Integration of Dual-Clutch Transmissions in Hybrid Electric Vehicle Powertrains

Transcription

1 POLITECNICO DI TORINO Cluster MOBILITA - Project ITALY 2020 gomma CRF PhD in Mechanical Engineering XXX cycle Integration of Dual-Clutch Transmissions in Hybrid Electric Vehicle Powertrains Torino, October 2017 PhD student: Ing. Guido R. Guercioni Tutor: Prof. Alessandro Vigliani

2 Research activity overview Integration of Dual-Clutch Transmissions in Hybrid Electric Vehicle Powertrains Vehicle dynamic performance EMSs for HEVs Gearshift control for HEVs Optimal control theory The automotive industry is experiencing an increasing demand for improved NVH performance and reduced fuel consumption! Sub-optimal real-time implementable EMSs

3 HEV model for vehicle performance assessment Powertrain description BAS 32 kw ICE 2.0 L Diesel engine AMT or DCT 6-speed transmission Synchronizers on output shaft Fuel tank Dry clutch or DCU Clutch damper incorporated Mech. Connection BAS SMF DCT or AMT PTU Diff. Fuel energy Electrical energy Battery EM Battery 18.9 kwh Li-Ion EM 112 kw PTU Mech. Connection between EM and AMT output shaft A single gear ratio is available to EM 3

4 HEV model for vehicle performance assessment A model is developed to simulate the longitudinal vehicle dynamics for both vehicles. For simplicity, only the model for the variant with the AMT is shown. Both systems are modeled using the same assumptions. The complexity of the model enables it to be used for: Control calibration Drivability assessment Model description Dry clutch LuGre dynamic friction model. Synchronizer Switching model Gear engaged: gearmesh Speed synchronization: LuGre dynamic friction model. T e is the sum of the torque provided by the ICE and the BAS T EM is the torque provided by the EM Clutch damper Two stage clutch damper with piecewise linear stiffness and hysteresis 4

5 HEV model for vehicle performance assessment Model description AMT bearings Torque loss is the result of two contributions: Coulomb friction Viscous damping Response delay of BAS, ICE, clutch actuator and EM is considered. Vehicle Longitudinal load transfer Road is assumed flat Aerodynamic drag force Wheels A DOF for each wheel Pacejka 96 transient nonlinear tire model Rolling resistance Half-shafts Modelled as linear stiffness and damping elements 5

6 Gearshift control Gearshift control with AMT Control strategy aims at reducing vibrations while satisfying the driver s torque request in the shortest possible time. Gearshift phases Phase I: Go-to-slip Phase II: Opening the clutch completely Phase IV: Disengage current gear / Speed synchronization / Engage new gear Phase III skipped since slip velocity is higher than a predefined threshold Phase III: Prepare to start closing the clutch Phase V: Clutch slip control for engagement Phase VI: Close theclutch completely Controller output signals Torque request to the EM Torque request to the ICE Clutch transmissible torque request Synchronizer position request 4 th to 5 th

7 Gearshift control Gearshift control with DCT Control strategy aims at reducing vibrations while satisfying the driver s torque request in the shortest possible time. Gearshift phases: US Phase I: Go-to-slip Phase II: Prepare for torque phase Phase III: Torque phase Phase IV: Inertia phase - Fast speed matching Phase V: Inertia phase - Clutch slip control for engagement Phase VI: Close theclutch completely Controller output signals Torque request to the EM Torque request to the ICE Clutch transmissible torque request (for both) 4 th to 5 th

8 Gearshift control Gearshift control with DCT Control strategy aims at reducing vibrations while satisfying the driver s torque request in the shortest possible time. Gearshift phases: DS Phase I: Go-to-slip Phase II: Inertia phase - Fast speed matching Phase III: Inertia phase - Clutch slip control for engagement Phase IV: Close the clutch completely Controller output signals Torque request to the EM Torque request to the ICE Clutch transmissible torque request (for both) 4 th to 3 rd

9 Gearshift control AMT Upshift: 1 st to 2 nd at 100% APP DCT

10 Gearshift control AMT Downshift: 4 th to 3 rd at 100% APP DCT

11 Gearshift control ҧ ҧ ҧ ҧ Metrics for gearshift quality assessment 100% APP 1 st to 2 nd Gearshift quality criteria Parameter H-AMT H-DCT t Vehicle dynamic GS [s] a performance v [m/s 2 ] T err [%] J Shift comfort RMS [m/s 3 ] J Max/Min [m/s 3 ] Shift efficiency E c,loss [kj] Torque fill energy consumption E EM [kj] % APP 4 th to 3 rd Gearshift quality criteria Parameter H-AMT H-DCT t Vehicle dynamic GS [s] a performance v [m/s 2 ] T err [%] J Shift comfort RMS [m/s 3 ] J Max/Min [m/s 3 ] Shift efficiency E c,loss [kj] Torque fill energy consumption E EM [kj] Note: Transmission ratios of DCT and AMT were aligned for comparison (DCT values are used)

12 HEV model for energy management Powertrain description Fuel tank EM 75 kw DCT 6-speed transmission ICE 1.4 L 110 kw Battery 18.9 kwh Li-Ion Modeling considerations Clutch EM Battery A low-order dynamic model of the powertrain, accounting for losses in all the major powertrain components, is sufficient to capture almost all the energy flows in the vehicle. DCU Only vehicle longitudinal dynamics are considered Gearshift losses are considered ICE start losses are considered Fuel cut-off functionality is considered Constant gear efficiency is assumed ICE and EM are modeled using efficiency maps Battery is modeled using a zero-th order model Constant inverter efficiency is considered Constant power loss from the DC/DC converter Constant efficiency for the DC/DC converter was assumed Regenerative torque is bounded by the vertical load acting on the wheels Mechanical brakes are modeled as an ideal torque source Tires are modeled using a perfect rolling model DCT Mech. Connection Fuel energy Electrical energy

13 Dynamic programming Dynamic Programming is a numerical method to solve problems requiring a sequence of interrelated decisions. Each decision transforms the current situation into a different one. The objective is to find the sequence of decisions which minimizes the sum of the costs resulting from decisions along the sequence. Bellman s principle of optimality Starting from the final step N, the algorithm proceeds backward using the sequence of controls that generate the optimal cost-to-go from the current state to the end of the optimization horizon Not a causal strategy Gives the global optimal EMS DP is useful to: Design rule-based strategies Generate a benchmark solution DP formulation A general-purpose DP algorithm is used [1]. The HEV model and the definition of the control problem must be given as inputs. The cost to be minimize is the fuel consumption. State constrains SOC with SOC k 0.32,0.93 Gear number with gn k 0,6 Quick-disconnect clutch state with QD s,k 0,1 ICE state with ICE s,k [0,1] EM torque limit counter T lim,k [0,20] Initial and final state constrains are also indicated State dynamics SOC k = SOC k 1 + f k (SOC k 1, u k ) gn k = f k (u k ) QD s,k = f k (u k ) ICE s,k = f k (u k ) T lim,k = T lim,k 1 + f k (T lim,k 1, u k ) Control variables Torque split factor with sf k 1,1 Gear number with i k 1,6 Quick-disconnect clutch command with QD c,k 0,1 [1] O. Sundström, L. Guzzella, A generic dynamic programming Matlab function, in 18th IEEE International Conference on Control Applications (CCA) & Intelligent Control (ISIC) (IEEE, 2009), pp

14 Dynamic programming DCT loss model for gearshift events The simplified powertrain model for energy analysis during gearshifts is presented in the next figure: T e J e T cd J m T m T d T c1 τ tot,1 Tg J v τ tot,2 T v T c2 The total power request is: From the torque equilibrium: T c (t) = T c1 (t) + T c2 (t) T g t = T c1 t τ tot,1 + T c2 (t)τ tot,2 T d (t) = T in (t) + T c (t) From the power equilibrium at the two clutches: T c1 (t)w EM (t) = T c1 (t)w 1 (t) + P c1,loss (t) T c2 (t)w EM (t) = T c2 (t)w 2 (t) + P c2,loss (t) P T t = T d t w EM t = T in t w EM t + T c t w EM t The total energy request is then: T spt E T t = න (t)dt 0 The mean total torque request is: T T (t) = E T T s T s ഥw EM

15 Dynamic programming DCT loss model for gearshift events For illustration a 2 nd to 1 st DS is presented in the following figures. The torque passing through the clutches: The speed profiles of the EM shaft and the two primary shafts: The energy dissipated in the clutches is not negligible. In this case, it is almost the same than the one needed to accelerate the powertrain components inertias

16 Dynamic programming Simulation results WLTC cycle (3 repetitions) No mechanical braking is needed. It can be present in the results (near the end of the driving cycle) if the SOC resolution is low enough due to numerical errors. True charge-sustaining operation is obtained

17 Dynamic programming Effect of the ICE start and Gearshift losses To assess the effects of the ICE start and gearshift losses, the results for a CS simulation of 3 repetitions of the WLTC cycle will be compared for three different cases: GS + ES losses GS losses No losses Total number if gearshifts Total number of ICE start events (+23%) 1504 (+237%) (+287%) 256 (+216%) Fuel consumption (-3.4%) 2524 (-4.5%) The effect of the ICE start losses on fuel consumption is higher than the one of the gearshift losses A more intermittent use of the ICE is encountered when the ICE start losses are neglected

18 Dynamic programming Effect of the ICE start and Gearshift losses Total number of gearshifts The following figures show the effect of the gearshift losses and their relation with the ICE start losses. ES + GS losses vs. No losses GS losses vs. No losses The number of gearshifts increases significantly when the losses associated to it are neglected. Neglecting the ICE start losses could lead to an unrealistic use of the ICE inertia to overcome gearshift losses.

19 RB + A-ECMS Algorithm overview A novel EMS is developed in which at each iteration: A rule-based approach is used for gear selection A-ECMS determines the torque split factor Main inputs Drive cycle information Vehicle speed request Total distance to be traveled Initial SOC Final SOC target EV-mode gearshift rules A-ECMS parameters Initial value of equivalence factor Sampling distance for equivalence factor adaptation Proportional gain for equivalence factor adaptation Fuel penalty to enforce ICE state and gearshift hysteresis Minimum fuel gain allowed when deciding to change ICE state Main outputs SOC Fuel consumption Gear ICE state Quick-disconnect clutch state Torque split factor ICE torque ICE speed EM torque EM speed

20 RB + A-ECMS Rule extraction from the gear schedule in EV-mode It was seen in the optimal solution computed with DP that the gear schedule when increasing the final SOC target could be interpreted as a deviation from the one seen for the driving cycles completed only in EV-mode. The idea behind the RB + A-ECMS approach is that if it can be understood how the optimal solution selects the gear in EV-mode, when the ICE is needed, gear selection is simply a matter of minimizing the equivalent fuel consumption. Rule extraction process 1. DP is used to find the optimal solution for driving cycles that can be completed entirely in EV-mode 2 nd to 5 th Based on DP results 2. Speed ranges are identified for each new gear after a gearshift event 3. Rules for single gearshifts are extracted 1. General rules are extracted from the EM power and vehicle speed plane 2. Detailed rules are extracted considering: vehicle acceleration, EM power request trend and EM power difference with respect to the no gearshift case 4. Rules for multiple USs are extracted based on the EM power request and vehicle acceleration 5. Rules for gearshifts prior to significant braking events are extracted considering: vehicle speed and acceleration, EM power request trend and EM power difference with respect to the no gearshift case For increasing vehicle speed, multiple USs are performed if the EM power request is close to the one of the single gearshift.

21 RB + A-ECMS Rule extraction from the gear schedule in EV-mode Rule extraction results Generally speaking, the gear schedule for DP is reproduced, but there are also several instances in which different decisions are made: However, the SOC profile obtained is very close to the DP solution: Main reason why the SOC profile is reproduced: The rule based approach is able to recognize the relevant EM power request and acceleration trends Limitations on gearshift losses

22 RB + A-ECMS A-ECMS Equivalent fuel consumption Equivalent fuel consumption at each time step will be computed as: m f is the actual fuel consumption s eq is the equivalence factor The units of the equivalence factor are then grams. Adaptation of the equivalence factor Equivalent factor is computed as: D s is the sampling distance D is the current covered distance K p is a proportional gain mሶ eq t = mሶ f t + s eq (t) ሶ SOC(t) S eq t = S eq D + D s = S eq D + S eq (D D s ) + K 2 p (SOC ref D SOC(D)) The SOC reference for the equivalent factor adaptation is linear. It is computed based on the total travel distance, initial SOC and final SOC target. Penalties to enforce gearshift hysteresis Equivalent fuel consumption is increased by a certain quantity if: An US is performed when a DS has been performed within a certain user specified time A DS is performed when an US has been performed within a certain user specified time Penalties to enforce ICE state hysteresis Equivalent fuel consumption is increased by a certain quantity if: ICE is turned on before being off for a certain user specified time ICE is turned off before being on for a certain user specified time Furthermore, when there is a change in the ICE state, a minimum gain in terms of equivalent fuel consumption is required.

23 RB + A-ECMS Simulation results A charge-depleting simulation of the WLTC cycle is presented. The final SOC is % (-0.29 % w.r.t DP result). The fuel consumption is g (+4.6 % w.r.t the DP result). The gear schedule does not present chattering. This is because gearshift hysteresis is introduced.

24 RB + A-ECMS Simulation results DP RB + A-ECMS The ICE speed and torque limits are respected by both solutions DP The EM speed and torque limits are respected by both solutions.

25 RB + A-ECMS Effect of restrictions on ICE state The following restrictions to changes in the ICE state are applied by adding penalties to the equivalent fuel consumption: ICE should be off for at least 4 s ICE should be on for at least 3 s ICE is turned on or off if the equivalent fuel consumption reduces at least 0.2 g Free ICE state Constrains on ICE Different torque split factors are selected w.r.t the DP results The ICE has a very intermittent on/off behavior Torque split factor selection is similar to the one obtained with DP The ICE state does not present chattering behavior

26 Publications Journal Guercioni G.R., Galvagno E., Vigliani A., An alternative method for order tracking using autopower spectrum, Advances in Mechanical Engineering, 2015, vol. 7 (11), pp ISSN: DOI: / Congress Galvagno E., Guercioni G.R., Velardocchia, M., Experimental analysis and modeling of transmission torsional vibrations, Recent Researches in Mechanical and Transportation Systems, 2015, pp ISBN: Galvagno E., Guercioni G.R., Vigliani A., Sensitivity Analysis of the Design Parameters of a Dual-Clutch Transmission Focused on NVH Performance, 2016 SAE world congress and Exhibition, Detroit (USA), April 12-14, 2016, pp ISSN: DOI: / Guercioni G.R., Vigliani A., Galvagno E., Midlam-Mohler S., Gearshift Control for Hybrid Powertrains with AMTs, 2017 International Conference of Electrical and Electronic Technologies for Automotive, Torino (Italy), June 15-16, 2017, pp ISBN: DOI: /EETA Accepted papers System Diagnosis and Mitigation Strategies for an Automated Manual Transmission Authors: Simon J. H. Trask, Gregory J. Jankord, Aditya A. Modak, Brian M. Rahman, Giorgio Rizzoni, Shawn Midlam-Mohler, Guido R. Guercioni. Conference name: Dynamic Systems and Control Conference 2017 Conference dates: October 11-13, Conference location: Tysons Corner, Virginia, USA.