Aston Martin DB9 is High-Tech Under the Hood
By: Mike Thomas | Ford Communications Network
Components within the Aston Martin DB9 are on the leading edge of computer technology. For more information on the DB9, visit astonmartin.com.
DEARBORN, Mich.-- In 1964, James Bond's Aston Martin DB5 scored a worldwide hit with its high-tech wizardry. Today, Aston Martin's DB9 contains a high-tech advance under its hood that, while not quite as flashy as dual-mounted machine guns and ejector seats, is on the leading edge of computer technology and a first in the industry.
The 2005 DB9 contains the first onboard neural network in an engine control module. Unlike traditional computer systems that need to be programmed for each step, neural networks are programs modeled on the way human brains learn and adapt. The DB9's module keeps tabs on engine combustion performance with a sophisticated software program that compares actual engine performance to the design specifications.
"The DB9 was quite a challenge because it has a V-12 engine," said Craig Stephens, manager of Research & Advanced Powertrain Controls. "On a V-12, the frequency of firing events is so high that the legislative requirements for misfire detection could not be met with conventional computing resources. Neural networks offered us a whole new paradigm for computing and the potential for a misfire detection system that would be fully capable of meeting every detail of the regulations, something that the whole industry struggles with on any engine with eight or more cylinders."
The key to neural networks is pattern recognition. Much of the information streaming into a system is random, a kind of "white noise" of data. Important features, however, occur and reoccur in significant patterns. To take a simple example, the human brain processes a large amount of sounds every day, yet most never reach conscious awareness. For example, the hum of the ventilation system disappears into general white noise.
The ventilation system, however, does have a definite sound (pattern of sound waves.) If the pattern is broken, say by a rattling loose part, the sound reaches consciousness like an alarm bell.
In the same way, neural networks in a vehicle are trained to detect specific patterns of data in a vehicle's operation.
"There's a pattern in the turning of a crankshaft," Stephens explained. "There's acceleration and deceleration in the turning of the shaft timed to combustion in the engine. A misfire is a disruption in the pattern. The misfires themselves may be isolated events, or they may form a pattern. That pattern is the signal we're looking for in all the noise."
To discover the patterns, Stephen's team drives a vehicle under every conceivable condition. They then force the vehicle into every conceivable type of misfire. The data collected during the test drives is returned to the lab and fed into a training program. By examining the patterns, the neural network is trained in a simulation to distinguish the normal patterns from the misfires.
"We intentionally hold back about 30 percent of the data," Stephens said. "Once the neural network is trained, we use the remaining 30 percent to test the network's capability."
Assuming the neural network passes its test (a kind of midterm for the program), the team then tests the network with data it's never seen before. Data is again collected on driving conditions and misfires, this time with a different vehicle, and used for the network's final exam.
"We test for two types of errors in the program," Stephens said. "Alpha errors are the most serious. They occur when the neural network detects misfires when there are none. Alpha errors can cause the Check Engine light to switch on, leading a customer to seek service for non-existent problems. Beta errors are the opposite. The neural network fails to detect a misfire."
The designers build a critical threshold number of misfires into the system. Once the neural network detects the critical number, the Check Engine light is triggered and the misfiring cylinder is shut down to avoid damage such as a melted catalytic converter.
"We now have the most complex vehicle application of a neural network in the world," Stephens said. "The system on the DB9's V-12 fully meets all the worldwide misfire detection regulations and allows us to protect the catalyst and engine. Initially, the network is contained on a separate chip (costing about $5), but someday we hope to include it in the main power control module."
Future applications of neural networks could include any area with predictable patterns, or as Stephens puts it, "Anywhere there is a signal buried in noise." Examples could include measuring engine torque or in maintenance diagnostics. Experience gained from the work done on the DB9's engine is allowing the same technology to be applied to mainstream Ford V-8 and V-10 programs.
Eventually, a neural network could theoretically learn the driving habits of a particular driver and then adjust the vehicle's operation or maintenance schedule. Outside of the vehicle itself, neural networks could be trained to recognize patterns of traffic flow (Monday rush hour as opposed to Sunday morning) and then time traffic signals accordingly.
For the time being, however, 007's fantasy Aston Martin has been trumped by a real-life breakthrough in the form of a $5 computer chip.