Ford has performed over 20,000 crash tests since 1954. This number has increased substantially in recent years as the company launched more products and performed additional tests required by safety regulations. These factors greatly increase the number of crash tests required each year and the complexity of planning and scheduling them. Prototypes built for product development testing can cost in excess of $200,000 each, because many of the parts and the full-vehicle prototypes are handmade and highly customized. Commonly, a vehicle program corresponding to development activities related to a product launch requires over 50 vehicles for crash testing; therefore, maximum utilization is critical for balancing engineering resources and program timing.
Maximizing utilization is not a simple matter. Crashes are destructive, and only certain tests can be combined on the same prototype vehicle. Program milestones and staggered prototype vehicle delivery create a difficult timing problem. The different configurations of vehicles offered (e.g., types of power trains, trim, body style) are an additional source of complexity, because a comprehensive testing plan must match crash tests to specific configurations.
Until recently, Ford manually developed crash-test plans using pen and paper and Excel spreadsheets. This process was tedious, and constructing a test plan could take several days or weeks. The schedule produced was highly dependent on the knowledge and experience of the individual planning engineer. Determining if the crash schedule was optimal in terms of the number of vehicles needed was nearly impossible. Reacting to delays and program changes required extensive rework of the test plan.
Competing for the Wagner Prizes places you and your author team among the upper echelons of those within the O.R. and Analytics fields today.
Researchers developed a custom-made crash-test scheduling system that transformed a labor-intensive process, which relies on high levels of expertise, to one that automates time-consuming scheduling analyses through mathematical optimization, while also institutionalizing expert knowledge. Instead of a more traditional assignment-based modeling approach, they developed an integer programming model using composite variables representing sequences of tests to be performed on a single prototype vehicle. To the best of their knowledge, we are the first to apply this modeling technique to problems that combine the features of both bin packing and complex sequential job scheduling. The researchers also describe a column-generation algorithm for solving our formulation, with a pricing problem structured specifically for crash-test scheduling.
The system produces optimized schedules in seconds or minutes (depending on program complexity and size). Engineers can use the time saved to run what-if scenarios including double-shifting prep time, working weekends and (or) holidays, and introducing flexibility for vehicle specifications. This gives planning engineers and program management personnel concrete data on resource requirements and potential areas of flexibility.
A key to delivering the technology and maximizing its benefits was designing the system for ease of use with engineers’ current working-document formats. The researchers developed interfaces that ensure data quality and completeness. They designed an optimization package to run on Ford’s High Performance Computing Center servers, while integrating it into a user-friendly web application with a back-end database.
The success of the Test Planning Scheduler Support System (TP3S) application is evidenced by its rapid technology adoption within the safety organization of Ford’s product development group. After a series of pilots and limited releases, the application was rolled out for use on all future programs. As of summer 2015, TP3S has been used to develop crash-test plans for the following vehicle programs: Ford SuperDuty, C-Max, Fiesta, EcoSport, Mustang, Fusion, and Edge; and Lincoln MKC and MKX. In addition to a host of benefits, ranging from improvements in planning processes and information infrastructure to valuable time saved by engineers, Ford has conservatively estimated TP3S will yield ongoing annual direct net savings of $1,000,000.