Automatic calibration of internal combustion engine parameters 3-axis
Automatic calibration of internal combustion engine parameters 3-axis
Cars manufacturers interest in both reducing costs and development time of their products is growing. A possible improvement can be given by applying optimization techniques to the automatic calibration of ECUs.
The main purpose of the collaboration between Alma Automotive and ESTECO is the implementation of a demonstrator able to highlight the added-value given by coupling National Instruments devices with modeFRONTIER framework, while exploiting specific applications developed by Alma Automotive for the engine set-up. The requirement of a short calibration time, a goal for the optimal management of an engine under test, is in opposition with the need for a robust and accurate optimization process, which is normally slow. Therefore the main challenge is to find the optimum as fast as possible, while maximizing the accuracy of the solution.
The test environment used to implement the proposed automatic calibration system can be described as follows:
- the engine and test bench controllers (connected to the engine and to bench transducers/actuators respectively) are implemented by using CompactRIO devices, properly programmed and equipped with FPGA boards;
- signals are sent from the CompactRIO, through an Ethernet connection, to two host PCs (equipped with Windows 7 operating system) which acquire, store the data and set up the operating parameters. The interfaces developed by Alma Automotive provide a simple way to interact with the system;
- modeFRONTIER is installed on the host PC connected to the engine controller.
In a typical working condition the engine runs on a steady operating point (identified by load and revolution speed); measures from the sensors are sent, via the CompactRIO device to the destination PC. Then modeFRONTIER can act on the engine controller (i.e. the ECU) in order to optimise the chosen target.
The environment just described can exploit the real-time processing characteristics of the CompactRIO and test the optimization algorithms provided by modeFRONTIER. The communication between systems is guaranteed by the LabVIEW integration node, which makes the connection fast and simple.
During the first test, the spark advance was controlled in order to achieve the Maximum Brake Torque value (MBT); the main target was to test the effectiveness of the chosen HW/SW environment (see Fig. 1) and to check the efficiency of the interaction between LabVIEW and modeFRONTIER. modeFRONTIER was running on the test bed PC, receiving signals from in-cylinder pressure sensors, properly processed to obtain the target to be maximised (Indicated Mean Effective Pressure, IMEP) (7). Before the test, the range of spark advance degrees was set on modeFRONTIER; the Simplex (Nelder-Miller version) algorithm was chosen in order to get the maximum torque. The goal was reached in about 10 optimization steps.
A more complex and meaningful experiment to test the effectiveness of the system is the fuel-film automatic compensation. The idea is to use modeFRONTIER algorithms to identify in real-time the value of the X and Tau constants in order to compensate the fuel-film dynamics.
On a spark ignition engine with the injection on the intake manifold (Port Fuel Injection, PFI) a fraction of the injected gasoline (identified by the X value) falls on the intake runner wall; the deposited fuel goes in the cylinder after the evaporation time (the Tau value) (1).
To highlight the effect related to the fuel-film dynamics and to extract synthetic and meaningful information, the engine operating point (controlled in open-loop) has been properly set and the injectors have been driven with step waveform. In an ideal condition, the injection profile should be reproduced from the lambda signal (put on the exhaust collector). With the compensation algorithm off, the variations of the lambda signal do not follow the imposed steps due to the low-pass behaviour of the system. Instead, with the compensation on, the more the signal waveform is close to the ideal one, the better is the calibration. The evaluated signal error is strictly correlated to the difference between the ideal transitions and the measured air/fuel ratio. In Fig. 2a and 2b the waveforms are shown respectively before and after the compensation,; the used optimization algorithm was the Simplex.
Conclusions and further developments
The tight cooperation between Alma Automotive test environment and the optimization framework from ESTECO, quickly provided a demonstrator for the automatic tuning of two ECU subsystems. In particular, the high flexibility of Alma Automotive solutions, mainly based on National Instruments devices, granted the complete control of the engine and the test bed, while the advanced optimization algorithms included in modeFRONTIER allowed to reach quickly and efficiently the targets. The next step aims to apply the optimization at different engine operating points, mapping completely the compensation system. A possible further development will be to apply the proposed approach to other ECU subsystems in order to build a complete automatic calibration environment for the engine, avoiding direct user interaction while reducing costs and time.
