Valmet Corporation, which serves the pulp, paper and energy industries, is investing full-on in big data, analytics and the industrial internet. The company offers its customers comprehensive advisory services to improve the efficiency of their paper and pulp production. The most important thing is to minimise the time taken by service-related shutdowns. To achieve its goals, Valmet cooperates with Houston Analytics and Teradata. They collect data with sensors in their customers’ production paper machines, which is analysed to optimise how the machines are operated and when to service them.
A single roll moves a line ahead at a speed of more than 1500 metres per minute while squeezing the paper with huge force. This obviously causes wear but until now, nobody knew exactly how, when and why. The answer to these questions lay in the analysis of data from the tens of thousands of sensors attached to the paper machine. A prediction model derived from the analysis was used to identify the causes of roll wear and create a model for the efficient production use of the paper machine. This model has extended service intervals by tens of per cent. Substantial savings have been achieved by thus increasing the utilization rate. A model has also been developed for pulp production, which recommends what measures to take and when, and provides forecasts for production according to the model.
“More and more of our customers want to buy demanding servicing and maintenance services from the original equipment manufacturer, which would be us. A new paper machine costs from 50 to 200 million euros and the price of an entire project is from 100 to 400 million euros. Paper machines are very complicated and have a large number of very expensive parts that wear in production use. We wanted to develop a model that would allow us to identify the factors that influence their life cycles and the optimisation of production use,” says Pekka Linnonmaa, Director, Paper Technology, EMEA, at Valmet.
The sensor data collected by a paper machine is very good at predicting things and there is also a lot of it. Performance will not be a problem when the analytics tool IBM SPSS Modeler and the Teradata Aster are used efficiently together. This means that work can be focused on understanding the phenomena discovered in the data.
“Data can be used to identify which events and series of events are relevant to the wear of the roll surface. This forms the bases of a predictive statistical model that identifies in good time when servicing is required. The model can also be used to create scenarios about how the paper machine should be operated. Collaboration between data analysts and specialists, based on the collected data, is seamless and enables the use of data analyses to confirm and identify even new physical dependences,” says Joonas Isoketo, Chief Analyst at Houston Analytics.