Umetrics Suite Blog

How multivariate data analysis can raise your manufacturing processes to new levels

September 5, 2017

multivariate data anlysis blog.jpg

Producing and distributing raw materials and foodstuffs with a low profit margin is a challenging business. One major supplier has made significant gains through applying multivariate data analysis (MVDA) to their manufacturing processes and logistics.

Like many manufacturers, the focus on modeling aims to boost yield (often with large economic savings), reduce carbon footprint, energy and water use, and increase capacity without excessive capital expenditure. Multivariate data analysis has become vital for their continuous improvement and maintenance of operational reliability because it helps them to pinpoint the key variables to focus on.

“It is important to get the expert operational staff on board when developing and maintaining models,” said the engineer who is driving the implementation of MVDA throughout the company. “Their expertise is vital to improving processes.”

While the main focus is on manufacturing processes, the company is also investigating the possibility of using multivariate data analysis in supply chain and business activities. Here are a number of case studies to illustrate the impact that MVDA using Umetric’s SIMCA and SIMCA-online analysis software is having on a broad range of their operations.

Boosting yield and operational efficiency in fermentation

Troubleshooting fermentation is very challenging due to the complexity of biological processes, and can involve hundreds of variables. The company has found that the MVDA approach provides them with a more structured basis for problem solving than just relying on gut feeling.

Indeed, multivariate data analysis software has been extremely useful in troubleshooting issues with yield to narrow down the top three or four factors. In one example, reduced yield was causing annual losses of over $2 million dollars. The initial feeling was that metabolic effects were to blame, but with MVDA the production team could drill down into the data and find the true cause of yield loss, which involved a handful of variables that could be addressed.

There are also operational benefits with multivariate predictions run online and linked to control charts. The software tool used by the production team (SIMCA-online) allows operators to monitor complex processes on just one screen instead of having to check fifteen or more, which can free up time and energy to focus on other issues.

Transporting animal feed

It is not only production that can lead to challenging and expensive problems. Full control over logistics also has a big effect on the bottom line. One example is an animal feed powdered product that was loaded into rail cars.

In recent months, Logistics had noted that the bulk density of the powder had dropped by 10%, lowering the amount of powder that could be packed into the rail cars, and resulting in an annual loss of around $1 million dollars. By applying MVDA to the problem, they determined in a couple of days that the problem was related to five or six variables controlling drying, and were able to solve the problem.

Dealing with solvent loss

Producing seed oil, such as granola and soya, relies on cracking and flaking the seeds, extracting the oil with hexane solvent, and distilling the extract to purify the oil. The residue from extraction is high-protein animal feed. Engineers were experiencing higher than normal solvent loss during the process, causing environmental and safety issues that threatened to limit production, with an annual loss of $500,000.

The plant engineers compared production runs with high and low solvent loss using a contribution plot and quickly developed a data-driven action plan based on a limited number of critical variables, including the vapor-capture system and flow rates.

Summing up

The key take-home message was that multivariate data analysis software is a key factor in improving the company’s operational functionality by pinpointing non-obvious issues that cause production loss.

Our contact at the company added, “I dream of having SIMCA-online installed in every manufacturing facility one day. Current users love the solutions and wonder how they ever managed without them. We have sixty active users and aim to double the number every couple of years”.

Learn more

Want to know more about the SIMCA multivariate data analysis processes and tools? Read more about how it works here


View Upcoming Webinars


Topics: Multivariate Data Analysis

Marie Wensley

Written by Marie Wensley

Marketing Manager at Sartorius Stedim Data Analytics

Search the Blog

    Subscribe to the Blog

    View the:

    Data Analytics Glossary of Terms

    List of Webinars

    Get a Free Trial