More answers on how to improve OEE, quality with advanced analytics

An analytics webcast March 29 explains how today’s Smart Factory leverages out-of-the-box and cloud-based overall equipment effectiveness (OEE) capabilities to accelerate improvement programs. Audience questions from the webcast are answered below.

By Cobus van Heerden and Gregory Dunn April 8, 2022
Courtesy: GE Digital


Learning Objectives

  • Understand details related to improving overall equipment effectiveness (OEE) and quality with advanced analytics.
  • Examine manufacturing questions and get answers about how analytics help.
  • Learn how industrial analytics are delivered and applied, as outlined from audience questions.

Industrial data analytics focus on process optimization, and the use of analytics allows a prescriptive approach to assets and production. A March 29 webcast with Cobus van Heerden, senior product manager – analytics and machine learning, GE Digital, looked at how to “Drastically improve OEE and quality with advanced analytics.” Gregory Dunn, senior staff technical product manager, GE Digital, helped van Heerden with the audience questions in the webcast and below.

Cobus van Heerden (left), senior product manager – analytics and machine learning, GE Digital; Gregory Dunn, senior staff technical product manager, GE Digital. Courtesy: GE Digital

Cobus van Heerden (left), senior product manager – analytics and machine learning, GE Digital; Gregory Dunn, senior staff technical product manager, GE Digital. Courtesy: GE Digital

Questions, answers on industrial data analytics

Question: What’s the best way to prioritize digitization efforts if I’m managing multiple plants?

van Heerden: To manage digitalization at multiple plants:

  1. Ensure you have automated, digitized and standardized your core operations using proven operational technology (OT) software from trusted sustainable global vendors like supervisory control and data acquisition (SCADA), manufacturing execution systems (MES) and historian software, and are calculating, benchmarking and tracking your production efficiency (OEE).
  2. Adopt proven easy-to-use analytics software from trusted sustainable global vendors to empower your existing operators and engineers to access and combine data from all or any of these OT systems, as required to start to continuously optimize your performance and efficiency OEE. Focus on your optimization efforts on plants where you have data, and much potential for improvement (low OEE).

Question: When can we start seeing return on investment (ROI) from implementing this type of solution?

van Heerden: Depending on the application, ROI can be visible in weeks to months, with significant improvements in one year.

Question: About organizational silos: How can we use analytics to collaborate rather than compete on resources to optimize?

van Heerden: Issues are seldom unique, and often repeat or are applicable to other machines, lines, and plants. Performance improvements or issues solved by one engineer on one machine, line, and plant can be shared and implemented on other relevant machines, lines, and plants. In this way, the right, easy-to-use analytics solution provides an innovation/intelligence/optimization layer on top of existing OT software to empower existing engineers and operators to collaborate for continuous improvements of OEE. That creates a positive snowball effect.

Question: Can you talk about using analytics to balance productivity (the #1 challenge in your survey) with other challenges, such as cost reduction, safety and compliance?         

van Heerden: Productivity is often the top challenge in more mature companies where issues like safety and compliance are already at sufficient levels, and productivity improvements can help them produce and sell more, or produce same amounts with more efficiency (at lower cost, etc.) Optimizing OEE with analytics help to utilize the manufacturing plant to maximum capacity (produce more), and/or to produce at maximum efficiency when you choose to. This will help reduce the cost/product unit produced and even improve safety as variation is removed from operations. It helps compliance, as quality will be good, improving customer and brand satisfaction.

Question: How can we ensure global analytic dashboard conclusions and plant-floor level conclusions agree?

van Heerden: Agree by making sure they’re based on the same reliable (single version of the truth) data.   A good data management strategy can help ensure this.

Question: When applying analytics tools, is there a need to or value in re-assessing OEE measurements?

van Heerden: OEE is continuously calculated by OEE or MES software while optimization is typically done by analytics software. When analytics software implements improvements, for example, to optimize recipe settings, the OEE should reflect and quantify those improvements. An example showed how resulting quality improved by 10% with total OEE improving by 5%.

Question: As organizations use analytics to advance in their maturity model from descriptive to prescriptive, what are some human-level skills to help the software?        

van Heerden: The most valuable skill of operators and engineers is their process domain knowledge.  The right easy- to-use analytics software will enhance/empower them to leverage their domain knowledge to accelerate the speed at which they can prevent, or troubleshoot and resolve issues that arise. Optimizing process performance using easy-to-use analytics software, doesn’t require them to be data scientists, and effectively makes them better at what they’ve been already doing. They are not replaced by analytics, but empowered. They become the superheroes in the story to deliver productivity improvements using analytics.

Question: Does finding the right mix of on premise or cloud-based analytics vary by industry, organization size, or primarily regulatory, criticality, and timing as discussed?

van Heerden: The mix of on-premise and cloud analytics is determined primarily regulatory, criticality and timing. Generally speaking, you will always require critical, real-time, close-loop analytics to run on-premise for safety, security and reliability reasons, and non-critical, non-real-time, enterprise or cross-plant analytics suitable for cloud execution.

Question: Have you seen analytics tie into existing alarm and notifications, or is it best to keep analytics alarms separate?

van Heerden: There should be a clear distinction between alarms and alerts. By definition alarms demand an immediate response and are produced by SCADA and distributed control systems (DCS) or safety systems. Alerts are non-critical and can be of varying severities and more suited for optimizing analytics applications that should not be mission critical. Generally speaking, critical alarms should be configured in a SCADA system or DCS, and analytics should be for optimization not mission critical or safety controls, and therefore will more often generate alerts, rather than alarms.

However, analytics software can and should be able to produce alarms back into the SCADA/control system. If an important issue is detected earlier than traditional SCADA/DCS simple alarm thresholds, the analytic software can generate an alarm that will provide more time for prevention or resolution. Or better even if the issues are detected with high confidence and can be safely avoided by switching something on/off or changing a setpoint, the analytic software can potentially do that automatically.

Question: Are there situations where the path toward optimization can seem counterintuitive, and how do you create trust/buy-in?  

van Heerden: Sometimes intuition and experience can be wrong, and that’s why data-driven optimization is so valuable because reliable data cannot lie. I’ve seen many examples where analytics mine new insights from data that is counter to what the plant operators expected and their experiences were. But also, more often it confirms and quantifies what they already suspected. You create trust by interactive quick improvements that create value for the business. Avoid trying to do large and perfect data-science projects. Focus on use cases with a solid business case and keep it simple. Yes, with analytics it is even more important to keep it simple, since with higher the complexity, the more difficult and costly it is to maintain in the long term.

Question: As we empower the people with data analytics, do you have advice about keeping plant-floor key performance indicators (KPIs) in line with organizational and supply chain KPIs?  

van Heerden: Keep KPIs in line by making sure they’re based on the same reliable (single version of the truth) data. A good data management strategy can help ensure this.

Question: What are out-of-the-box capabilities?          

van Heerden:  Different vendors offer different out-of-the-box (OOTB) analytics capabilities. GE Digital analytics offerings can help manufacturing companies analyze, monitor, predict, simulate and optimize OEE out of the box.

Question: How does the analytics software work?       

van Heerden: The software uses available OT data to analyze, monitor, predict, simulate and optimize OEE.

Question: Can you explain how the software provides closed-loop feedback to the equipment?

van Heerden:  The analytic software can write calculations, alerts, predictions or recommended settings and setpoint changes back to control systems in real-time closed loop via protocols like OPC.

Question: Do we need TEEP if we are calculating OEE?

Dunn: Total effective equipment performance (TEEP) and OEE provide separate information. Whether or not both are required depends on your use case. TEEP, where potential availability is calculated all the time for a given period, can be used, for example, to estimate capacity utilization. OEE, where potential availability only includes the scheduled production time, gives more focus on issues impacted expected production time.

Question: How does the tool or platform help operators or engineers interface with the system to get good information? Frequently the issue is not lack of data but more often lack of context relevant information such as reasons for downtime (reason codes), editing information that was incorrect, identifying issues that don’t make sense, etc.        

Dunn: During the webinar, we described operators adding more context through the existing tools (MES, SCADA, etc.) in situations where there were perhaps generic fault codes to provide more information. Another approach would be to provide additional OT data inputs to the analytic software, which can use to provide more context to the conditions that create the generic faults.

Cobus van Heerden is senior product manager – analytics and machine learning, and Gregory Dunn is senior staff technical product manager, both at GE Digital. Edited by Mark T. Hoske, content manager, Control Engineering, CFE Media and Technology,

KEYWORDS: Applying industrial analytics, on premise analytics, cloud analytics


Are you getting real-time answers with analytic software to optimize your processes?

Cobus van Heerden and Gregory Dunn
Author Bio: Cobus van Heerden is senior product manager – analytics and machine learning, and Gregory Dunn is senior staff technical product manager, both at GE Digital.