Why diagnosis in the automotive field? The field of auto-motive engineering has seen an explosion in the presence of electronic components and systems on-board vehicles since the 1970s. This growth was initially motivated by the introduction of emissions regulations that led to the widespread application of electronic engine controls. In addition, the presence of microcontrollers on-board the vehicle led to a proliferation of functions implemented through electronic systems and related soft-ware, related to safety and customer convenience, creating the need for more sophisticated on-board diagnostics.
The original motivation for the introduction of real-time on-board diagnostics in automotive vehicles originates from the California Air Resources Board (CARB) requirements introduced in the early 1990’s to guarantee the integrity of the engine exhaust emissions control systems. The idea behind the original on-board diagnostics regula-tions [obd05], [EPA] was to guarantee that the exhaust emissions control system would be functional for a pe-riod of time associated with warranty or with regulated requirements. OBD regulations mandate that any fault in the emission control system aﬀecting software algorithms, sensors, actuators or other hardware that could lead to an increase of tailpipe emissions such that the vehicle would no longer meet the emissions regulations, should be detected in real-time and codified according to a set of on-board diagnostic codes that are described in the OBD legislation. These regulations first came into eﬀect in 1988 and were further expanded in 1994 through OBD-II regulations, and aﬀect every single component or subsys-tem that could increase engine exhaust emissions above a pre-specified threshold. With the growth in complexity in exhaust emissions regulations, and the attendant increase in complexity in the hardware and software required to meet such regulations, the task of meting OBD regulations has become quite challenging.在汽车上引入实时车载诊断的最初动机源自于1990年早期引入的加利福尼亚空气资源委员会（CARB）要求，以保证发动机排气排放控制系统的完整性。最初的车载诊断规则是为了保证废气排放控制系统在保证或调节要求的时间内起作用。OBD法规规定，排放控制系统中的任何故障，如软件算法、传感器、执行器或其他硬件，可能导致尾气排放的增加，使得车辆将不再满足排放法规，应实时检测。根据OBD法规中描述的一组车载诊断代码进行修改。这些法规于1988年首次进入E.E.ECT，并于1994年通过OBD-II法规进一步扩展，并对每一个部件或子系统进行了改进，使发动机废气排放量高于预先规定的阈值。随着排放法规的复杂性的增长，以及满足这些法规所需的硬件和软件的复杂性的增加，制定OBD法规的任务变得非常具有挑战性。
A second motivation for the introduction of on-board diagnostic algorithms has been the introduction of safety systems on-board vehicles. In recent years, increasing attention to safety has led to the introduction of anti-lock braking systems, traction control systems, electronic stability control systems, and passive and active restraints. Many safety functions are also the subject of increasingly stringent regulations. The introduction of active systems that can aﬀect the safety of a vehicle, such as braking, traction and stability control, and the introduction of by-wire systems to implement these functions, has generated deferent needs in diagnostics. In this context, diagnosis is a precursor to fault-tolerant control: if a safety-critical component is malfunctioning because of a fault or failure in a sensor, actuator or other component, or a malfunction in one of the software algorithms, then it is necessary to identify such safety-critical failures very quickly so as to be able to take corrective actions and ensure the safety and reliability of the vehicle.
The third area that has seen a growth in diagnostcs is related to customer satisfaction. There may be some significant advantages in having diagnostic algorithms on-board the vehicle for the purpose of guaranteeing customer satisfaction and overall quality.
PROBLEMS AND CHALLENGES
In the face of the deferent requirements outlined in the preceding section, there is growing interest on the part of the automotive industry in the ability to systematically design diagnostic algorithms. Further, automakers have also shown a desire to extend warranty periods to provide consumers with a worry-free experience. As a consequence, in addition to on-board diagnosis of deferent functions, the prognosis of various functions and subsystems in the vehicle has also become important. Manufacturers would like to be able to predict when maintenance or replacement may be needed for specific components, for example the 12V battery, or components in subsystems related to the emissions control system. So, prognosis is beginning to take on a role in automotive electronic systems that was not on the horizon even just five or ten years ago.
The implementation of diagnostic and prognostic algo-rithms of this type in automotive systems presents a num-ber of challenges due to the scale of the implementation. Such algorithms must be adaptable to millions of vehicles and must be robust enough to be valid over a broad range of diﬀerent realizations of the same vehicle platform, with choice of diﬀerent engines, transmissions, and accessories. Further, vehicles that might be architecturally identical, will unavoidably require diﬀerent software calibrations in diﬀerent markets. Thus, the design and implementation of OBD algorithms is not a ‘one size fits all’ kind of design approach.
The second issue is related to the fact that automo-tive systems tend to be complex and highly nonlinear. For example, engine and exhaust emissions processes are characterized by complex thermochemical behavior (com-bustion processes, exhaust emissions formation), that is strongly aﬀected by chemical reaction kinetics, fluid mo-tion and heat transfer. Further, the presence of sensors and actuators, such as fuel injectors, or systems that could be pneumatically or hydraulically actuated increases the overall complexity of an engine emission control system. Therefore, it is diﬃcult to imagine that simple, linear algo-rithms could be very eﬀective unless a substantial amount of thinking and a deep understanding of the physics of the processes goes into their design.
Another important aspect is the speed of execution, in the face of limited computational capabilities (both CPU speed and memory). On-board computers used in automotive applications have relatively low power relative to the number of functions that they perform, because cost is a significant constraint in the automotive industry. So, one of the main challenges is to develop eﬀective diagnosis algorithms that can be implemented in fixed-point arithmetic microcontrollers with limited amount of memory and limited CPU speed. Some algorithms may require truly real-time implementation. For instance, in safety-critical diagnosis algorithms (e.g.: vehicle stability control, or brake-by wire or steer-by wire applications), one is obviously concerned with the implementation of these algorithms in real-time so that any fault that is detected can be compensated for in a fault-tolerant control scheme or by entering a limp-home mode as safely and as quickly as possible. On the other hand, other types of algorithms, such as those that may be used to diagnose malfunctions in the emission control systems, may not have such stringent real-time requirements, in the sense that on-board diagnostics regulations typically require that the diagnosis be carried out within what is called a ‘one trip’
Finally, diagnosis must be as transparent as possible to the user, while the designer must be very cognizant of the relative weight of false alarms vs. missed detections: such weighting will vary depending on the application, with missed detections being especially costly (to the user) in safety-critical applications, while false alarms can be very costly (to the manufacturer and consumer alike) when non-safety-critical applications that have warranty implications are considered.
In short, the subject of system diagnosis in the most complex consumer device in existence today - the automobile is one that presents numerous technical challenges that range from the theory of estimation and detection, to real-time software implementation issues.