Aratubo – Energy Efficiency and Processes
Aratubo
Sector: Industry
Business Case
Predict the energy consumption of the six different production lines associated with each manufacturing family. The problem is that we currently only have energy consumption per ton of pipe manufactured and per line. The challenge is to associate the energy consumption of each of the parts of the six production lines to the product family that is being manufactured at that moment in order to make decisions.
Objectives
To predict consumption across the six production lines linked to each product family and identify variations from the optimal levels: 1) Make our processes more efficient by optimising their operation (parameters, settings, etc.) and reducing energy consumption. 2) Identify and associate the energy costs derived from the production to each product family. 3) Adapt the production schedule to manufacture each family in the most efficient line. 4) Obtain consumption patterns and trends to detect and anticipate failures (predictive maintenance) and predict the end of life of the different components of the process.
Use case
We use analysers at various stages within each production line, collecting consumption data stored in a database hosted in Power Cloud. This consumption can be monitored using Grafana. We need to cross-reference and align consumption data with various factors such as product families, production lines, and raw materials, which are currently distributed across separate locations. With all this data, we can create and train algorithms to obtain consumption patterns and trends in order to compare them with the real ones and make decisions about them.
Infrastructure
On premise Cloud
Technology
Machine learning and deep learning
Data
Consumption through analysers for at least the last 5 months. Production data for more than 3 years; families, references, temperatures, power, line speed, etc.
Resources
Multidisciplinary team composed of data scientists and domain experts.
Difficulties and learning
KPIs (business impact and metrics of the model)
We use analysers at various stages within each production line, collecting consumption data stored in a database hosted in Power Cloud. This consumption can be monitored using Grafana. We need to cross-reference and align consumption data with various factors such as product families, production lines, and raw materials, which are currently distributed across separate locations. With all this data, we can create and train algorithms to obtain consumption patterns and trends in order to compare them with the real ones and make decisions about them.
Funding
Collaborators, Partners
Data Value Management BAIC
