concept
In modern logistics and material ordering the “just in time” principle is actively lived since several years. Basically it means that the subsuppliers shall deliver their material or parts right before they will be needed. it optimizes the cash flow of the company (order and pay for the material as late as possible) and minimizes the storage space and associated cost. Most car makers work exactly this way. They receive the parts just before they will be needed in the assembly line. of course, it does not only come with benefits. Besides cost savings this approach increases the vulnerability of the supply chain and may lead to disruptions in production. The world pandemic has given us a big lesson in this regard. Despite of that most companies will probably stick to the “just in time” principle or perform only minimum changes (as many companies simply do not have the storage capacity anyway).
Just in time predictive maintenance
But let’s go back to predictive maintenance. Here the “just in time” approach brings similar benefits and possibly less drawbacks. Simply said the use wants to perform the maintenance works exactly as needed:
- Not too late (unplanned downtimes and liquidated damages)
- Not too early/too frequent (reduced availability, unnecessary maintenance cost)
The trick is to find the right balance -> just in time maintenance.
Predictive maintenance is quite easy to realize in variable frequency drives (VFD). One advantage is that we practically have all the required hardware. In most cases no extra sensor is needed. The VFD utilizes existing sensors used for control and protection. Several other quantities can be observed from other signals. Again, VFDs have long tradition and experience with this principle. Great example is sensorless speed control used already for several decades. The machine speed is determined by a motor model using measured electrical signals. This method eliminates the need for a speed or position encorder. In fact, most VFDs are operated encoderless.
How does predictive maintenance work?
Each manufacturer has its own proprietary algorithm. That is where the know-how and previous experience flows into. The algorithms are also refined and retuned based on experience from the specific plant. You might have exact same application and exact same VFD in two different installations. However, their maintenance intervals may be quite different. For instance, one VFD is installed in a clean environment while the other one is exposed to dust and other particles. By means of predictive maintenance it will be recognized when a service intervention is required.
Predictive maintenance strategy typically examines trends of key parameters or indicators. An illustrative example is air cooling of a VFD. Air-cooled VFD normally has dedicated filters at the air inlet. Over the time, the filters get polluted by dust and other particles. For the fresh cooling air it will be more difficult to enter the VFD enclosure. This is a natural thing that every air-cooled VFD experiences. However, the rate of change of pollution is installation specific. Without predictive maintenance the exchange interval of the filter mats shall be short enough to be on the safe side. A more advances approach can monitor the pressure drop or observe slowly increasing temperature rise for given loading over the time.
Data and information
Any predictive algorithm requires data. As already mentioned, VFD is inherently equipped with several sensors to measure specific signals (e.g. current and voltage) and other quantities can be estimated based on those measurements. Data is a necessary precondition. But data itself is not so much useful – what is really needed is information. The algorithm processes the data and creates the information out of it. Those can be various indicators describing the health state of the equipment.
Feedback for the manufacturer
There is also a benefit for the equipment manufacturer. He can use the information (if authorized by the user, of course) to recognize weak spots in the design or to identify (sub)components with quality issues. In this way, the product can be further improved and enhanced reliability can be achieved.
Summary
While preventive maintenance is still the most common approach in variable speed drive systems, more and more users recognize the advantages of predictive maintenance, such as reduced downtime and maintenance cost, optimized spare parts kit, increased equipment lifetime or reduction of severe faults.
References
[1] Reliability and availability, https://mb-drive-services.com/category/reliability-and-availability/
[2] ABB drive services, https://new.abb.com/drives/services
[3] Introduction to predictive maintenance with MATLAB, The MathWorks Inc., 2019, https://ch.mathworks.com/
