Data-driven insight: a win-win for manufacturing productivity and the environment
By Conor Clifford, Future Energy Ventures
Globally, industrial operations account for more than a third (37%) of final energy usage¹ and about a fifth of Greenhouse Gas emissions², so finding ways to drive decarbonisation³ throughout industry is a critical element in delivering a sustainable future for us all.
The decarbonisation of industry can be addressed from a number of angles. On the demand-side, as consumers, we can reuse and recycle more so that the need for primary products reduces (e.g. Trove which facilitates the circular economy for consumer brands). On the supply side products and processes can be re-designed to use less raw material and/or lower-carbon alternatives, so that the product itself has a lower carbon footprint per unit produced (e.g. Carbicrete which is developing cement-free concrete). Exchanging the primary energy inputs to production processes can also be considered, like replacing natural gas with green hydrogen, or a combination of electrification (e.g. Boston Metal is trading the blast furnace for Electrolysis in the production of aluminium) and renewable energy. All of this will be crucial in order to achieve net-zero carbon emissions by 2050 and much work has already been done. However, dramatic shifts from the modi operandi are still needed: large-scale changes in consumer behavior; many person-years of product development and marketing; and the continued transformation of the energy sector. Scaling these actions across industries will take plenty of time – which is not on our side – and money to achieve.
A simple concept in theory
Meanwhile there is something that can be done today, which will not only contribute to the overall decarbonisation goal, but will make manufacturers who adopt it more competitive as well – improved production efficiency! More specifically for manufacturers, production efficiency is embodied in the term Overall Equipment Effectiveness (OEE). It is an indicator of how efficiently they can produce a product of the required quality. So, increasing OEE is directly related to increased profitability, and at the same time inversely related to carbon emissions, because less scrap is generated and less energy has been used to create the same volume of saleable product.
While this is simple concept, and manufacturers have been running continuous improvement processes to eke out small percentage point gains in OEE for years, it is not so simplistic to achieve.
With the advent of Industrial IoT (or Industry 4.0) came the promise of transforming manufacturing through digitalization, with commensurate gains in OEE. There have been many companies, large and small, over the years that have built products and digital solutions to improve OEE, but none have achieved deep penetration into the market or scale, and we are still waiting for someone to crack the problem… or are we?
Why have so many companies tried and failed?
The challenge lies in mapping data, which is continuously streaming from the factory floor (a.k.a. OT data), into a so-called digital twin. Most companies approach the problem with the same top-down mental model; because machines are the basic building blocks of a production process, they must also be the basic building blocks of the digital twin. So, these companies start by creating templated models of each machine in a production line, which are then connected together to create the digital twin of a production process. This might work on a one-off project, but the problem arises when coming to the next production line, and then the next one. Because there are many thousands of different machines, there are many thousands of different templates required, which becomes a portability and scaling issue. At the same time, every plant floor is generating a continuous stream of messy OT data, which needs to be transformed and mapped to these many different templates. This gives companies the compound of complex challenges on both the data and modelling sides, and is the reason why so many have not succeeded so far.
How can we fix this?
At Future Energy Ventures, we think our portfolio company Sight Machine (a data platform for manufacturing) has cracked it by focusing on the problem in a different way. Instead of building templates for the thousands of machines, Sight Machine has broken down production processes (disregarding the machines for a moment) into four foundational building blocks, often referred to as Common Data Models, or Standardised Schema. By stream-processing OT data into these models, and performing scores of transformations continuously, Sight Machine’s platform turns this OT data into standardised, useful information.
These Common Data Models represent four universal aspects of Production: the work performed in each “cycle” of production (which can be denoted as a cycle in discrete processes, or a time period in continuous processes); each downtime event or unplanned interruption to work; each part; and each aspect of a part that is not intended, such as defects. Together these schema represent two basic but powerful dimensions of production:
- a structured and infinitely rich record of work done by assets on a product, i.e. each value-added step performed; and
- materials flow, which is represented by associating process data at each step to identifiers of materials, components, batches, and parts throughout the digital thread.
Under this approach, all plant data, whatever the source, is structured into four foundational building blocks and then combined to represent any machine, line, plant, or part. And since there are only a few building blocks with a common and consistent structure, the task of relating data across assets and activities to deliver insights becomes much less complex. This allows Sight Machines’ software to deliver the insights for manufacturers to leverage into focused improvements in production (OEE), and in turn reduce carbon emissions.
Sight Machine has applied this approach to over 20 industries with meaningful outcomes:
- Within the first weeks of implementing Sight Machine’s models and dashboards, a highly regarded paint manufacturer achieved a more than 7% lift in OEE at an already high-performing plant. Gains have continued at this and other plants in the manufacturer’s fleet, and work is extending to cover use of materials, water, and energy.
- An automotive manufacturer has used Sight Machine models along with machine learning techniques to resolve stubborn scrap issues. Scrap in this manufacturer’s process, typically run about 10% and in the first month of use, scrap was reduced by almost 50%. Applied to manufacturers and industries at scale, these kinds of gains will improve EBITDA margins disproportionately – by as much as 15-20% for a 5% improvement in yield – and will drive corresponding gains in sustainability.
- In heat-intensive industries, the opportunities for yield improvement are especially substantial. In the glass industry, for example, first-pass yield rarely exceeds 85%, which means at least 15% of the energy used is wasted. Sight Machine has developed approaches to predict defects that anticipate and help operators prevent about 80% of the defects in glass-making. Energy savings from these gains are sizable, and Sight Machine is actively partnering with one of the world’s leading glass automation providers to extend this work across the glass industry.
Industry players must seize opportunities to leverage policy actions from governments and accelerate our transition to a net-zero carbon emissions by 2050.
There are a number of actions that need to be taken to decarbonise industry towards the goal of net-zero, many of which will take considerable time and investment. However, there is one major point of leverage that is ready to be implemented and requires relatively little investment. It is all about effectively leveraging data to boost productivity, improve the bottom line and lower carbon emissions across industries and borders.
 GHG emissions are measured in carbon dioxide equivalent (CO2-eq.), and collectively referred to as carbon emissions, while the reduction in these emissions referred to as decarbonisation, within this article
 Data associated with machines and systems on a factory floor is commonly referred to as Operations Technology (OT) data, while data associated with back-office systems is commonly referred to as Information Technology (IT) data
Photo by Ameer Basheer on Unsplash