The Path from Data to Actionable Information as a Driver for the Industrial Ecosystem
equipment. For example, predictive vibration analytics can find a pattern for undesirable vibration when a machine is making a 20mm slot in steel with a 40mm endmill having three cutting items and rotating at 2500 RPM. The analytics will be continually refined as more data is collected and the model is improved. This update cycle allows the local systems to continue functioning and ensure safety even if they are not able to communicate with other parts of the ecosystem.
Prediction is often referred to as knowledge since one is building models that are capturing the causality relating to the objective truth. It has the potential of adding tremendous value to the manufacturing processes since it allows for the avoidance of loss and reduces unexpected process disruptions. With surrogate models and simulations, first principle models can be calibrated to the reality of the actual manufacturing execution and become more prescriptive.
PRESCRIPTIVE
Following from predictive analytics, focusing on avoidance of problems before they occur, prescriptive analytics allows the system to avoid problems by predicting future outcomes and working around situations that are highly likely to cause problems or determining best practices. Examples of prescriptive analytics are technologies that prescribe optimal process parameters when using certain tools to cut a feature in a certain material, in this case, the information will inform the Computer Aided Manufacturing( CAM) engineers to better specify how tooling is used. One can also prescribe optimized material flows at the enterprise scale to increase on-time delivery and machine utilization. Maintenance strategies can be significantly improved by prescribing when repairs should occur based on the machine capabilities and the required activities in the job queue.
Prescriptive analytics avoid losses before they occur. By combining the IIoT information streams with the intent-based models of the product geometry and inspection plans, the causality of decisions and impact on outcomes can be better understood. The eventual goal is to get to a level of prescription where the outcomes can be forecast to the extent that ondemand scheduling can adapt to rapid changes in product requirements and new orders, down to individual parts.
As with predictive models, the prescriptive models will be created using large amounts of historical data. The models will be updated as they are refined and better predictions become available. The standards that are currently applicable to the prescriptive analytics are SysML for system engineering, STEP, specifically AP-242ed2 19 for solid model geometry and GD & T as well as QIF for quality reporting and statistics. Models like AP-238 or STEP-NC can also be used to provide the expected execution stage models to compare the engineering intent.
19 http:// www. ap242. org
IIC Journal of Innovation- 31-