Aluminum Maker Crunches Big Data for Top Performance
Increases production, reduces energy consumption, improves productivity
"We wanted to define productivity for every cell, or batch, of aluminum solvent for every day of operation," says a process control engineer. "We thought that if we could look at all the variables that correlate with high efficiency, we could use the analysis to predict and ultimately control daily, rather than monthly, outcomes."
The engineer estimated that on any given day, 15 to 50 variables could impact productivity. These might include cell voltage, stability, cell structure, and so on. "Previously to achieve sufficiently precise measurements, we would have to collect data for weeks, and of course then we couldn't respond in real time. We wanted to analyze the many factors that directly or indirectly affect the quantity of aluminum produced."
The company implemented a predictive analytics solution based on TIBCO® Data Science and used multivariate statistical process control (SPC) to help monitor key performance indicators and improve control of the daily production process. The company now uses advanced statistical modeling to identify and predict the interplay of variables that affect the process.
Reduced Analysis Time
The manufacturer's process team saw immediate and positive results. "With the TIBCO Data Science solution, we have significantly strengthened our understanding of the production process," says the team's process control engineer. "With the expanded knowledge of production dynamics that TIBCO Data Science provides, we increased total aluminum production while reducing the time spent analyzing key performance indicators."
Before TIBCO Data Science, the engineering team spent hundreds of hours analyzing data before making decisions about what to adjust in the production process. The engineer estimates that the team has cut data analysis time by about 25 percent, thus increasing employee productivity and making better use of resources. With TIBCO Data Science, the company has created best practices and automated standard processes, saving significant time for the engineers and scientists who monitor processes, and the R&D group who develops and enhances new ones.
Using TIBCO Data Science, the manufacturer can now readily detect new variables affecting the smelting process. "Suddenly, we can see that a certain parameter we've never looked at before seems to have a pretty important effect on the results of a particular production batch," says the process control engineer.
Thorough analysis enables the team to slowly change parameters on poorly performing production cells to bring them in line with those that are doing well. Precision analytics reduces the risk of these changes and results in higher yields.
The solution's automated neural networks use historical data to represent a massive range of operating scenarios. Process technicians and engineers can now easily see the effects of adjusting various inputs (temperature, cell voltage, cell stability) and identify which ones have the largest influence on manufacturing outcomes. The solution also helps users make predictions based on planned events as well as on parameters that are out of the process team's control—such as a change in the impurity concentration of raw materials.
"We know this kind of event can generate a loss in the total quantity of metal produced, but with the analytical models we've developed, using the normal behavior of our process, we can understand how to make changes in other areas of the process to minimize the effects of unexpected events,” says the engineer.
TIBCO Data Science allows the team to quickly crunch data and find unexpected variables for a specific effect. With automation and sophisticated analytics, they've achieved greater precision in less time and found new and innovative ways to approach problems and improve processes. Because they can quickly detect and resolve emerging issues before they become significant, it saves the company millions of dollars each year.
Besides striving to increase the quantity of aluminum that's produced, the manufacturer is constantly analyzing energy consumption. TIBCO Data Science helps the company keep track of energy and how to best adjust processes to save it. "If we can produce aluminum using 5 percent less energy, it's valuable for the smelter and better for the environment," says the engineer. Without the analytics provided by TIBCO Data Science, it would be difficult to make changes that address the company's sustainability concerns. "We are very proud to be one of the most efficient primary metal producers in the world. To keep our place in the industry, and maintain our commitment to sustainability, we have to continuously work to improve our understanding of even the smallest details that could affect our production processes and reduce energy consumption. For that, we need some very sophisticated analytics tools.
"TIBCO Data Science has certainly given us a competitive advantage within the R&D organization. It definitely helps with our overall production process, too, but in terms of research, it's a very good advantage."