September 21, 2000 U.S. Environmental Protection Agency Office of Transportation and Air Quality Engine Control in Transient Operations on a Dynamometer Test Bench (This note will discuss how differing strategies have been used to yield satisfactory performance for some problematic engines.) In basic terms, transient operation is defined as a series of different engine speed and torque pairs that represent a recognized "cycle." Running transient operations on a typical on-highway heavy-duty diesel engine for purposes of generating emissions information can involve applying various laboratory-specific dynamometer and engine control strategies. A typical approach includes measuring torque between an engine and dynamometer using a suitable in-line torquemeter. In such cases, the test cell computer provides signals that are used to control dynamometer speed and throttle position. A servomotor is usually used for varying throttle or "pedal" position, which translates into engine fueling changes that target achieving a given in-cycle torque command. Success in driving transient cycles depends largely on how well the engine responds to changes in throttle position. Attempts to improve or tune a throttle control algorithm emphasize minimizing overshoot, undershoot, and settling time of the actual versus commanded engine speed and torque values. Subtle variations in software and hardware control strategies can be implemented to improve transient cycle performance. For example, programming slightly different linear slope and offsets alter speed and torque command signals, and command signal gain changes can be made to vary the response of servomotor and/or dynamometer operational amplifier circuits. However, when reasonable efforts fail to yield satisfactory transient performance, it may become necessary to seek alternative control solutions. In recent years, this has often been the case for running some engines fueled on compressed natural gas (CNG), for which unusually poor transient cycle regression (actual versus command) statistics can result despite significant efforts by laboratory staff and engineers to tune typical controls. Testing engine exhaust emissions on many different engines over various transient cycles requires either successfully tuning the existing control software and hardware, or working to develop a new control strategy for each unique engine or engine system combination. This iterative process can be further confounded in emissions research activities. For example, an engine manufacturer working to develop an engine control module for a new spark-ignited CNG-fueled engine may struggle to improve emissions by modifying various spark and fuel control parameters between emission tests. An emissions measurement laboratory in such cases would face a moving target with regard to engine performance, and hence a successful transient control strategy would need to anticipate or actively adapt to that evolving engine to meet cycle performance criteria. Another example where transient cycle performance has presented problems is with some nonroad engines. Trying to run selected transient cycle operations on some nonroad engines has required implementing modified versions of those same adaptive dynamometer and throttle control strategies. Of course, to date, new nonroad engines must comply only with emissions regulatory limits based on steady-state engine performance. In general, a successful transient control strategy may utilize a combination of throttle position mapping for developing a "feed forward" expression, and some form of active or adaptive control. Implementation requires defining parameters that represent proportional, integral, and derivative time constants, referred to as PID control algorithm constants. Combining a feed-forward expression and PID algorithm has often resulted in stable engine and dynamometer control, with overall improved cycle validation and regression statistics. The throttle position mapping exercise simply establishes servomotor command voltages for a discrete number of engine speed and torque pairs. This voltage information is then used to develop a closed form mathematical expression. This surface exists in three-dimensional space, with the XY-plane representing engine speed and torque values as ordinate and abscissa. In application, the computed Z-axis value is the throttle position or servomotor command voltage expected to yield the targeted operating point. A fine grid will result in a more representative surface, but experience has shown that 10 to 15 percent increments are usually sufficient for this process. In actual usage, the applicable servomotor command voltage is calculated and output for each point in the cycle. Actual engine speed and torque information is measured and used by the PID control to "trim" the servomotor position to actively meet transient performance targets. In addition to trim, the successful control strategy computes and reacts to rates of change in the measured speed and torque values and either adds or subtracts a large or small voltage accordingly. Such combined strategies can be further tuned to minimize overshoot and shorten settling times. In cases of developing a new engine, the throttle mapping exercise will need to be repeated if significant changes are made by the manufacturer. However, for an existing nonroad engine, for example, one throttle position mapping exercise is usually sufficient. Depending on the chosen speed-torque grid density and, possibly, the stability of the engine's steady-state performance, the process typically requires two hours to complete.