Analysis of operating parameters and data on previous breakdowns obtained from 410 sensors allowed to identify patterns in the operation of the paper machine
“The long, laborious process of threading the web after a break leads to significant losses in productivity due to downtime,” says Andrey Nilov, head of the paper mill at Segezha PPM (Karelia, Russia).
“Receiving predictive information from the PM’s digital twin, the operator promptly changes the technological parameters of the paper machine and resumes the load after localizing the problem.”
“The mathematical model has learned to analyze the readings of the machine and warn the operator about the risk of breakage,” says Dmitry Bocharov, chief internal auditor of Segezha Group.
“Based on the results of the pilot project, the model was able to predict more than 60% of possible breaks.”
Since PM is a complex piece of equipment with many operating parameters, the human operator has limited ability to efficiently analyze multiple sensor readings, establish relationships between data streams, and predict breaks.
This problem was solved with more noticeable success by a mathematical model developed on the basis of machine learning, created at the Segezha PPM during a pilot project.
At the same time, machine learning implies continual improvement of the model as it constantly learns and improves performance.
“A key criterion for evaluating a new digital twin technology is the ability to solve a specific business problem, eliminate the causes of deficiencies in processes and reduce risks,” says Pavel Vakhnin, board member, vice president for information technology and process automation at Segezha Group.
“Currently Segezha is working on a full-scale implementation of the technology with advanced predictive analytics functionality on all its paper machines.”