(58b) Fixing AI’s Identity Crisis in the Industrial Space

Managing chemical processes at their optimal can often come undone when raw material variability, energy overages, recurrent and unplanned downtime, maintenance issues and ageing assets come into play. The sector is also under tremendous pressure to increase throughput while protecting their margins. Beyond their plant floor, the chemicals industry is also facing a sustainability challenge. According to the U.S. Energy Information Administration, industrial companies account for about 10% of global total final energy consumption and 7% of greenhouse gas emissions.

These challenges mean that the industry needs to optimize their batch processes in order to improve their top and bottom-line results. Artificial Intelligence (AI) and Machine Learning (ML) enable industrials to predict how to run their processes more efficiently to improve yield, quality and reduce downtime. As a result, AI is helping the industrials to reduce variation between batches and from plant to plant.

However, deploying AI is not without challenges. If not implemented properly, it can prevent the company from realizing the benefits of the deployment. Gartner says that 85% of the AI projects will continue to fail by 2022. Right from selecting the right AI use case to empowering the industrial workforce to use AI-powered systems and utilizing the right technology, the industry needs the benefit of best practices to achieve the benefits from AI.

This session will focus on three key best practices critical to taking the benefits of AI from aspiration to reality.

1. Understand your problems

One of the prerequisites of a successful AI implementation is identifying the operational challenges that keep you awake at night. How has the plant surprised you (good or bad) and do you know why? Are you satisfied with how your assets are performing? How well do you understand your data and why? By interrogating these questions helps you to identify use cases that will be both A) best suited for AI and 2) will deliver impactful value to your operations. One of the critical reasons why AI projects fail is because it is chosen to solve random problems. It is crucial to identify areas that will generate savings and revenue and have measurable Return on Investments (ROI). The use case should also be scalable in other functions or locations to increase the returns.

2. Understand the critical difference between AI and Industrial AI

Robust technology to manage scale and capability is required to ensure the success of AI initiatives. This is why it is critical to understand the difference between generic AI platforms and fit-for-purpose Industrial AI platforms. While generic AI platforms that are designed to be used by all types of businesses and industries (i.e. finance, healthcare, retail) and are built for common functions. Industrial companies need an Industrial AI platform that helps in resolving their specific problems. A generic AI platform is typically built to be used by data scientists and require coding skills. However, industrial companies are competing against a global shortage of data scientists and may not want to invest in hiring teams of them.

Industrials require an Industrial AI platform made especially for them because industrial data is complex with several variables and is collated from different data sources. Organizations generate around 3TB of data each day, but it is believed that approximately 40% of it is spurious and cannot be used. A generic AI platform is not designed to collate and work with data from so many different sources and the amount of noise that needs to be extracted from data. It is estimated that a single AI prediction can cost over $1000, which can become a costly and time-consuming exercise if your models feeding the predictions are built on noisy data and are not explainable. Instead, in order to achieve their desired impacts, industrial companies need an Industrial AI platform that cuts through noise and can extract clear visual trends to observe and action, prevent undesirable events before they happen, and go beyond model predictive control.

3. Invest in augmenting your workforce with AI

Possibly the most significant challenge faced by the industry in scaling AI is the lack of digital talent. There is an acute global shortage of data scientists, and it will be a few years before this problem is resolved. One solution is for industrials to hire consultants to execute AI deployments. But, even then, this approach may backfire because usually, the consultants are not domain experts and therefore can exhaust a lot of internal resource to contextualize the data so that it can be applied to the company’s operational processes and assets. For AI to truly succeed and become ubiquitous across their operations, industrials need to augment their workforce with AI skills to reduce dependency on black-box tools and consultants. This is possible with today’s no-code Industrial AI platforms, which automate the majority of data science functions, putting the operators in charge of their data and how best to extract valuable insights. The shortage of data scientists should not delay industrial companies from leveraging AI to realize their net-zero ambitions. This need is further heightened when the aging workforce comes into consideration, whereby institutionalizing knowledge and removing variability across plants becomes a priority.

This presentation will include real-world case studies of industrial companies applying AI to their processes including some batch operations. Examples that will be showcased include how a food manufacturer is using AI to accurately predict when the production of its animal feed additive product was completed in order to reduce batch-to-batch variability and maximize asset utilization.

By applying these three best practices, closing the gap in process variability is within reach with AI.