Inside Advanced Scale Challenges|Wednesday, December 12, 2018
  • Subscribe to EnterpriseTech Weekly Updates: Subscribe by email

The Process Industry: 4 AI Questions to Ask before Implementation 

(Source: Shutterstock/branislavpudar)

Food processing, beverage production, chemical manufacturing, water treatment, and other verticals in the process industry know that performance monitoring of industrial assets and processes plays a critical role. It’s all about monitoring the key parameters, predicting what’s going to happen next, and taking the right action to increase operational efficiency. Connected industrial assets generate pools of data that can be used to predict failures before they occur and optimize processes for increased production.

An organization that connects their asset data with financial metrics can gain insights that will enable them to achieve better business outcomes and increase profit margins.

However, turning large amounts of raw data into human-readable, actionable intelligence is the challenge. Data types differ greatly, especially across industries, and successfully extracting information from those varied datasets can mean the difference between experiencing a production line shut down and avoiding one. The extent to which AI and machine learning (ML) can help the process industry is profound. Drawing insights from enormous amounts of data is where these technologies shine. Plant operators no longer need to rely on just their instincts and whatever information their legacy systems can gather to make decisions. AI is critical to effective Asset Performance Management (APM).

But as with everything, there is a right way to go about implementations and a much costlier, wrong way. Here are the questions organizations should ask when evaluating the implementation of an AI solution to monitor and predict the behavior of  industrial processes.

What is the business goal?

Organizations can often rush to deploy an emerging technology, enamored by marketing-driven luster. Deploying technology just for the sake of it usually leads to a failed project. Defining the problem beforehand will allow the organization to understand the goal. It’s important to understand how the technology will help them overcome a specific problem and identifying the issue will enable that organization to define what goal they want to achieve and how to get there.

Outlining specific objectives and showing how results will be achieved is also a good way to gain executive buy-in for the pilot. A clearly stated roadmap from start to successful finish greatly improves the odds of a project getting approved.

Do you have the data to support the project?

AI and ML project failures are often the result of an organization learning too late they don’t have the necessary data. There are three key factors when it comes to data: quantity, quality and access.

Since AI projects use historical data to train algorithms that can predict future outcomes, the more data the better. It’s always best to have too much than not enough. While not all data may get used, access to it allows data scientists to flush out any correlations and identify even causal effects that can prove useful. A lack of data can make this step challenging, but it doesn’t necessarily lead to failure. Gaps in data can be overcome using data science techniques.

Do you have data scientists and subject matter experts?

Data scientists are critical to any data project. However, they are not the only required experts. As vital is a strong partnership with subject matter experts who understand the business goals and what parts of the process are being optimized. Without their collaboration, the project will likely fail.

Data projects for the foreseeable future will require the knowledge of both to be successful. It’s an evolving technology realm and skill sets will need to adjust. This may eventually lead to a new blended role wherein a single person possesses both data science and the domain expertise necessary to build a new generation of machine learning models.

Can a solutions provider help?

Unless you possess deep AI expertise in-house, an organization will need to seek an outside source to help guide the evaluation and execution of a project. Choices range from specialized analytics or consulting engineering firms, to all-in-one providers capable of end-to-end support. Whatever route you go, the key is to identify the entity with proven experience who can provide those services within a reasonable budget.

But be sure not to be “penny wise and pound foolish” – ensure that you’re making an investment that will produce results.

In vetting any firm, be sure they can demonstrate an initial ability to perform the data analysis. They should also be able to ascertain, within limits, that the analysis will produce prescriptive recommendations to support the project goals. Finally, if a firm asks for a large fee up front, chances are that payment is funding the firm’s learning curve.

AI and ML represent much more than just emerging technologies still trying to find a place in the business world. Though in their infancy, they are not just improving bottom lines and efficiency, they are changing the way things get done in the process industry.

This will result in new job responsibilities, capabilities, and entire evolutions of business models. All of these will prove requirements for succeeding in the changing and more competitive landscape of our modern global economy.

Prateek Joshi is the founder and CEO of Plutoshift.

Add a Comment

Do NOT follow this link or you will be banned from the site!
Share This