You are here

System Diagnostics for Smarter, Safer Vehicles

When the “check engine” light illuminates, drivers take their cars to a repair shop and a repair technician fixes the problem. But how did the vehicle recognize a fault? Through system diagnostics.

“Diagnosis is more or less about recognizing differences between expected behavior and anomalies,” explains Dr. Giorgio Rizzoni, Director of The Ohio State University Center for Automotive Research (CAR). “You would like to be able to detect anomalies in the behavior of a system before they result in a problem.”

Rizzoni has been conducting research in system diagnostics since the late 1980s, when the topic was first recognized as an important component of engine and emissions control systems. “The government announced that they were going to require on board diagnostics (OBD) systems for every vehicle that was going to go into production. In 1994, the second generation, OBD-II, went into effect. So, why are on board diagnostics important in that 1980s-1990s generation of vehicles?”

In 1994, California implemented a set of regulations broadly known as OBD-II, which were adopted nationwide by 1996. These regulations required automakers to implement sophisticated onboard diagnostics to verify that all components related to the exhaust emission control system were continuously diagnosed.

Today, system diagnosis is an integral element of all of a vehicle’s electronic control systems. There are two main classes of diagnostic systems: those that are government-regulated (mostly related to environmental protection) and those that are linked to vehicle safety.  Reliable diagnostic systems also provide quick identification of malfunctions and therefore faster and more reliable service and maintenance procedures.

As electronic control systems become more common through the use of hybrid-electric powertrains and connected and automated vehicles (CAV), the need for accurate diagnosis is steadily increasing. Diagnosing systems has traditionally used mathematics- and physics-based models, which predict behavior and anticipate malfunctions. But using these methods to diagnose increasingly complex systems leads to extremely complicated models. “What if I had 20 sensors in my car that measure what I need to know to automate the driving of my vehicle? Can I really create a model that links and ties all of these things in such a way that I can use it for real-time diagnosis?” asks Rizzoni. “Yes, I probably can. But then I need the Ohio Supercomputer Center to run the model. I can’t do it onboard the vehicle.”

CAR-affiliated researchers are tackling system diagnosis through a combination of physics-based models, machine learning and artificial intelligence methods, and cloud computing to comply with ever-changing technologies and regulations. Their work produces safer, more cost-effective vehicles on the cutting edge of modern transportation.

Diagnosing Leaks and Reducing Emissions

Dr. Ruochen Yang, who recently earned a PhD in Electrical and Computer Engineering, is working to improve evaporative emissions control systems (EVAP) by improving the detection of small leaks.Dr. Ruochen Yang, who recently earned a PhD in Electrical and Computer Engineering, is working to improve evaporative emissions control systems (EVAP) by improving the detection of small leaks.Automotive system diagnostics has expanded beyond its roots in exhaust emissions systems, but that original system is still challenging researchers. Dr. Ruochen Yang, who recently earned a PhD in Electrical and Computer Engineering, is working to improve evaporative emissions control systems (EVAP) by improving the detection of small leaks.

“The regulations of the Environmental Protection Agency (EPA) and the California Air Resource Board (CARB) require detections of any hardware or component that may malfunction and affect the emissions performance in the system,” says Yang. “One of the most challenging faults to be detected and that is required by the regulations is small leaks.”

How small? Leaks as small as 0.020 inches in diameter must be detected and repaired. “Auto companies tend to be very conservative about detecting them,” Yang remarks. “So sometimes, a good system without leaks may be considered as a small leak.”

Those false alarms are often unable to be reproduced by a mechanic, leaving customers dissatisfied. Yang is attempting to reduce false alarms while still detecting true leaks, so that customers are satisfied and vehicles comply with ever-stricter regulations.

“Diagnosis will be an area that requires a lot of advancement in the industry to comply with the standards while remaining cost-effective,” Yang predicts. But as the transportation industry moves toward electric vehicles, gasoline emissions may become a smaller concern.

Building Safer Energy Storage Systems

Senior Design Engineer Prashanth Ramesh knows the pain of trying to drive a car with an unexpectedly drained battery. A research project to predict when batteries would fail to start a car allowed him to improve user experience and solve a day-to-day problem. Solving tangible issues is what Ramesh truly enjoys about system diagnostics.

As the transportation industry moves toward electric vehicles, Prashanth Ramesh is racing to provide reliable diagnostics for their essential energy storage systems.As the transportation industry moves toward electric vehicles, Prashanth Ramesh is racing to provide reliable diagnostics for their essential energy storage systems.As the transportation industry moves toward electric vehicles, Ramesh is racing to provide reliable diagnostics for their essential energy storage systems. “Companies are constantly adding more and more batteries to the different applications, like vehicles,” he says. As the usage of batteries grows, so does the need to understand when things could go wrong in order to set warranties and ensure the safety of passengers.

Diagnosing potential faults in batteries is tricky, because batteries are surprisingly delicate and faults can be dramatic. “A battery is very sensitive to temperature and how you use it. So, they need to be able to make sure that it works really [well] in a hot place like Arizona or Florida, but also works pretty well in a really cold place where it's snowing. Temperature has very adverse effects on batteries. So, if it gets too hot things could lead to where the battery explodes, catches fire. But,” Ramesh says, “it could also just fail to function.”

Batteries are not only environmentally sensitive; they’re also meant to last for years. So, how can researchers test them in an efficient way? Ramesh and his team begin in the lab, performing experiments in carefully controlled environments. Based on that data, a model can be created to predict a battery’s behavior. “Once we generate a model, we can run simulations and try to understand if the model can predict failures,” says Ramesh. Using the models allows researchers to predict 10 or more years of usage in a few weeks.

But battery technology is evolving even as Ramesh diagnoses existing systems. New batteries are being produced and existing technologies are being integrated at a deeper level. “It's not just being looked at as trying to solve problems on a battery but more from trying to make sure the system works,” Ramesh says. Soon, he predicts, every system on a car will be powered by energy storage systems. "It's a more complex system than it's ever been... becoming more complex.”

Modeling Faults in Steering Systems and Tire Degradation

While earning his PhD in Mechanical and Aerospace engineering, Tianpei Li researched system diagnosis of electrified power train and vehicle chassis systems. Li uses the classic method of physics-based models to improve the diagnostics of new vehicles, especially electrified and automated vehicles. “People have used [models] for a long time, but we apply different methodology and different objectives,” Li says. “Overall, it still works well.”

Li currently works on a project modeling vehicle steering, suspensions and tires. Using experimental data, he creates models to predict the behavior and collaborates with the Driving Dynamics Lab at CAR to improve them using their driver-in-the-loop simulator. “Once we design the diagnostic strategy and algorithms, we can do simulation in the loop using that simulator by providing real human driver input. So, we can simulate more realistic driving scenarios to validate all the diagnostics design,” Li says.

But no model is perfect. “Models always have uncertainty and inaccuracy,” Li notes. “There's always modeling error. Making sure the modeling error is within control when you actually use it to apply diagnostic designs... that’s a big challenge.” The other main challenge in Li’s work is the reliance on onboard vehicle sensors to perform system diagnostics in the real world. Because those sensors are limited in their ability and external noise affects their performance, designing the diagnostic system becomes more complicated.

Despite the challenges, Li is improving the ability to identify and predict system degradation and failure, which in turn improves the safety of the vehicle and saves on maintenance costs, making consumers and auto makers more satisfied with their vehicles.

Facing the Future of Safety and Security

Research Associate Professor Qadeer Ahmed leads the Cybersecurity@CAR lab, where a diverse team of researchers diagnose new vehicle systems in an effort to keep them safe and secure. Since OBD II was mandated, diagnosis has been happening within the vehicle. With the addition of Advanced Driver Assistance Systems (ADAS) like lane keep assist and adaptive cruise control and in-vehicle connectivity via WiFi and cellular connection, onboard diagnostics systems must become smart enough to diagnose these new systems.

These new features increase a vehicle’s connectivity to the world around it, but they also make vehicles more susceptible to outside influence. “Let’s say [a hacker] can spoof your GPS signal. Your lane keep assist is working on your GPS signal; that signal is being changed or influenced. This will affect your lane keep assist. Now, how do we identify that and make sure that lane keep assist module is still behaving the way it should?” asks Ahmed. “These are the types of problems we are looking at in cyber security.”

Those problems span multiple disciplines and require varied viewpoints, so Ahmed is building as diverse a team as possible. “If you don't have the diverse team, we may be addressing part of the solution, which may not fit into the bigger picture,” he says. As new technology evolves, the challenge grows. Vehicles may connect to other vehicles and transportation infrastructure, making the possibilities for system faults nearly limitless.

But there are advantages to technology that continues to change. System diagnostics researchers are able to understand how these technologies are evolving and even influence them, so that vehicles remain as safe and secure as possible.

Written by Georgia Drost, CAR Writing Intern