Artificial Intelligence Based Calibration and
Predictive Control for Future Engines
Professor Hongming Xu
International Summit on Break-out Technology of Engines and Fuels
Vehicle & Engine Technology Center, University of Birmingham
State Key Lab of Automotive Safety & Energy, Tsinghua University
August 21, 2018, Tianjin
Outline
• Introduction
• Research Facility and Methodology
• Results and Discussion
• Future Outlook
• Conclusion
23/08/2018 2
Future vehicles/engines on the way
portal.ku.edu.tr/.../VEHICLE%20TRACKING%20SYSTEM.ppt
Artificial intelligent
and unmanned control! Wearing the Google goggles or using a Google driveless car?
3
Research Facilities
23/08/2018 4
Engine test bench
Advanced Engine control
Rapid Control
Prototyping (RCP)
INCA-EIP INTECRIO
Design of control
strategy
SIMULINK
Research Facilities
23/08/2018 5
AJ200D
Signal Line
Fuel Flow
Water Flow
Low Temperature Flow
High Temperature Flow
Mechanical Connection
Open
ECU
Crankshaft
On-board
Combustion
Analyzer
AVL I60 DMS500
(Emission Analyzer) Puma
Control Cell
Low Temperature Chiller
(Minimum -29 ℃)
Pedal
Actuator
Diesel Supply Pipe
Filter
AVL Fuel Mass Flow
Meter
Valve
Fuel Condition
System
(Cold Water Flow)
Coolant
System
(Cold
Co
ola
nt)
(Hot
Co
ola
nt)
Oil
Condition
System
(Hot Water Flow)
1.5 Bar
Control & Data Acquisition
System
(Hot
Oil)
(ASAP 3)
Cold Air Flow
(CAN)
(In-cylinder
Pressure)
(Lambda Sencor)(Exhaust)
Cold OilHot Oil
Valve
(Minimum -10 ℃)
(Cold
Oil)
(Cold
Wa
ter
Flo
w)
(Hot
Wa
ter
Flo
w)
Desiccant
Dehumidifier
Air
Condition
System
(Cold Air)
(Air Inlet)
(Hot Air)
Hot Air Flow
(CAN)
(Enthernet)
INCA
Transient Test Bench
Open ECU
AVL Transient Testing Bed (PUMA), -20oC
Research Methodology
23/08/2018 6
Test Bench Validation
Control Algorithm
Hardware In the Loop Test System
Offline Test
Online Test
Controller
Development
Control-oriented Vehicle Engine Model
Research Methodology
Artificial Intelligence Intellectualizing the Future
Evolutionary Algorithm
Machine Learning
Fuzzy Logic
Application • Intelligent Engine
calibration • Intelligent
Component Sizing • Top Level Online
Intelligent Control
Application • Model-Free
Predictive Control • Complex System
Modelling • Driver Behavior
Prediction • Advanced Model
Predictive Control
Application • Lower Level System
Control • Electric Motor
Control • Driver Model for
Front-forward Vehicle Simulation Platform
• Clustering and Classification
Research Methodology
• Fuzzy Logic Controller (FLC) • Fuzzy clustering
• Strength Pareto Evolutionary Algorithm 2 (SPEA2)
• Chaos-enhanced Accelerated Particle Swarm Optimization (CAPSO)
• Model Predictive Control (MPC) • Model-free Predictive Control (MFPC)
23/08/2018 9
Calibration Methods
He Ma, Ziyang Li, Tayarani M., Lu G., Xu H.. Yao X., Model-based Computational Intelligence Multi-Objective Optimization for GDI Engine Calibration, Proc. IMechE , Part D: Journal of Automotive Engineering, (https://doi.org/10.1177/0954407018776743)
• Calibration with Evolutionary Algorithm
23/08/2018 10
• Transient Calibration with Chaos-enhanced Accelerated Particle Swarm
Optimization Algorithm
Y. Zhang, H.M. Xu*, Intelligent Transient Calibration of a Dual-loop EGR Diesel Engine using Chaos-enhanced Accelerated Particle Swarm Optimization Algorithm, Proc. Part D: J. Automotive Eng (https://doi.org/10.1177/0954407018776745 )
𝑔𝑖,∗
x𝑖,𝑗
𝜷(𝑔𝑖,∗ − x𝑖,𝑗)
α
α ∙ 𝑟𝑗(𝑖)
x𝑖+1,𝑗
Step 1: moving the computing agent towards the best position with a proportion 𝜷
Step 2: moving the computing agent randomly within the circle with radius α
The best positon of current iteration
Calibration Methods
Block diagram of the PI-like FKBC
Setpoint /
Desired value
Measured value
/ Actual value
Error “e”
• Self-adaptive Fuzzy Logic Control (FLC)
23/08/2018 11
Block diagram of the PI-like FKBC
“if-then” rules
𝑒 ∗ 𝐾𝑖
∆𝑒 ∗ 𝐾𝑝
∆𝑢
Fuzzification Defuzzification Input
Input
Output
The rule in column 4 and
row:
IF 𝑿𝟏 is Zero, AND 𝑿𝟐 is
Negative Medium, THEN
PI output = Positive Small.
Control Methods
Ziyang Li, Ji Li, Quan Zhou, Yunfan Zhang, and Hongming Xu*, “Intelligent Air/Fuel Ratio Control Strategy with a PI-like Fuzzy Knowledge Based Controller for GDI Engines”, Proceedings of the Institution of Mechanical Engineers, Part D: Journal of Automobile Engineering. https://doi.org/10.1177/0954407018779180
23/08/2018 12
Engine Plant
Prediction Model Bank
MPC Controller
Bank
Controller Weight Bank
2D-map Based Switch
Scheme
𝑖𝑓 𝑁𝑒𝑛𝑔 ∈ 𝐴𝑖 , 𝐵𝑖 𝑎𝑛𝑑 𝑄𝑓𝑢𝑒𝑙 ∈ 𝐶𝑖 , 𝐷𝑖
𝑡ℎ𝑒𝑛 𝑊𝑦 = 𝑊𝑦𝑖; 𝑊𝑢 = 𝑊𝑢𝑖
; 𝑊∆𝑢 = 𝑊∆𝑢𝑖
𝑎𝑛𝑑 𝑖𝑛𝑡𝑒𝑟𝑛𝑎𝑙 𝑚𝑜𝑑𝑒𝑙 𝑁 𝑖𝑠 𝑎𝑐𝑡𝑖𝑣𝑒)
𝑡ℎ𝑒𝑛 𝑠𝑤𝑖 𝑖𝑠 𝑁 (𝑖. 𝑒 𝑀𝑃𝐶 𝑁
• Model Predictive Control (MPC)
Control Methods
Y. Zhang, H.M Xu*, Tuneable Model Predictive Control of a Diesel Engine with Dual Loop Exhaust Gas Recirculation, Proc. IMechE , Part D: Journal of Automotive Engineering, First Published Oct 2017
• Model-based Computational Intelligence Multi-objective Optimization for GDI Engine Calibration
23/08/2018 13
ISFC
ISPMN
ISPMM
3.2%
16.5%
10.3%
He Ma, Ziyang Li, Tayarani M., Lu G., Xu H.. Yao X., Model-based Computational Intelligence Multi-Objective Optimization for GDI Engine Calibration, Proc. IMechE , Part D: Journal of Automotive Engineering, (https://doi.org/10.1177/0954407018776743)
Results and discussion
• Computational Intelligence Non-model-based Calibration Approach (CINCA)
23/08/2018 14 Advanced Engine control
BSFC PMn PMm
3.1% 6.8% 6.9%
Ma H., Li Z., Tayarani M., Lu G., Xu H. Yao X., Computational Intelligence Nonmodel-Based Calibration Approach for Internal Combustion Engines. ASME Journal of Dynamic System, Measurement, and Control. Vol 140 (4), 2018
Results and discussion
Convergence of particles after 50 loops of iteration
(1500 rpm / 8.3 bar BMEP )
• Intelligent Air/Fuel Ratio Control Strategy with a PI-like Fuzzy Knowledge Based Controller for GDI Engines
23/08/2018 15
Self-adaptive and self-tuning
Settling time
ITAE
56.25%
58.67%
Ziyang Li, Ji Li, Quan Zhou, Yunfan Zhang, and Hongming Xu*, “Intelligent Air/Fuel Ratio Control Strategy with a PI-like Fuzzy Knowledge Based Controller for GDI Engines”, Proceedings of the Institution of Mechanical Engineers, Part D: Journal of Automobile Engineering. https://doi.org/10.1177/0954407018779180
Results and discussion
23/08/2018 16
• Intelligent Transient Calibration of a Dual-loop EGR Diesel Engine using Chaos-enhanced Accelerated Particle Swarm Optimization Algorithm
Y. Zhang, Q.Zhou, Z. Li, J.Li, H.M. Xu*, Intelligent Transient Calibration of a Dual-loop EGR Diesel Engine using Chaos-enhanced Accelerated Particle Swarm Optimization Algorithm, Proc. IMechE , Part D: Journal of Automotive Engineering, (https://doi.org/10.1177/0954407018776745 )
Results and discussion
Results and discussion
23/08/2018 17
• Tuneable model predictive control of a turbocharged diesel engine with dual loop exhaust gas recirculation
Y. Zhang, H.M Xu*, Tuneable Model Predictive Control of a Diesel Engine with Dual Loop Exhaust Gas Recirculation, Proc. IMechE , Part D: Journal of Automotive Engineering, First Published Oct 2017
23/08/2018 18
1300 1305 1310 1315 1320
0
2
4
6
8
ide
al_
lpeg
rflo
w (
g/s
)
time (s)
ideal_lpegrflow
PID_lpegrflow
MPC_lpegrflow
1300 1305 1310 1315 1320
0
5
10
15
ide
al_
hpe
grf
low
(g/s
)
time (s)
PID_hpegrflow
MPC_hpegrflow
ideal_hpegrflow
1300 1305 1310 1315 1320
0
50
100
ideal_
lpegrs
plit
(%
)
time (s)
ideal_lpegrsplit
PID_lpegrsplit
MPC_lpegrsplit
1300 1305 1310 1315 1320
0
20
40
60
80
idea
l_eg
rrat
e (g
/s)
time (s)
ideal_egrrate
PID_egrrate
MPC_egrrate
1) MPC makes more use of LPEGR instead of HPEGR than the PID controller 2) Less EGR overshoot. Produces more torque for the same fuel delivery 3) More energy through turbo because less HPEGR use – more efficient
Results and discussion
23/08/2018 19
The mechanism of the BSFC reduction via the MPC-based controller is the improved VGT efficiency. Compare with the PI controller, the MPC-based controller achieves better performance on regulating the H/LPEGR fractions, while keeping the total EGR rate as the target value.
• Tuneable model predictive control of a turbocharged diesel engine with dual loop exhaust gas recirculation
Results and discussion
Algorithm interface Optimal vehicle performance
Optimisation process
Intelligent Sizing of for the Hybrid Engine
• An algorithm for hybrid electric powertrain intelligent sizing is developed
• The proposed CAPSO algorithm is capable of finding the real optimal result with much higher reputation.
• The CAPSO gave more reliable results and increased the efficiency by 1.71%.
Zhou Q., Zhang Y., Li Z., Li J, Xu H.*, Oluremi O., Cyber-Physical Energy-Saving Control for Hybrid Aircraft-Towing Tractor based on Online Swarm Intelligent Programming, IEEE Transactions on Industrial Informatics, 2018,
Modelling of energy-flow
Target vehicle Controller framework
HiL test result (vs. CD/CS)
The OSIP can optimize the vehicle performance in real-time with a maximum prediction horizon size of 35s.
The vehicle with OSIP outperforms the system without it in energy saving at all initial battery SoC level
The proposed energy management method is robust and reliable, and up to 17% fuel and 13% total energy saving
Energy Management for Hybrid
Zhou Q., Zhang Y., Li Z., Li J, Xu H.*, Oluremi O., Cyber-Physical Energy-Saving Control for Hybrid Aircraft-Towing Tractor based on Online Swarm Intelligent Programming, IEEE Transactions on Industrial Informatics, 2018,
Outlook
Model-free predictive control framework
Learning process of Alpha-go
Learning process of model-free predictive EMS (considering different prediction length)
Vehicle learns the optimal
control policy from Real-world
like Alpha-go
Summary & Conclusion
• The development of Artificial Intelligence technology has
provided a new horizon for the design and operation of future
combustion engines. Further improvement and optimization of
the engine system will be possible beyond the conventional
present possibilities.
• Engine calibration will transit from human knowledge-based
methods to AI based methods, which can resolve much more
complex problems involving multi-variables and multi-
objectives in much shorter time and at lower cost.
• Predictive optimal engine control will come into application,
from the linear-model-based to nonlinear-model-based and
finally to model-free predictive control with machine learning
capability.
23/08/2018 23
Publications
23/08/2018 24
1. Zhou Q., Zhang Y., Li Z., Li J, Xu H.*, Oluremi O., Cyber-Physical Energy-Saving Control for Hybrid Aircraft-Towing
Tractor based on Online Swarm Intelligent Programming, IEEE Transactions on Industrial Informatics, 2018, DOI
10.1109/TII.2017.2781230.
2. Zhou Q., Zhang W., Cash S., Olatunbosun O., Xu H.*, Lu G., Intelligent sizing of a series hybrid electric power-train
system based on Chaos-enhanced accelerated particle swarm optimization. Applied Energy 2017:189:588–601.
3. Ma H., Xu H., Wang J., Schnier T., Neaves B., Tan C., et al. Model-based Multi-objective Evolutionary Algorithm
Optimization for HCCI Engines. IEEE Trans Veh Technol, 2015;64:4326–31. doi:10.1109/TVT.2014.2362954.
4. Ma H., Li Z., Tayarani M., Lu G., Xu H. Yao X., Computational Intelligence Nonmodel-Based Calibration Approach for
Internal Combustion Engines. ASME Journal of Dynamic System, Measurement, and Control. Vol 140 (4), 2018
5. M. H. Tayarani-N., X. Yao* and H. Xu, "Meta-Heuristic Algorithms in Car Engine Design: A Literature Survey," in IEEE
Transactions on Evolutionary Computation, vol. 19, no. 5, pp. 609-629, Oct. 2015. doi: 10.1109/TEVC.2014.2355174
6. Y. Zhang, G. Lu, H.M Xu*, Ziyang Li, Tuneable Model Predictive Control of a Diesel Engine with Dual Loop Exhaust Gas
Recirculation, Proc. IMechE , Part D: Journal of Automotive Engineering, First Published Oct 2017
7. Y. Zhang, Q.Zhou, Z. Li, J.Li, H.M. Xu*, Intelligent Transient Calibration of a Dual-loop EGR Diesel Engine using Chaos-
enhanced Accelerated Particle Swarm Optimization Algorithm, Proc. IMechE , Part D: Journal of Automotive Engineering,
(https://doi.org/10.1177/0954407018776745 )
8. He Ma, Ziyang Li, Tayarani M., Lu G., Xu H.. Yao X., Model-based Computational Intelligence Multi-Objective
Optimization for GDI Engine Calibration, Proc. IMechE , Part D: Journal of Automotive Engineering,
(https://doi.org/10.1177/0954407018776743)
9. S. Cash, O., Olatunbosun, ‘Fuzzy logic field-oriented control of an induction motor and a permanent magnet synchronous
motor for hybrid/electric vehicle traction applications’, International Journal of Electric and Hybrid Vehicles, Vol.9 (3),
2017 page 269-284
10. Ziyang Li, Ji Li, Quan Zhou, Yunfan Zhang, and Hongming Xu*, “Intelligent Air/Fuel Ratio Control Strategy with a PI-like
Fuzzy Knowledge Based Controller for GDI Engines”, Proceedings of the Institution of Mechanical Engineers, Part D:
Journal of Automobile Engineering. (https://doi.org/10.1177/0954407018779180 )
Acknowledgement
23/08/2018 25 Advanced Engine control
Many thanks to
All co-investigators and authors/co-authors of the publications used
in this presentation
Support from Jaguar Land Rover, Etas National Instrument and PI
Technology
Technical Staff at the Vehicle and Engine Research Centre
University of Birmingham.
Thank you for your attention
Professor Hongming Xu, FIMechE, FHEA, FSAE
School of Mechanical Engineering
University of Birmingham, B15 2TT
Tel: ++44 121 414 4153
Email: [email protected]
Publication list http://www.birmingham.ac.uk/staff/profiles/mechanical/xu-hongming.aspx
Contact