Skip to content

3 Ways to Cut Chemical Manufacturing Costs With White-Box Machine Learning

By: Fero Labs Logo light
• October 2022
Micheile dot com Z Vprb Bm T8 QA unsplash

Between Europe’s energy crisis and global raw material price increases, chemical manufacturers face growing pressure to trim budgets. This article explores 3 ways white-box machine learning can be an effective tool for cost savings.

Under pressure

The chemical industry faces several challenges, which are putting sales volume and profits at risk.

First, increased energy and raw material costs put pressure on companies in every segment to raise prices. Yet by passing the cost on to the customer, they risk losing sales. Meanwhile, increased costs lead to a growing emphasis on yield, as rising costs force manufacturers to focus on the highest yield possible in order to optimize profit margins. Put simply, everyone is concerned about wasting valuable material.

At the same time, reclamations due to quality issues are a continual problem. These not only force the producer to re-work and re-ship, but can also lose them business, particularly bad in an environment where every delivery counts.

Now more than ever, every low-cost opportunity (compared to massive capital expenditures) to improve in the above mentioned areas is highly welcome. Machine learning can be this opportunity. Unlike older optimization techniques such as Six Sigma, white-box machine learning software can analyze a process in real time. Based on that data, the software pinpoints ways to make production more efficient, from reducing energy consumption to improving yield and first time right percentages.

Here are 3 concrete use cases where white-box machine learning can improve your bottom line:

1. Boosting first time quality percentage

No one ever wants to ship a bad product, but in today’s economy, doing so is more costly than ever. Any item that fails customer quality standards will need to be re-worked and re-shipped, eating away at your already slim profit margin. Sometimes, re-working the product isn’t even possible.

Additionally, with increased prices, your customer may already be looking for a reason to get out of business with you. Quality issues could give them that reason. If this happens over and over, it could steadily decrease your volume—and you want to book every sale you can in these economically challenging times.

White-box machine learning can prevent quality issues. Its capacity to analyze a process in real time means you can see when products are going to be low-quality—and fix them before it’s too late.

Conventionally, if you receive a shipment of inferior raw materials, you’re not able to do anything about them. With a white-box machine learning solution, however, engineers can model the initial processing units upstream of the main reactors and learn which operational changes would improve reactor feed quality. For instance, if the raw feed comes in too concentrated, they can increase the amount of solvent used to dilute it. This way, production teams can quickly and easily adjust production for each new shipment of raw material to ensure quality never dips below the usual standard.

2. Reducing energy usage

Even if you have your production quality under control, rising energy prices may be eating at your profit margin and forcing you to push the boundaries with price increases.

Traditionally, significant energy savings in the chemical industry are connected to significant capital expenditure projects. However, in today’s crisis, liquidity matters and energy prices are so high that every saving counts. In such an environment, low-cost options like software can be a valuable option to generate savings without spending a lot of capital.

White-box machine learning software can optimize your process to be as efficient as possible, so you can get the same yield at lower temperatures. One challenge faced by continuous commodity chemical factories is prolonging catalyst life while keeping energy costs low and quality within specification. A white-box analysis can quickly model catalyst performance and tell engineers how to adjust feed rates and steam to operate more energy-efficiently. This allows them to save unplanned downtime and energy costs for time periods where they might be otherwise operating with highly inefficient catalyst.

Internal data scientists can set up similar recommendation engines; however, typically, this takes weeks and lengthy back-and-forth collaboration between the engineering and data science teams. A machine learning analysis can be set up in as little as two hours and doesn’t require input from data scientists, hastening the time between idea and efficiency improvement.

3. Maximizing yield and recovery

Since every producer in the chemical industry faces the challenges of increased energy cost and supply shocks, your raw material prices have likely also skyrocketed. That means every loss caused by suboptimal yield cuts directly into your P&L.

White-box machine learning can help maximize yields and increase raw material recovery. This is especially crucial in the specialty agricultural chemical process.

A white-box analysis can quickly scrutinize the distillation train and reactor operations and predict losses before they happen, then recommend process set points to prevent them. For a large chemical plant, this could allow them to recover as much as $400,000 of raw material that would otherwise be lost.

These are just a sampling of the many ways white-box machine learning algorithms can help chemical manufacturers cut costs. As energy and raw material costs show no sign of abating, efficiency is likely to be the buzzword for the near future. Fortunately, new technology provides manufacturers a chance not only to survive these tough times, but to emerge ahead of competitors.