This JMP Control Chart shows stable processes in the manufacture of Bloom Energy Servers.
Bloom Energy uses JMP® analytics to produce clean, affordable power
|Challenge||To refine processes, increase yield, and improve reliability and performance in the production of a more sustainable source of energy.|
|Solution||Bloom Energy uses JMP® every day for design of experiments, statistical process control, data analysis and product reliability enhancement.|
|Results||Over a short time, yields have increased significantly and product reliability has improved.|
Bloom Energy has an ambitious mission: to deliver affordable, reliable, clean energy for all. And it begins with sand.
The company, based in Sunnyvale, CA, has developed unique on-site power generation systems that employ an innovative fuel cell technology. Starting with a sand-like powder, rather than a precious metal such as platinum, the technology produces clean, reliable, affordable power practically anywhere. The systems can run on a variety of fuels, including biogas and natural gas.
Bloom’s Energy Servers, which are about the size of a parking spot and can power an average-sized office building, significantly reduce electricity costs and dramatically lower greenhouse gas emissions. Established in 2002, Bloom serves an impressive customer base that includes Google, AT&T and Walmart.
Making this technology marketable involves many processes, and Bloom’s founder, former NASA scientist K.R. Sridhar, has built a team of experts across a wide range of disciplines.
Among them is David Trindade, a Fellow at Bloom Energy and the company’s Chief Officer of Best Practices. Trindade’s job is to provide direction and leadership in the implementation of statistical process control (SPC) for manufacturing and testing. He works daily with Bloom engineers to refine processes, increase yield, and improve reliability and performance.
Enhancing the team effort that Trindade helps cultivate at Bloom Energy is the relationship he’s built with the developers of JMP statistical discovery software from SAS. Trindade has been a JMP user almost from the day the software was introduced more than two decades ago, and along the way he’s provided valuable insights into how JMP can be applied most effectively in the field.
Delving into the data
Prior to Trindade’s arrival, a few Bloom engineers had turned to JMP for data visualization and analysis, but there was no broad-based commitment to using the interactive software for design of experiments (DOE), SPC, reliability enhancement or modeling and predictive analytics.
Today, Bloom relies on JMP extensively to optimize processes throughout design and production. “Every percentage-point improvement we have in our processes is literally worth hundreds of thousands of dollars to us,” Trindade says. Recently, yields have increased significantly.
One of Trindade’s favorite tools is JMP Graph Builder, which lets him create interactive graphs on the fly. “I think it’s fantastic that I can drag different variables into a graph and have it instantly appear, giving me all those capabilities of looking at data in many different ways,” he says.
Process improvement begins with DOE, and JMP helps Bloom identify those factors that improve manufacturing, performance and efficiency. Says Trindade: “Another huge asset for us is JMP’s Custom Design platform for experiments that involve, in many cases, quite a few factors and responses.”
Next, Bloom monitors a variety of equipment involved in the manufacturing process using SPC capabilities in JMP. “They’re just excellent,” Trindade says. “SPC in JMP helps us achieve stable processes.
“We have many different suppliers,” he explains, “and we investigate the performance of each across our different systems using JMP’s control chart features, which allow us to see, simultaneously, a number of charts.”
The control charts are then reviewed on a weekly basis.
Trindade and his colleagues also regularly use the Fit Y by X and the Fit Model platforms in JMP to extract useful information from their data, then monitor and analyze the reliability of products in the field with the expanded reliability platform in JMP.
Just getting better
Among the relatively new features that JMP has introduced, Trindade likes the ability to model both repairable and non-repairable systems and degradation processes.
The Degradation platform in JMP allows users to analyze product deterioration data over time to help predict product quality and warranty risk. Users can define the best model (linear or nonlinear), create pseudo-failure times and then analyze with the Life Distribution and Fit Life by X platforms to predict and improve reliability, cut costs and prevent catastrophic failure.
“We use that a great deal,” Trindade says. “JMP allows us to choose a suitable reliability model for modeling individual components. It also allows us to monitor our systems out in the field using the recurrence analysis feature and to measure degradation processes and model those as they proceed.”
JMP can also analyze competing risks, which Trindade hails as “a great feature. When you have competing risks, you have different failure modes that can occur, and, basically, one acts as a censoring event to the other failure mode. JMP allows you to analyze that very easily.”
The Life Distribution platform “is just extremely powerful,” Trindade says. It automatically evaluates a large range of reliability distributions to find the best fit. Users can opt for nonparametric procedures or manually select and compare parametric distributions. JMP profilers allow users to interactively determine lifetime estimates and extrapolate to future time periods.
“I’ve seen dramatic improvements in the reliability capabilities, especially the incorporation of recurrence analysis and degradation analysis,” he says. “Defect modeling in JMP is a feature added after I mentioned that to the JMP developers. The DOE, the analysis capabilities, looking at distributions – all these features continue to improve.”
Still, what continue to impress him most are capabilities for DOE and the Graph Builder: “Those two things far outshine anything that I’ve ever seen anywhere else, and we use those extensively.”
A collaborative effort
Receptiveness to user input is among the attributes that Trindade most appreciates about the JMP organization. Over the years he has observed, and participated, as JMP developers have collaborated with users of the software to add features that help them speed discovery and work smarter.
“I really appreciate that openness to input, and the willingness to incorporate new ideas,” he says.
“And the technical support has been excellent. That’s another great feature. If I have any comments, I get a quick response from technical support, and they’re very thorough in investigating any questions I have. I think that’s a big plus.”
In short, JMP enables continual improvement at Bloom, from the earliest iterations of design through the modeling, manufacturing and testing processes. Trindade says, “The modeling and data analysis capabilities guide us in improving process yields, thereby reducing costs, which eventually translates into more affordable and clean energy for the world.”
The spirit that guides Bloom Energy is the belief in a better way – the same spirit that has guided the development of JMP.
Every percentage-point improvement we have in our processes is literally worth hundreds of thousands of dollars to us.
Chief Officer of Best Practices