He commented: “Within the process control environment there is a huge amount of data coming off the line – too much for any one person to analyse. There is a real danger that operators are being blinded by a blizzard of information and are missing real opportunities for process and quality improvement.”
He continued: “Manufacturing is complex and hard to control, and existing technologies often fail at pre-empting production glitches and fail to spot the changing conditions that can impact quality. Each process step can have hundreds of variables that affect the quality of the output and operational effectiveness.”
Mr Cronje said that with traditional statistical process control software, control limits are programmed by human experts and often fail to capture the complex interactions between all process variables. The solution, says Mr Cronje, is to use AI to find complex patterns in datasets and develop distinct operating paradigms for the production facility.
DataProphet’s Omni product is an AI solution to reduce the cost of non-quality for heavy industries, which can then produce control plans that bring production into a stable and optimised state – developing a path from the current plant state to the optimal state.
He continued: “All of this is possible, using cutting-edge machine learning techniques, which allow Omni to understand the effect that a small change in an arbitrary process step will have on all downstream process steps. Process steps are no longer operated and controlled in isolation and for each prescribed change, Omni can quantify the expected magnitude of its effect on production.
Crucially, Mr Cronje says that AI is delivering real ROI for heavy manufacturers. “We are seeing incredible results, such as a large foundry that has reduced internal and external scrap by more than $1 million per annum, and an automotive bodyshop has reduced stud and spot-welding faults by more than 50%.”
Mr Cronje also encouraged manufacturers to learn from historical manufacturing data and not just assume that historical data has no value.
“We need to learn from historical data, from PLCs and even old logbooks, to exploit the collective knowledge and experience of all those who have influenced the plant in the past. We need to institutionalise those years of operational experience for the benefit of the manufacturer’s current and future workforce – not lose them.”
DataProphet www.dataprophet.com