Getting a grip with data analysis

Paul Boughton

A tyre and rubber company designed an experiment to solve challenges overcoming a production bottleneck and certifying new tires meet government standards. Eston Martz reports.

Bridgestone Corporation, the world’s largest tyre and rubber company, manufactures tyres in locations around the world for use in a wide variety of applications.

Among the company’s facilities is its Costa Rica plant, which produces 12,000 tyres every day for 23 markets in Central America and the Caribbean. But the plant faced two challenges: certifying a new tyre by meeting US Department of Transportation regulations, and overcoming a bottleneck that limited the overall production of light truck tires at the plant.

Six Sigma black belt Kenneth Quirós Acuña embarked on a project that, if successful, would address both challenges. He and his project team used Minitab Statistical Software to identify critical factors affecting both production levels and the certification tests, design an experiment to find optimal settings for those factors, and develop a model the company could use to produce and certify more tyres with more speed.

The challenge

When engineers at the plant looked at how capacity might be increased, they found that the chief bottleneck in production was a lack of tread. The plant had two types of tread-making machines: an extruder that makes tread in a single piece, and a ribbon-tread machine that makes tread by extruding many small ribbons.

The ribbon-tread machine was being underused. But efforts to get more use out of the machine faltered when tyres created with it failed to meet the standards needed for certification.

Plant managers believed it was too difficult to make qualified light truck tyres using the ribbon-tread machine, which involved multiple steps and complicated variables such as feed rate and increments width.

“It is not easy to reproduce a tread shape using ribbons because you have to find the correct combination of factors,” Quirós says. “The feed rate, drum velocity, extrusion velocity, and other factors all need to be in balance to produce a good tread.”

Quirós and his team realised that with so many steps and potential factors in the tyre-making process, he first needed to narrow his focus to create a manageable project. “When we looked at different combinations of factors to better understand why we had issues with the ribbon-tread machine, we realised that about 80 per cent of the problems involved a specific tire certification test.”

Tyres are built and cured according to exacting technical specifications. If a tyre passes preliminary building and curing tests, it proceeds to other tests required by the US Department of Transportation. Tyres that meet these specifications go on to the plunger test, in which a rounded plunger is forced into the centre of the tread of an inflated tyre. Testers measure the energy required to either penetrate the tire, or contact the surface of the tire rim. The Costa Rica plant’s ribbon-tread tyres were not meeting the guidelines.

“Some tests have a bit of flexibility,” Quirós explains, “but a tyre that can’t pass the plunger test can’t be certified, so that’s where we focused our efforts. We set out to create a model that yielded specifications for tires that passed the plunger test without hampering their performance on other tests.”

Solving the problem

First, the team members ranked variables in the tyre-making process according to their importance for certification. After an initial analysis, the team prioritised four factors for further investigation.

“One of the most important tests is the diameter, so we needed first to understand whether those four factors affected diameter,” Quirós explains.

By performing multiple regression analysis on these factors and tyre diameter data, the team discovered the most important factors were angle and weight.

“So in our model, those input values had a constraint, and we were able to generate an equation to set those input values appropriately.”

Now Quirós used Minitab’s Design of Experiment tools to quickly gather the data he needed to develop his model. “Because testing tyres is a destructive process, it’s very expensive to set up experimental runs,” he notes. “But DOE lets us study multiple variables simultaneously using the least number of runs, so we get enough data for reliable results without wasting time and resources.” He selected a two-level factorial design that would let him assess high and low settings for each of the key process input variables with only 16 runs.

Before collecting their data, the team performed a measurement systems analysis with Minitab to ensure they were collecting good data. They also assessed the baseline capability of the ribbon-tread process using Minitab’s Process Capability Sixpack. The results revealed that while the process was stable, the capability of the process to meet certification standards was poor. Now they were ready to produce their 16 experimental tyres, and put them through the certification tests.

When Quirós and his team analysed the experimental data in Minitab, the results showed that all four factors, and the interactions between them, had a significant effect on plunger test results. This was great information, but using it to derive a precise model would be difficult without a tool that could take the experimental results and let the team forecast how different variable settings would affect the tyre’s performance.

Fortunately, Minitab’s Response Optimizer tool does just that, making it easy for Quirós to fine-tune the process settings to produce the best results. “Using the optimiser helped us attain and surpass our goal for the plunger test,” he says. “We used it to create our predictive model and define our optimal input settings. Then we gathered more data at these settings and used Minitab to perform multiple regression analysis and validate the factor values.”


Based on the results of those analyses, the team redesigned how they used ribbon-tread to make light truck tyres. Before new tires were made, the plant managers would review the last specifications. Previous and proposed values for angle and drum width would be evaluated in a spreadsheet programmed with equations from DOE. If they were acceptable, data about the other two factors in the process would be added.

The new model was tried on a tire size that had proven particularly challenging in the past. The team used the model to determine effective factor settings, then made some tyres with the new process settings. To validate the results, they used the Minitab Assistant’s Capability Analysis tool. They also created a before-and-after control chart, which revealed how dramatically the new model improved the tyre’s performance in the plunger test.

“We achieved our results,” Quirós says. “Whenever we changed the level of a parameter, and ran the tests again, the average was very close to what the model predicted. We’re now exceeding the target for the plunger test.”

Use of the new model soon was expanded to other tyre sizes, too. The model was next applied to a different size of light-truck tyre, one that had not passed certification with ribbon tread before. This time, the process was qualified after the first trial. Since then, Quirós and his team have also successfully applied the model to tires used in agriculture.

The success of the project has increased production at the Costa Rica facility, saving hundreds of thousands of dollars and resulting in more than $2 million in opportunity costs.

Eston Martz is with Minitab Inc., State College, PA, USA.