A new industrial age is being propelled by companies wanting their assets to generate more revenue without further investment or infrastructure upgrades. Artificial intelligence and the industrial internet of things can make this a reality. With a system intelligently assessing conditions that affect manufacturing processes — driven by a flow of real-time data from connected devices — machines can learn and the environment itself can “make decisions.”
This allows operations to improve with little or no direct involvement from personnel, leading to lower costs and downtime, and an ability to produce faster, as well as a slew of other benefits.
Sounds good, but there’s more to this.
Large data sets are too time-consuming for standard analytics to process, especially if attempted manually. AI is used to find correlations and the cause to specific processes. Add in a good application performance management system and AI algorithms can offer advanced analytics that deliver a clear view of business outcomes, even what the future may hold.
It’s an exciting time and a lot of companies are ready to rush right in. But if AI was simple and success guaranteed, everybody would already be on board. It’s an evolving field, and if not done right, it can go very wrong. Before turning these new technologies loose, consider the following questions.
1. What are we trying to solve?
Not identifying key business pain points to solve is a reason many AI pilots flounder. The thing is, even when these initiatives appear successful, they will stall at some point. You have to know what you’re trying to achieve and, most importantly, make sure leadership is aware. This will enable you to continue, despite obstacles. Here are a few examples that grab executives’ attention and commitment:
- Reduce unplanned downtime: Forecast performance metrics and schedule maintenance to keep operations up and running.
- Reduce energy costs: Take advantage of off-peak energy prices.
- Reduce production material cost: Purchase and use resources more cost-effectively, such as lowering chemical dosing amounts.
2. What improvements will be reached?
When pilots succeed but don’t progress, it’s often because results weren’t as powerful as anticipated. The fact is results are still positive even when performance improvements weren’t obtained but a clear reason why is determined.
The challenge is to find a project with which everyone feels comfortable — getting some kind of pilot off the ground just to get an evaluation started is actually reasonable. This is where concrete, meaningful improvement goals become important. Your solution provider should lead this charge since they know what’s possible.
3. What access to data will you have?
When it comes to data, three key aspects make up the backbone of an AI project — quantity, quality and access. AI projects use historical data in order to train algorithms to predict future outcomes. The more data the better. It may not all come into play, but data scientists will want to tease out any and all correlations and look for causal effects, so access is crucial.
Even so, while less data poses challenges, project goals can still be met. Even gaps in data — such as a lack of one or more sensor inputs — can be overcome. It’s important to know what you have to work with, so bring in a data science team to conduct an investigation before beginning.
4. Do we have data scientists and subject matter experts?
It’s important to involve, and have strong collaboration between, data scientists and subject matter experts (SMEs) who understand the process to be optimized. Without this, the project will likely fail. Some solution providers have good AI expertise, others have SMEs. These types of projects require a combination of both.
5. How do we proceed?
There’s a lot of approaches you can take to evaluate and execute a plan. Do you involve an analytics company if have your own SME? Should a consulting engineering firm organize the project? Do you get a one-stop solution provider to do the whole thing?
All of these are viable options. The key is to know the analysis can be done, and access to historical and near-real-time data is crucial.
Data analysis should be completed and vetted up front. Your team or provider must be able to tell you, within certain limits, that you’ll get the prescriptive recommendations necessary to meet your project goals. If a significant payment is needed before any analysis occurs, you could be funding someone else’s learning curve.
Improving your processes is a process. The key is to be realistic, patient and persistent.
All IoT Agenda network contributors are responsible for the content and accuracy of their posts. Opinions are of the writers and do not necessarily convey the thoughts of IoT Agenda.