SAP for Chemicals Blogs
Explore blog posts by chemical industry leaders and SAP experts, showcasing transformative solutions powered by SAP. Contribute a blog post of your own.
Showing results for 
Search instead for 
Did you mean: 
0 Kudos
The topic of machine learning keeps coming up. At the time, I started writing this googling machine learning resulted in about 49,700,000 results. Originally I ignored the topic thinking machine equaled equipment and the topic only concerned equipment operating better and better due to continual small adjustments and investigation of the results. Something that the Original Equipment Manufactures installed on their equipment and that it had no general use. But the topic is much bigger and more generalized than that.

Machine learning has nothing to do with equipment but is a method of data analysis. And the machine is a computer. Together they allow for the automation of analytical model building. By using algorithms that are designed to learn iteratively from data, machine learning allows computers to find hidden insights in the data without being explicitly programmed where to look.

The general concept of machine learning is very well explained in a  series of videos on YouTube by Andrew Ng of Stanford University

And Machine learning is a sub set of Artificial Intelligence. Which begs the question, what is artificial Intelligence? Merriam Webster defines Artificial Intelligence as follows:

  1. a branch of computer science dealing with the simulation of intelligent behavior in computers

  2. the capability of a machine to imitate intelligent human behavior

Stanford University defines AI as “the science and engineering of making intelligent machines, espec....”

A detailed definition of AI is tricky (what makes up AI), because as soon as machines can perform a task that previously only humans could do, then the task in no longer considered to be a mark of intelligence. Being able to read used to be considered a criterion for AI, now Optical Character Recognition is just an everyday technology.

And then there are chat bots. A chat bot is a computer program that is designed to simulate conversation with human users either via voice (recorded) or by text. Chat Bots are typically used in for various practical purposes including customer service or information acquisition.

One thing that all these technologies have in common is the need for large amounts of data. And an iterative process to analyze, generate the result, and correct.

All these parts can be imbedded in multiple business processes. Chat Bots can be used to solicit more information; machine learning can be used to suggest solutions or recommendations.

Let’s take a maintenance scenario and look at the following steps

A problem is discovered with a piece of equipment

  1. A maintenance worker is sent with a work order to correct the problem

  2. The maintenance worker starts to repair the equipment by going through a series of diagnostic steps as stated in the maintenance procedures

  3. Once the problems are identified then the worker can repair the faults.

  4. Maintenance worker repairs the faults

  5. Corrective actions are recorded

And the same process with AI, machine learning, and chat bots:

A problem is discovered with a piece of equipment. A signal is sent to the AI machine where it uses this information to identify the probable causes and solutions for the problem

  1. Armed with this information the maintenance worker is sent to rectify the problem

  2. While repairing the problem the chat bot gathers addition information by guiding the worker through a series of additional questions. This additional information in fed into the AI process to utilize further machine learning algorithms to identify the root cause of the fault.

  3. The maintenance worker repairs the fault and the root cause.

With the AI / Machine Learning / Chat Bot process there is a higher probability the fault will truly be corrected since there is the root cause analysis that the worker is guided through, along with suggestions to remedy the cause.

This has the potential to radically change the way we do maintenance and the results of our efforts. However, these benefits are dependent on what we are doing today. More specifically in gathering data about causes, our corrective actions, and of course putting all this information into a form that can be accessed for machine learning. My concern is that we are not rigorous in gathering and recording data about our maintenance efforts.

Several enterprises I have encountered just process the work order, with materials and times. But do not record changes to the work order in detail, nor do they record the causes & corrective actions associated with the initial problem report. This lack of data means that when we start looking at embedding AI / Machine Learning & Chat bots into our processes that we will not have sufficient data to analyze and gain improvements. We, I fear, will spend more time correcting the automated suggestions than gaining benefit. This I believe will cause us to drop the entire effort.

If you believe that this technology has benefit in the future, then we need to start now. We need to ensure that the databases that will be analyzed are robust and full of data. In maintenance, this just means doing what the experts have been talking about for years; recording problem cause & effects, record solutions, record observations, and record what was done on the work order is sufficient detail to be worth analyzing. Nothing too hard, just what we are supposed to be doing.

Learn more about SAPPHIRENOW and secure your spot today!
Labels in this area