Virtual Calibration for Mild-Hybrid Powertrain Development
Tushar Chodvadiya
E&D, TATA Technologies Limited, Pune, India
Karishma S. Patil, Chandrakant M. Awate
Engineering and Research Centre, TATA Motors Limited, Pune, India
ABSTRACT
Control variables, being nerves of any control strategy, play important role in improving performance of vehicle. In Mild hybrid powertrain development program, initial phase of control strategy development is characterized by setting initial values to these variables on the basis of pre knowledge of vehicle behavior say road data. This is followed by fine tuning of such calibration variables to achieve the desired performance and fuel economy. This calibration can’t be done using repetitive vehicle level tests as it is time consuming process. Hence we have used model based virtual calibration process to bring the control strategy to a quite matured level before performing vehicle level test.
INTRODUCTION
Considering the complexity of hybrid control system, it needs more controllable parameters and thus increases the control algorithm development time and cost. The traditional approach to calibrate the system is becoming too time consuming and more expensive. With current market competition, developing the control strategy, testing and calibration must be completed quickly and effectively. The V-model design process provides a methodical approach to design, test, calibrate, and validate control algorithms. Model-in-the-loop (MIL) and hardware-in-the-loop (HIL) simulations are used to validate the control algorithm before implementing into the vehicle. MIL and HIL are used to make sure the control system is robust, but the algorithms still must be calibrated to provide reliable operation and optimized performance.
In mild hybrid system, along with conventional powertrain, electric machine also becomes add-on part of it. To determine optimal setting of engine and electric machine with respect to performance, fuel economy and emissions, the characteristics of powertrain system must be accurately measured. Because of so many variable parameter, a complete survey becomes extremely time consuming. The time spent on testing and calibration of developed control systems is quite substantial, and this chapter will explain a new approach, the model-based calibration techniques which reduces the calibration time as well as cost and improves the quality of control systems.
In virtual calibration, initially we have separated lots of available road data from conventional/ base vehicle into different driving pattern for example City drive, highway drive and hill terrain. We used those road data and driving patterns for setting initial values for calibration variables. What value of particular variables triggers which mode is defined using this available data. Control strategy was then tested and results were analyzed in terms of closeness towards expected results. We already had known importance and significance of each mode. At the same time, simulation in AVL cruise gives fuel economy and Battery SOC balancing results. Thus comparing these outputs with expected outputs and having information of significance of each mode, calibration can be changed.
Need for virtual calibration
12V Mild hybrid application uses small electric machine whose power rating is very less as compared to power of engine; for example, 12V electric machine would deliver 1.2 kW continuous power during torque assist function which is comparatively very less power to greatly affect the engine performance. This power comparison between 12V electric machine and engine gave an idea that engine performance during driving will not be greatly changed by this 12V electric machine. It was found that emission reduction is mainly due to stop-start functionality as compared to boost and idling from e-machine. Apart from stop-start, fuel economy directly depends on time duration of assist and generation function, because during generation (alternator mode), loading the engine leads to increased fuel consumption; whereas assist function will cut off alternator load from engine and increase the fuel economy. We had used these facts during testing and calibrating our control strategy.
To completely simulate the vehicle behavior and performance, plant model for mild hybrid system is required. After considering prior studies of 12V mild-hybrid vehicle, it was seen that mild-hybrid vehicle performance can also be analyzed by observing the function duration. Along with the function duration, battery SOC also should be considered because energy should be balanced before and after the vehicle drive. Battery SOC will tell about how much electrical energy is available for boost and idle mode. Battery charging and discharging current had been derived from basic battery and motor model in AVL cruise simulation. Average value of battery current for each modes and time duration for each mode would be used to estimate battery SOC level before and after test drive.
So, tests were conducted in model-in-the-loop (MIL) and hardware-in-the-loop (HIL) simulations without using plant model just to validate the functionality. Control algorithm was calibrated based on mode duration and battery SOC balancing instead of complete vehicle behavior. This reduced a lot of efforts for calibration in term of time and complexity. We have used test data recorded from city drive, high way drive and chassis-dyno cycles for conventional vehicle.
VIRTUAL CALIBRATION PROCESS ADOPTED FOR 12V MILD-HYBRID PROJECT
Following figure illustrate the process:
- Select logged data from test drive cycle for which we want to calibrate control algorithm. We had multiple test drives data of conventional vehicle which we had used in our project. For specific test, we had also written test cases which is in excel format. For manually written test cases, we may require vehicle behavior knowledge, because we don’t use plant model.
- Logged data in excel sheet is then imported to signal builder block. Because logged data is in excel format and model is in MATLAB, so we need to bring both on same format/platform. After importing excel file to signal builder, individual signals are generated, those will feed all the required inputs to control algorithm.
- Set values for calibration data, run the simulation and record modes output from control algorithm. Recorded mode output is a time sampled data, so export this data into excel sheet for further processing and analysis.
- Exported data of hybrid modes is converted as time distribution chart same as shown in figure 2. Here, each mode is represented in term of time taken during complete test cycle.From this data, we can identify that how long is charging modes and how long is discharging modes. And based on average battery current in each mode and each mode duration for given test cycle, we can calculate gained and consumed amount of battery energy for individual mode.
- Analyze the time distribution of each modes and make corrective changes in calibration data to improve the model behavior.
- Repeat the test with updated calibration values and compare the output with previous test data as show in figure2.
- If requires, repeat the complete process for different test data or manually written test inputs.
CASE1: Mode Duration
In above example of two different set of calibration values, we can identify changes in mode duration particularly for Idle and Alternator mode. In test-1, we ran the simulation with some set of calibration values and we observed that charging events (Alternator, Coast down, Recuperation mode) were more and hence battery SOC was increased after completing the test cycle. Here, electrical energy gained from Coast down and Recuperation mode are regenrative energy i.e. free enegy; but energy from Alternator mode is from fuel consumed by engine. Hence, fuel economy can be increased by reducing this extra charging events i.e. Alternator mode and at the same time also esures that SOC is maintained.
After re-calibrating the value, we ran this test and recorded the mode as Test-2. Here, we could identify the change of Alternator mode duration. As alternator mode duration is reduced, there was no extra charging event which burns fuel and charge the battery. Hence fuel economy would be increased.
- CASE2: Electrical Energy Balance
In figure 3, Test1 pie chart represents energy flow during individual mode. From this chart, we identify that total discharging energy (Stop, Start, Boost, Passive Assist mode) is more than total charging energy (Alternator, Coast-Down, Recuperation mode). This leads to discharge the battery after the drive cycle. One way to improve this is by calibrating the setpoint voltage for alternator mode. Test2 chart represents the improved data after updating the alternator mode setpoint voltage.
Here charge accumulation is nothing but the accumulation of averge battery current for each mode during complete test cycle. We got average battery current for each mode from AVL simulation. For that, we ran the same test drive in AVL and recorded the average current. This way we can calibrate thoes parameters which mostly affect the battery energy balancing.
This process goes on and on till best results are achieved.
If it had been done on vehicle every time, it would have consumed too much of time and resources. There is possibility of hazard while testing directly on vehicle by changing calibration. Hence doing it in virtual environment, saved lot of time and gave best possible combinations of variables. Hence control strategy is brought to a mature level before doing actual vehicle level test and help to mitigate risk and reduce wastage of resources.
This way, we had calibrated out mild-hybrid control algorithm for different drive cycle and came up with the most suitable calibration value for all test drive cycle. We had found this approach very easy, time and cost effective.
Conclusion
By conventional approach, the time spent on testing and calibration of developed control systems is quite substantial. That also requires calibration engineer to work on actual physical environment and vehicle which increase cost and efforts.Model based calibration more specific virtual calibration approach improves calibration quality and minimize experimental efforts and cost.
Acknowledgments
Authors would like to thank Mr. Sanjay Patel for technical inputs on electrical system and Mr. Prathmesh Sawarbandhe for helping in AVL simulation of hybrid vehicle.
References
- Ryan Vincent Everett-An Improved Model-Based Methodology for Calibration of an Alternative Fueled Engine:The Ohio State University, 2011.
- Ramesh Kumar, Ramanathan Annamalai, Shivaram Kamat- Virtual Simulations in ECU Software Development: Control Measurement and Calibration Congress, 2015.
- A Cost Effective and Optimal Energy Management Strategy for Hybrid Electric Vehicles (HEV) based on Emission Analysis: 15ITEC-0063.
- Document on test method, testing equipment and related procedures for testing type approval and conformity of production (COP) of vehicles for emission as per central motor vehicle (CMV) rules 115, 116 and 126. Document number: MoRT/CMVR/TAP-115/116.
CONTACT
Tushar Chodvadiya received Bachelor degree in Electronic and Communication Engineering from L.D. College of Engineering, Ahmedabad, India in 2010. He worked for Larsen & Toubro Technology Services, Mumbai, India for 4 year in area of control system design, model based development and testing of Off-Highway vehicles. He has been working with Engineering and Design division of TATA Technologies Limited, Pune, India since 2014 in area of automotive embedded software development, model based design and testing for hybrid electric passenger vehicles. He is reachable at .
Karishma Patilis Assistant Manager at TATA Motors Limited, Pune, India and she is reachable at .
Chandrakant Awateis Assistant General Manager at TATA Motors Limited, Pune, India and he is reachable at .
Definitions, Acronyms, Abbreviations
ECU: Electronic Control Unit
MBC: Model based Calibration
MBD: Model based Development
MBT: Model based Testing
MIL: Model-In-the-Loop
HIL: Hardware-In-the-Loop
BSG: Belted Starter Generator
E-Machine: Electric Machine
APPENDIX
Figure 1:Virtual Calibration Process for 12V Mild-Hybrid Vehicle
Figure 2: Test results with different calibration values
Figure 3: Energy distribution improvement for two different test