Batch On-Line Data Analytics

On-Line Data Analytics

At Emerson Exchange 2009, Lubrizol’s Bob Wojewodka and Emerson’s Terry Blevins presented, Benefits Achieved Using On-Line Data Analytics. Lubrizol and Emerson have worked together to develop on-line batch analytics as a beta test. The objective was to demonstrate on-line prediction of quality and economic parameters and evaluate different means of on-line fault detection and identification.

The intent was to document the benefits of this approach and learn from the test with data and usability feedback for product development of these batch analytics capabilities. Some of the challenges of applying online data analytics in a batch process are process holdups, access to lab data, variations in feedstock, varying operating conditions, concurrent batches, and assembly and organization of the data.

The foundation for this project was to form a multi-discipline, collaborative team that includes plant operations. They developed a workflow process to capture team input using an “input-process-output” data matrix to capture and explain the information required for the data analysis. It was important to integrate lab and truck shipment data by creating a workflow process with the plant’s SAP enterprise planning system.

Calculated property estimation was performed on feed tank quality and other non-directly measured properties. The instrumentation was surveyed and loop tuning performed to improve process control. And, a formal training program was established so that everyone was knowledgeable about the new work processes.

For the data analytics, it was important to identify among a wide number of inputs and process variables how these variables relate and which ones have the greatest impact on product quality. These analytics could also predict the end of batch quality while the batch was running. The team had good correlation between the predicted end of batch, what the lab samples indicated, and what the end of batch time actually turned out to be.

The analytics were developed for two batch processes, where the output of the first fed the second process, as well as provided finished output.

Over the three-month period, a process fault was detected using the on-line analytics. A problem was identified a problem with the mass flow meter. This occurred during the initial training, so the operations team quickly embraced these analytics. The on-line analytics also discovered a problem with a hot oil heating system. They discovered the process with the first batch and estimated it would have taken weeks to find with traditional methods. This alone paid for the efforts by the collaborative team’s time in this effort. The benefits have been seen from the operators through senior management.

Leave a Reply