Our Sites

Improving advanced manufacturing practices through AI's Bayesian network

Populated with accurate data, BN can help shops predict problems before they become serious

Artificial intelligence can improve shop floor efficiency.

Artificial intelligence can assist metal forming and fabrication companies on their journeys for continuous improvement. Getty Images

Artificial intelligence (AI) is a set of mathematical models we use to predict possible outcomes of observed conditions and events. Given good information and assumptions, both employees and AI can predict outcomes. If a set of current conditions indicates potential failure, we can act to prevent a failure from happening. Manufacturers also can learn to manage the efficiency and quality of systems by observing how conditions and events influence time, equipment wear, and product quality.

With experience, we learn awareness of events and conditions in our plant environment. As our experience matures, we learn the possibility of a given set of events and conditions resulting in certain outcomes. Computational models can perform the same service by capturing events and conditions, then calculating the probability of certain consequences. If the probability of an anticipated outcome is unacceptable, our computers can inform us of a condition needing attention or address the situation themselves. This, along with collecting meaningful volumes of relevant data, is the core of AI.

One mathematical model employed in AI is the Bayesian network (BN), which is a graph that defines the relationships between conditions or events and their possible consequences. The conditions or events are random variables that are identified on a BN as a node. The node also includes a table indicating the probability that the condition will be true or false. For instance, the BN in Figure 1 addresses the probability of rain.

When we take the next step, to define causes and effects, we show relationships between events. As an example, assume our experience with weather tells us:

  • There is typically a 50% chance that it will be cloudy (sounds like Michigan!).
  • We also know that if it’s cloudy, there is an 80% chance of rain. If it’s not cloudy, there is a 20% chance of rain.
  • If it’s cloudy, there is a 90% chance the sprinkler will be on. If it’s not cloudy, there is a 50% chance the sprinkler will be on.

Such relationships, called causal influences, go beyond true/false and can be demonstrated in a more detailed BN. Relationships between nodes are identified with lines, called edges, to indicate the effect of a node or combination of nodes on a given outcome. Outcomes show their probability of occurrence based on the possible combination of the nodes that influence the outcome.

The BN in Figure 2 illustrates causal influences for the given weather data and shows the probability of the grass being wet given possible combinations of conditions for sprinklers and rain.

Plugging information from your shop into this elementary type of BN can help you:

  • Identify the most probable causes of events.
  • Monitor activities that are likely to cause issues.
  • Determine what conditions are optimal for producing quality components efficiently and will lead to the least amount of unanticipated downtime.

Initially, the probabilities should be entered by experts with experience on the equipment and processes being measured. With enough data, you can populate the probabilities with reliable historical evidence. The greater the accuracy of your BN, the better you will be able to predict and respond to events before they become serious issues.

Your network will gather information and experience with your manufacturing environment. As time progresses, the BN will reflect an accurate view of the probability of conditions and events occurring. With accurate and comprehensive information, you will be able to identify the strongest relationships between events and their consequences. With this knowledge you can fine-tune your maintenance schedules, learn and reinforce good practices, and conduct troubleshoot activities with greater confidence.

AI offers tremendous opportunity to predict consequences and troubleshoot events. Your AI network can manage large volumes of information and offer feedback quickly. The BN is one of the many probability tools available to help your AI network support your goals. The more you understand how it works, the more you will understand your need to monitor the right equipment and events and ensure your monitors remain clean and calibrated.

About the Author
4M Partners LLC

Bill Frahm

President

P.O. Box 71191

Rochester Hills, MI 48307

248-506-5873