CSP 2017 Online Program

Blending Big Data Visualization Tools with Statistical Analysis: Improving Automotive Lubricants (303408)

Keywords: big data, visualization, general linear model, compare, lubricants, automotive

Real world performance data is more meaningful than laboratory test results in developing automotive lubricants with improved thermal properties. To compare several experimental lubricants, four taxi-cabs in a major US city were equipped with several thermocouples to monitor ambient temperature, pinion bearing temperature, and lubricant temperature. Data was captured every 30 seconds for the duration of 25 weeks of testing, providing roughly 2 million observations to determine if differences in thermal performance of the axle can be attributed to variations in the axle lubricant. Ambient conditions vary throughout the day and across such a long period of testing. Driving patterns and vehicle down time also fluctuate, so a method must be developed to address both the long term (week-to-week) and short term (hour-to-hour) trends in order to make sound decisions about the lubricant. Big-data visualization tools are used to address the trends and summarize the massive data set and then general linear model techniques are used to compare the performance of the different lubricants.