A few blogs back we looked at growing interest in extracting value from IoT data through industrial analytics. This interest has not sprouted up overnight. Since the beginning of computer-assisted control systems, plant engineers and managers have been using their increasingly powerful and sophisticated tools to gather data, and then use the data to improve their processes.
For much of that time, the idea was to collect data in a database, and then at the end of the month or quarter, run various analytical tools on the data to see where the problems and bottlenecks were, and what could be changed. This approach had some value, but it is essentially a reactive model. Today, there is a general trend underway to go beyond simple reaction like this, and move towards the ability to predict problems, and if possible prescribe a solution. In a recent blog, Blurred Lines Between Predictive and Prescriptive Analytics Mike Guilfoyle at ARC Advisory Group explains the value of each of these approaches to analytics, as well as their differences.
He breaks down this kind of pro-active analysis into three parts: performance, predictive and prescriptive, distinguished as follows:
- Performance describes what is happening or has happened, and is the starting point of all analytics, reactive or pro-active. The focus here is on current or past performance.
- Predictive looks forward to what is most probable to happen, given the current conditions, using Big Data, machine learning, and other IT tools.
- Prescriptive uses all of the above inputs, and adds to that a knowledge base and decision-making algorithms to prescribe what action can or should be taken. In some instances, the system might actually even carry out the action, which is referred to as “closed-loop control.”
Guilfoyle goes on to identify some of important differences between predictive and presciptive analytics. In fact, he will be leading a session on analytics best practices at the ARC Industry Forum this week. You may not be able to attend, but his article is a good introduction.
In any case, the trend towards predictive and prescriptive analytics and any kind of closed-loop control based on such approaches highlights the need for secure, real-time access to plant data. It is yet another example of the closing gap between OT and IT, and is an unmistakable benefit of the Industrial IoT.