Seebo Solution Demo

Predict & prevent losses in production quality and waste

The problem

Charlie, a process engineer at a global chocolate bar manufacturer, was tasked with reducing quality and yield losses in the company’s wafer production line.

Charlie attempts to analyze the time series data from the company’s data historian, together with the quality results of the end product from the ERP system.

But lacking data science expertise, is unable to extract meaningful insights into the root causes of quality issues, such as broken wafers, overweight, and packaging faults. Charlie is also unable to predict when such issues will occur in the future to prevent them from happening.

The solution

In order to mitigate quality losses, Charlie and his production team must answer 3 questions:

  • WHAT are the most painful losses? Using digital twin analytics
  • WHY are these losses happening? Using automated root cause analysis
  • WHEN will these losses happen next? With process-based predictive analytics

Charlie turned to the Seebo Predictive Quality solution for the job.

PINPOINT WHY PREDICT WHEN DISCOVER WHAT Charlie models the production processes using Seebo’s quick, easy and code-free Digital Twin Modeler - generating a process-based data schema of the production line.

Machine learning algorithms are applied against this data schema to deliver accurate insights that consider the company's specific production flows and dependencies.

Capture the Production Context – Process and Assets

PINPOINT WHY PREDICT WHEN DISCOVER WHAT The insights are presented in the context of a Digital Twin of the production line, enabling Charlie to quickly prioritize the waste issues - based on their operational and economic impact.

Charlie must next uncover the root causes driving each of the top waste issues.

Prioritize the issues driving waste

PINPOINT WHY PREDICT WHEN DISCOVER WHAT The Seebo solution applies supervised machine learning algorithms to generate the primary suspects for root cause of events, prioritized according to their statistical contribution to the issue and presented in the digital twin.

The insights reveal 5 primary suspects causing the problem Charlie is currently investigating.

Automatic Root Cause Analysis

PINPOINT WHY PREDICT WHEN DISCOVER WHAT Charlie investigates further by using the composite root cause analysis, which indicates that the combination of the two parameters increases the probability of causing a quality problem by 283%.

Composite Cause Analysis

PINPOINT WHY PREDICT WHEN DISCOVER WHAT

Charlie learns that the combination of an increase in the baking temperature, and a parallel decrease in the baking conveyor speed are the most dominant causes of the problem.

This kind of information gives Charlie a good lead as to how to solve the problem that is taking place.

Understand how the composite cause analysis impacts the quality issue

PINPOINT WHY PREDICT WHEN DISCOVER WHAT Once Charlie and his team confirm the cause of the problem, Seebo applies predictive analytics to anticipate future occurrences of similar issues, making it possible for them to prevent the issue from occuring again.

Predictive analytics enables Charlie’s team to see warning alerts along with the statistical probability of the occurrence happening again along with the estimated time of occurrence.

Predict specific losses before they happen

PINPOINT WHY PREDICT WHEN DISCOVER WHAT The team can investigate the root causes of the prediction - in this case there are five. By addressing the root cause, the team is able to prevent the issue from happening.

Predict specific losses before they happen

PINPOINT WHY PREDICT WHEN DISCOVER WHAT Lastly, for operators to close the loop, Seebo presents the company's Standard Operating Procedure to address the issue at hand.

This gives them access to important organizational knowledge, at the right time.

Address the process inefficiency with Standard Operating Procedure