The global manufacturing sector has gone through significant change over the past several decades, with advances ranging from mass production to computerization and automation. The Industrial Internet of Things (IIoT) promises to be another evolution, allowing everything (or nearly everything) on the plant floor to be connected. The advantages of this inter-connectivity are numerous but, to me, the biggest benefit — and the biggest challenge — with IIoT is what to make of the millions of data points being generated.
While a majority of manufacturing companies are already generating enormous amounts of data, few are equipped to deal with it and, most often, struggle with where to get started. This is where artificial intelligence (AI) plays such a critical role; converting massive amounts of operational data into insights that can be used to dramatically reduce unscheduled downtime, optimize production processes, and optimize overall energy consumption.
In the context of IIoT and the interpretation of the millions of data points being generated, AI technology closes a critical gap in terms of accelerating insight. Traditional statistical methods are time consuming and require constant updating by a trained, highly skilled data scientist / analytics expert. In most cases, data generated from most modern manufacturing facilities is so complicated, produced so rapidly, and in such high volumes, that traditional systems are not designed to handle such volumes nor can they scale.
AI technology gives manufacturers a new competitive edge by enabling predictive models that automatically learn from each new data point, accelerating the analysis of large volumes of data. With AI-enabled data models, manufacturers can more accurately predict equipment failures, production yield, and resource allocation. By applying innovative AI technology, the initial data analysis can also be automated, driving further efficiency across the entire plant floor.
A leading global Fortune 500 Food & Ag company identified high output variability in one of their most expensive production processes, resulting in negative product margins. The existing systems were unable to handle the volume and frequency of the data being generated. Further, the standard data models generated were not configurable and required manual upkeep, which increased time to market on the already low-margin product.
Using AI technology, we were able to automate the entire data modeling process for the customer and help them optimize their production process by five percent, or roughly one million dollars in savings for a single process, in one plant. The customer is now looking to measure the overall savings associated with the energy reduction that the production equipment itself is being utilized more efficiently.
In another case, a global Fortune 1500 Food Processing company wanted to reduce fuel consumption of their gas-powered turbines by better predicting utilization. Since the plant requires large amounts of energy to run its processes, a small change in energy cost has a dramatic impact in the customer’s operating budget. For the customer, fuel costs are highly dependent upon being able to accurately predict consumption quantities. In order to minimize fuel costs, they needed to be able to optimize fuel usage.
Using the Canvass AI-enabled platform, our automated predictive models have helped the customer initially achieved a 4.11 percent reduction in fuel costs. In parallel, the customer is also able to reduce their greenhouse gas (GHG) emissions by several million pounds of CO2 per year.
One of the questions that I am asked most often when it comes to automation and AI is: “Where should we start?
Like most things in life, it is always best to take an incremental approach. Start with one project, or one plant, as most of our customers have done. From there, our customers are scaling into tons of processes and then into hundreds of plants. By taking a focused approach, you have the ability to be more agile; learn fast and scale efficiently.
In my experience, the learning never ends. It is not a matter of “one and done”, but a process of constant improvement over time. This is where automation plays a key role; these AI-based predictive models continue to learn and adapt.
