1943. C.H. Waddington, a scientist of the Royal Air Force, noticed that something was wrong with the aircrafts’ maintenance strategy. In fact, by collecting data on repairs and analyzing it, he discovered that breakdowns were more likely to happen after scheduled maintenance sessions. This phenomenon, which later became known as the Waddington effect, shows that Preventive Maintenance can often be more damaging than no maintenance, as it interferes with a satisfactory status quo.
Many attribute the invention of Predictive Maintenance, which now constitutes a market worth $16 billion, to Waddington.
Although Predictive Maintenance is one of the main changes boosted by Industry 4.0 and it has been around for a long time, it has not yet plateaued.
An Overview of Maintenance Methodologies
Maintenance methodologies can vary according to many parameters and depend on the monitoring of different metrics, such as sound, temperature and vibration. In particular, vibration analysis is a very advantageous way to assess the performance of industrial machines, as changes in their behaviour almost immediately reflect in the vibration signals emitted by the machines. However, such a deep level of monitoring is not yet common practice, and many operators still rely on older methodologies. The following outlines the most common maintenance strategies (image 1).
- Reactive Maintenance implies that intervention only happens whenever a machine breaks down, without executing prior analysis or optimization.
- Preventive maintenance relies on repeated maintenance interventions. During each control session, operators or service suppliers physically inspect the plant’s equipment. They may or may not detect a malfunction and intervene accordingly. However, as the Waddington paradox proves, this strategy is not always effective. Intervening with the machine very often means interfering with its normal operation, which can negatively affect its performance. Until now, this methodology, together with Reactive Maintenance, was the standard.
- Condition-based monitoring is based on the analysis of vibrations emitted by the machine. However, intervention is only performed if the signal exceeds a critical, predefined threshold value. This can indicate three different things:
a. The machine is not really broken. Since the threshold is fixed, there are no further insights on the meaning of the vibrations and the signal might have exceeded the threshold for other reasons.
b. There is a real malfunction but operators do not know what it is or how much time is left until failure. Once again, the vibration analysis is not thorough enough to provide that information.
c. There is a real malfunction, and operators know exactly what it is and how to fix it. However, they had to carry out their analysis and repairs in the high-stress, unplanned context of an emergency, stopping the machine mid-production.
4. Predictive Maintenance goes a step further, by collecting a wide range of signals and using Machine Learning algorithms to analyze them. This makes it possible to gain additional insights about the machine and allows operators to schedule maintenance sessions around production, such as during planned shutdowns.
5. Machine Diagnostics actually go beyond Predictive Maintenance and represent the most advanced level of maintenance of those mentioned here. This strategy allows a thorough knowledge of the equipment and its parameters. It does not only calculate the time to failure, but also gives more insights about the root causes of the error. In this way operators gain a deeper understanding of the error occurred and can easily initiate the right measures. This in-depth analysis is what the AiSight solution can provide.
Why upgrade to Predictive Maintenance
Improving your maintenance strategy has two main benefits: Increased performance and decreased costs.
As reported in a survey, administered by the VDMA, of businesses that implement Predictive Maintenance, increased performance is a greater advantage than the cost decrease, according to those. Increased performance results mainly from:
- Longer machine availability
- Longer useful life of the machine
- Safer and more sustainable operations
- Increased quality of the process and end product
On the other hand, decreased costs result from saving on the following:
- Repairs and spare parts
- Communication with service providers
- Downsizing of service staff
Moreover, assessing whether you are allocating too many or too few resources on your maintenance strategy is very easy. The most powerful instrument to do so is the ROI, which helps you understand whether there is a balance between the expenses arising from machine repairs and the money invested in the implementation of Predictive Maintenance. More on ROI will be written in future articles.
What Stands Between Words and Action?
Though 81% of German firms recognize Predictive Maintenance as an important trend, only 40% have relevant offerings or technologies in place to address it. If the benefits of implementing a Predictive Maintenance strategy are so widely acknowledged, then what is preventing more firms from adopting one?
One reason is that such a strategy is difficult to implement. To do so, a company would need the following: sensors, to detect the vibrations of your machine; a system that gathers all the data in an aggregated way; a central hub, where all the data can be stored, safely; algorithms to analyze them; specialized, skilled people, to make the algorithms work and draw conclusions. All of these elements also mean additional costs.
Riding the wave of Industry 4.0
Luckily, amongst the great innovations brought forth by Industry 4.0, are new solutions aimed at making the upgrade to Predictive Maintenance (and beyond) an attainable objective for all types of firms. Among these, is AiSight Sensorkit.
The hardware component is a sensor node equipped with state of the art accelerometers with exceptionally high frequency bands, WiFi communication, as well as high performance microcontrollers enabling pre-processing on the edge. The flexibility of the solution allows operators to easily install the sensor kit on a machine, using magnets or screws, and run it without having to undergo additional training. Data is collected by the sensor kit in real time and transmitted to the recently revamped dashboard. The information is then analyzed by machine learning algorithms and operators are alerted of any anomalies.
The broad spectrum of vibration signals that can be caught by the sensor node allows it to detect even the smallest deviations from the normal behavior of the machine. Our algorithms can simultaneously perform a Root Cause Analysis to determine the cause of the deviation. By doing this, the sensor node will help ensure that the machine is operating efficiently. (Learn more about how this can positively impact your plant’s cost.)
If you want to know more about AiSight’s research drop us an email!Request more information