Making Factories Smarter Through Machine Learning
Another point of view can be seen in Figure 6, where the acceleration level related to the shaft is plotted against servomotor power. In this case, the clustering technique distinguishes between idle, acceleration and deceleration and maximum power in terms of acceleration levels. The acceleration level is independent from the power consumption. However, power levels can be distinguished more clearly than in the data shown in Figure 5. Therefore, from the point of view of predictive maintenance, it is expected that a servomotor should maintain this fingerprint acceleration level at all power consumption levels. Since the acceleration is related to the shaft angular speed, a malfunction of the servomotor could be detected when anomalies are outside the clusters: for example, anomalous vibration levels at a given acceleration state.
5. SUMMARY
This system represents the convergence of the OT and the IT. Through this convergence and advances in sensor fusion, edge and cloud computing, machine learning is seeing adoption in many ways and is now playing a central role in smart Factory predictive analytics. This article provides an example where both unsupervised and supervised machine learning are used in predictive maintenance:
�
�
Unsupervised Learning: as a first step to detect or find new information within data and for machine monitoring. Supervised Learning: to answer specific questions regarding the spindle and overall machine behavior providing a 20 % increase in improved asset availability, estimated at an additional three hours per day of use.
Maximizing system operation and lifetime provides valuable advantages to further advance the intelligence of manufacturing facilities. Specific areas directly related to Predictive Maintenance include: o Machine knowledge discovery o Transparency to the machine and process behavior o Increased understanding of system weaknesses o Ability to tune the operation to increase asset utilization o Optimizing the system for maximum productivity o Preventing Unplanned Downtime; detect anomalous operation; predict in advance of failure o Extending asset and system life
6. FUTURE OUTLOOK: SMART FACTORY OPERATION: FROM PREVENTATIVE TO PREDICTIVE TO PRESCRIPTIVE
The path to intelligence is through“ listening” to and“ learning” what the data is saying. Gaining actionable insight is only valuable when acted upon in a meaningful way that provides value. Today a human operator is involved in many areas for preventative and predictive maintenance.
- 38- January 2017