5 Corporate Sustainability Goals You Can Hit With ML
With manufacturing accounting for as much as 45% of global emissions, and facing growing pressure from both regulators and society, industrial firms are setting ambitious goals to become more sustainable. It’s now common for digital executives and whole teams to be tasked with reducing greenhouse gas emissions.
To achieve these goals, both existing and new facilities must cut down on waste and energy usage. This means manufacturers have to optimize their processes in increasingly challenging conditions, and they need to optimize fast.
This is where machine learning can play a crucial role. In contrast to offline process optimization methods such as Six Sigma, white-box machine learning algorithms can rapidly “understand” the production process and recommend concrete steps to achieve specific goals. By creating a more efficient process, such solutions can both optimize for costs and emissions, by as much as 10%. And, crucially, this software-based approach to sustainability can be downloaded at the push of a button and work with existing hardware systems already in place.
In this post, we examine 5 common corporate sustainability goals—from minimizing raw material use and energy consumption to cutting out waste—and reveal how a machine learning solution can help teams achieve them.
1. Minimize virgin feedstock in raw material usage
How ML helps: Reveals exactly how much virgin feedstock to use in recycled processes
Traditionally, plants use more raw materials than absolutely needed, as a way of erring on the side of caution and making sure product is up to spec. White-box machine learning algorithms can alert operators in real time how much ingredient needs to be added to each heat or batch, reducing “overdesign margins” while still making sure the products are high quality. Using fewer virgin raw materials has a positive impact on reducing Scope 3 emissions of manufacturers, percolating up the entire supply chain.
In recycled EAF steel production, for example, scrap steel is melted down and re-rolled. Along the way, various alloys are added to ensure the final product has the desired composition. Since it costs thousands of dollars to remelt a flawed batch of steel, unnecessarily high amounts of alloys are typically added to minimize that risk.
With machine learning, however, operators can know exactly how much alloy they need to add, achieving as much as a 10% reduction in raw materials—and valuable cost savings.
Some manufacturers can choose between using new and recycled materials. It’s often trickier to use the latter, since these come with a high degree of variability. Machine learning can help operators adapt production, making the necessary changes to ensure that the quality of the products doesn’t suffer from the variability inherent in the ingredients.
2. Reduce waste
How ML helps: Prevents mistakes that would result in scrapped products
Mistakes are costly and wasteful, for both the environment and the company’s bottom line. Say a CPG company makes detergent in batches and then bottles it for consumer use. Once it’s bottled, if any product properties are discovered to be beyond spec—perhaps the viscosity is too high, or the pH is off—the entire shipment of bottled goods must be scrapped.
With white-box machine learning software, any problematic properties can be predicted during production. Hours ahead of time, operators will get an alert telling them the issue and how to fix it before it’s too late—meaning the problem can be fixed at no environmental or financial cost. This directly minimizes Scope 1 impact of the plant.
3. Reduce energy and heat use
How ML helps: Reveals how long to perform a specific step in the process
Reducing energy use is a common sustainability goal, as industrial heat accounts for nearly half of manufacturing emissions.
Just as manufacturers typically use raw materials too generously, the same is true for energy, with operators believing it’s better to overdo a process than under-do it and risk being under spec. If a paste must be dried to a specific level, a candy manufacturer might over-dry it, just to be safe.
Machine learning software can avoid such inefficiencies, telling you how long to dry the ingredient. Thus, both money and energy consumption are saved—another direct Scope 1 minimization for the facility.
4. Increase use of alternative energy sources
How ML helps: Adapts to rapidly shifting factory conditions, reducing uncertainty and process instability
Many companies are shifting towards new mixes of energy sources, striving to meet their green goals. However, this also introduces variability into the process, which can result in new kinds of waste.
In the cement industry, for example, many companies are achieving their sustainability benchmarks by changing the energy they use to produce heat, often to waste energy sources. This has a strong environmental benefit, avoiding landfilling and also getting energy from waste and even biogenic sources. But introducing these fuel can create variability and process instability, resulting in overconsumption of energy and potentially lower quality or off-spec product.
With machine learning-powered recommendations, cement companies can understand the parameters needed to minimize process instabilities, reducing uncertainty and allowing to maximize the substitution of fossil fuels with waste derived energy.
5. Reduce water use
How ML helps: Optimizes equipment to use (and thereby contaminate) as little water as possible
Increasing access to clean water is a key UN Sustainable Development Goal. With 2.2 billion people around the world still lacking safely managed drinking water, limiting high-volume industrial use can prevent water scarcity and contamination.
Many companies have publicized their aim to use less water, corresponding with this goal. How can this be achieved with machine learning?
One example falls under the category of waste reduction. If a product is scrapped, all the water that went into creating that product is wasted. Preventing mistakes that cause scrapped heats, therefore, simultaneously reduces water use.
Another example is optimizing equipment that relies on water. Say you have a blancher or some other piece of equipment that uses water in direct contact with a product, and thus contaminates the water. Optimizing the equipment to use as little water as possible produces as little contaminated water, which is then easier to treat.
Manufacturing has evolved far beyond the days when waste was ignored. Today some version of the above goals appear on most manufacturers’ sites, testifying to the increasing importance of sustainability. But amid pressure to reduce costs and increasing demand, actually delivering on those goals will be a challenge. With a machine learning solution, manufacturers can build a more efficient process that jointly optimizes for cost and sustainability, killing two birds with one stone.