About Valmet
Valmet is the leading global developer and supplier of process technologies, automation and services for the pulp, paper and energy industries. The company has over 200 years of industrial history. Valmet Industrial Internet is a set of data-driven services that combine advanced monitoring and prediction applications, Advanced Process Controls (APC), dynamic process simulators and remote services from Valmet Performance Center into comprehensive solutions.
Antti Sirkka, Chief Data Architect at Valmet, recently presented the company’s success story at Data Innovation Summit Sweden. Antti explained how Valmet Industrial Internet provides value to customers by connecting data from different plant-level and process automation systems. As a result, customers can move towards autonomous mills and plants. Customers become resource-efficient, flexible in operations, clean and safe, and less dependent on operating location.
The key elements of Valmet Industrial Internet are Industrial Internet applications, the Valmet Performance Center and the Valmet Customer Portal. In addition, a solution ecosystem brings industry-leading players and startups together to co-create new, data-driven solutions.
Ensuring data quality
As Antti Sirkka, Chief Data Architect at Valmet sought to move the business into a cloud data platform, his first step was to establish the data architecture principles. Valmet’s data architecture needed to be multi-tenant, data-agnostic, multi-tier and support virtual private clouds.
Valmet’s modern paper machines generated a large amount of data, with 20,000 – 50,000 different process tags and 5-10 TB of raw data per year. Scalar data from one line in one mill generated 1 billion rows and 75 GB of data per year. Profile data from one line in one mill generated 100 million rows and 15 GB of data per year.
Most of the data comes in from IoT sensors, with multiple sources transmitting data constantly. Data from 200 mill production lines generate 20 TB of compressed data per year. The data platform would need to support hundreds of concurrent queries from thousands of end users.
Sirkka established a set of metrics his team would measure to ensure data quality: completeness, consistency, timeliness, integrity, accuracy and standardization.
Powering end-user applications via the Snowflake Data Cloud
After a thorough evaluation of several data platform solutions in the market, Valmet chose the Snowflake Data Cloud. According to Sirkka, “Snowflake was the best fit. It’s secure and provides governed access to all data. It provides secure and governed data sharing. Another big factor was concurrency and scalability. We are able to auto-scale during peak load hours.”
Data from various automation systems is loaded into Amazon S3. Data is then ingested from S3 into Snowflake. Data is moved from a staging area to a raw data area. Business rules are applied to transform the data, with the help of machine learning. The transformed data is moved to a curated data area, which feeds application data out to end-user applications.
Data-driven cost savings in the printing process
Snowflake serves as the foundation for a Valmet application called Anomaly Detector. The application detects machine failures, web breaks and their root causes. “Web breaks” refers to the breaking of paper that occurs on a printing press during production.
The Anomaly Detector uses data to highlight key challenges that optimize the paper making process. This application helps customers prevent deviations in their process, and has been proven to catch 15-20% of potential web breaks. The improved efficiency generated by Anomaly Detection generates an annual savings of 300,000 euros.
Generating paper fiber cost savings using data
The Online Quality Predictor application uses data from automation and other process systems to predict real-time paper strength levels and alert operators to take necessary actions. The result of these operator actions is a 1-7% bottom line savings in paper fiber depending on produced grade.
Dashboards provided by the application display multiple properties and their predicted strength values and trends in near-real time. The “operator advisor view” in one dashboard helps operators understand how variables can be controlled and how to optimize the manufacturing process.
Conclusion?
As a set of data-driven services, Valmet Industrial Internet appreciates the importance of a cloud data platform to connect data across different plant-level and process automation systems. By connecting this data, customers can move towards autonomous mills and plants, becoming resource-efficient, flexible in operations and less dependent on operating location.
Snowflake was the best fit for Valmet because it is secure and provides governed access to all data. The concurrency and scalability of Snowflake enabled Valmet to auto-scale during peak load hours. Valmet built applications on top of Snowflake including Anomaly Detector and Online Quality Predictor. These applications saved costs and improved the manufacturing process.