Background

In one of the distillation units, one of the product temperatures is being lower than expected for around 1 year, leading to lower product recovery, as the separation is not as efficient as it should. 

This means a 10% recovery loss in the company, which is an important economical loss.

Challenge

Troubleshoot the causes for this deterioration in performance, by detecting the influencing variables and the respective influence of each.

Solution

-Load all the related tags, as well as the key temperature

-Run the correlations engine to find correlations against the potentially correlated variables and against all the variables of the asset

-Layer 2 periods (high and low recovery) and compare statistics and trends of the potentially influencing parameters

Results and value

-Two process variables were identified as the main influencing factors. 

-It was identified that both variables weren't acting at the same time, but separately. That's why by themselves, the correlation is not to high, but combined, the change in the temperature can be explained.

-A linear regression equation based on the 2 most influencing factors was built, to predict temperature (and thus, recovery) based on them.

-In the last year, the deterioration in product recovery was around 10%, which can be prevented with TrendMiner.