Testing and Tuning Engine Controllers

In all areas of engineering today, market pressures and regulatory issues dictate design criteria. In the automotive industry, designers of electronic engine controllers face conflicting demands: maximizing performance, ensuring smooth driveability and minimizing emissions—all while maximizing fuel efficiency.

On-board diagnostics add layers of self-monitoring to the task list. And most controllers interact with other subsystems such as transmission control, anti-lock braking, traction control, climate control, power steering and anti-theft systems. At least one major manufacturer is now using 32-bit CPUs to cope with these increasing demands.

Despite increasingly sophisticated computer modeling, there is a constant need for empirical measurement to verify, calibrate and troubleshoot the models. But the role of instrumentation has evolved from a primary means of development to a feedback tool for refinement of the models.

Rather than scanning meters of strip chart, performance data is digitized and numerically compared to prediction. As computers have evolved, so too have data acquisition systems.

Controller Development

During development of an engine management system, the complete engine is prototyped and subjected to a variety of load conditions on a dynamometer. A mapping process, usually automated, sweeps over a range of variables while recording slowly changing parameters such as throttle position, air, coolant and exhaust temperatures, manifold air pressure, fuel and air flows, rpm, torque and exhaust gas recirculation.

After collection, workstations compare the observed behavior to expectations. Years of development and analysis have fine-tuned this process to a high degree of precision. As a result, today’s vehicles perform better proportionally than the muscle

cars of the 1960s, with twice the fuel economy and a hundred times fewer emissions.

A more detailed study of performance requires higher-speed instrumentation to investigate dynamic behavior. For example, a lean fuel mixture improves economy and CO emissions but increases NOx emissions. Up to 95% of the total pollutants in the EPA FTP-75 emissions test are generated in the first three minutes after a cold start; but until the thermally reactive oxygen sensor is warmed up by the exhaust gases and provides accurate readings, the ECU is running open loop.

Designers must characterize these intricate relations and balance the competing needs. Since vehicles are already so highly tuned, an incremental miles-per-gallon is ever harder to obtain.

Internal Combustion Analysis

Although the American auto industry is celebrating its 100th anniversary this year, there is still much to be learned about the internal combustion process. Propagation of the flame front is so acutely sensitive to tiny variations that mathematical chaos theory is being applied to identify causes of seemingly chaotic behavior.

Because simulations cannot account for the last few percent of variation, the final tuning relies on actual measurements. Along with the previous quasi-static variables, typical internal combustion analysis (ICA) measurements include cylinder pressure, fuel injector pulse width and ignition timing in each cylinder.

Digitizing rates of 10 to 100 kS/s per channel are needed to adequately describe the millisecond pressure pulses (Figure 1). The measurements are made in degrees of crankshaft angle, with 360 to 3,600 samples per revolution. If the pressure is plotted vs cylinder volume, the area of the P-V curve is proportional to the work produced (Figure 2).

Engine compartment packaging constraints mean it is unlikely that all cylinders perform identically. Due to small differences in airflow of intake and exhaust passages, the pressure from cylinder to cylinder may vary by several percent. The control software can be tuned to partially compensate.

Another key measurement is consistency. A single misfire can fail an emissions test. Hundreds or thousands of combustion cycles are recorded and peak pressures statistically analyzed. Because of the sensitivity to tiny variations, it is difficult to reduce the coefficient of variance below 1%.

Types of Instrumentation

Because of their statistical nature, ICA tests generate huge amounts of data. Traditional methods have been limited to two techniques:

Specialized ICA equipment measures performance in real time over a large number of combustion cycles. Abnormalities are detected in real time and alert the operator to end the test. While effective, the high cost of such dedicated instruments has limited them to the major automakers.

High-speed tape recorders capture raw data during the test, then it is transferred to a workstation for data reduction. This method minimizes cost, but is time-consuming and lacks real-time feedback.

Some of the newest data acquisition systems incorporate current PC and digital signal processing (DSP) technologies to analyze high-speed events in real time. While DSP is often considered a frequency-domain tool, there are many useful time-domain applications as well.

Detecting cyclic peaks, computing mean values and comparing to setpoints are within the capabilities of inexpensive DSPs (Figure 3). In combustion analysis, a plot of thousands of peaks can be displayed in real time with an alarm on any anomalies.

For investigation, a trigger can capture unusual cycles in detail, whether too high a peak due to knocking or too low due to misfire.

Massive hard disks driven by the PC market offer the storage and bandwidth to record high-speed engine data for hours, and allow easy data transport by network or backup media. Like PCs themselves, new instrumentation using PC technology offers increasing capability at decreasing cost.

Application Examples

Applications where an instrument of this capability are proving attractive include:

A high-performance engine builder tests a new V-8 design. The ECU must be reprogrammed from previous settings to accommodate greater air and fuel flow. Due to the prohibitive cost of dynamic ICA instrumentation, oscilloscopes monitor incipient knock and cylinder-to-cylinder variation during the mapping process.

For numerical analysis, manual measurements are made from the traces and transcribed to a spreadsheet. The Odyssey, a data acquisition system from Nicolet, provides the real-time display of a multichannel oscilloscope while capturing statistics of each cycle. A report is sent immediately to analysts via Ethernet without the delay of manual post processing.

A heavy engine manufacturer performs exhaustive testing on each electronic fuel- injection system. A battery of tests currently consumes a full day since more than 70 waveforms are digitized on a transient recorder, then imported to a large spreadsheet for comparison to standard values. A test technician manually enters UUT information, rpm, fuel flow and temperature. The dataset is so large the spreadsheet requires several hours to produce the test report.

The data acquisition system makes calculations in real time to improve statistical validity and places them automatically into the report, eliminating tedious file translation.

Multitasking allows the operator to examine and print test results while the next recording is underway. Due to doubled test throughput, the payback period is much less than one year.

The examples illustrate several advantages of the new data acquisition systems:

Real-time measurements and data reduction via DSP.

Easy data transport to other software and other computers.

Increasing power and decreasing cost with mass-market PC technologies.

As these instruments become more common and even more powerful, design engineers can directly compare simulation results with dynamic test data without the need for painful translations. Considering the tremendous time savings, instrumentation engineers may disagree with Thomas Watson’s 1943 statement that “I think there is a world market for maybe five computers.”

About the Author

Gary Schneider is the Senior Product Manager at Nicolet Instrument Technologies. He has been with Nicolet for seven years, with previous positions in training and manufacturing. Mr. Schneider studied electrical engineering at the University of Wisconsin and holds a B.S. degree in marketing from the University of Upper Iowa. He is a member of the IEEE, SAE and the Midwestern Council of Sports Car Clubs. Nicolet Instrument Technologies, 5225-4 Verona Rd., Madison, WI 53711-4495, (608) 276-6172.

Gary Schneider is the Senior Product Manager at Nicolet Instrument Technologies. He has been with Nicolet for seven years, with previous positions in training and manufacturing. Mr. Schneider studied electrical engineering at the University of Wisconsin and holds a B.S. degree in marketing from the University of Upper Iowa. He is a member of the IEEE, SAE and the Midwestern Council of Sports Car Clubs. Nicolet Instrument Technologies, 5225-4 Verona Rd., Madison, WI 53711-4495, (608) 276-6172.

Copyright 1996 Nelson Publishing Inc.

October 1996


Sponsored Recommendations

Comments

To join the conversation, and become an exclusive member of Electronic Design, create an account today!