Heat exchanger fouling was a major issue for an oil and gas customer, leading to significant annual production losses.
Challenge
The customer faced production losses due to heat exchanger fouling, necessitating frequent cleaning and/or repair. Planning these cleanings was difficult, as individual heat exchanger performance wasn't consistently measured. Additionally, equipment without previous failures wasn't consistently monitored, increasing the risk of future failures.
Solution
Arundo trained a supervised machine learning model using data from a well-monitored heat exchanger to predict performance. This model was then applied to other heat exchangers without performance measurements, enabling a better understanding of their status. Cloud-based machine learning models were deployed to predict performance every 10 minutes using real-time data, and alerts were set up for operators in case of deteriorating performance.
Impact
The solution resulted in a reduction of 5-7 million USD in annual production losses from heat exchanger cleaning and failures.
