The empact CVT : dynamics and control of an electromechanically actuated CVT

T.W.G.L. Klaassen

Research output: ThesisPhd Thesis 1 (Research TU/e / Graduation TU/e)

1137 Downloads (Pure)


The large ratio coverage of a CVT and the possibility to choose the engine speed in a wide range independently of the vehicle speed enables the ICE to operate at more fuel economic operating points, making the vehicle potentially more fuel efficient. Unfortunately, because the energy dissipation of the CVT itself is higher than that of a manual transmission, this efficiency improvement is partly lost. The main power losses in the CVT are due to the inefficient hydraulic actuation system and the excessive clamping forces used to prevent the belt from excessive slippage. Direct control of the slip can significantly increase the efficiency. Due to the low actuation stiffness at low hydraulic pressures, the hydraulically actuated CVT is not well suited for slip control. The Empact CVT, developed at the TU/e, is an electromechanically actuated pushbelt type CVT, which has a high stiffness at low clamping forces and is suitable for slip control. This system reduces the steady-state losses, which are dominantly present in a hydraulic system. The goals of this research are to achieve optimal efficiency of this system, to obtain good tracking performance and to prevent the pushbelt from slipping excessively. These objectives are experimentally validated at a Empact prototype, which is tested at a test rig and implemented in an Audi A3 2.0 FSI. The Empact CVT uses two servomotors to actuate the moveable pulley sheaves. To decouple the rotation of the input and output shaft from the servomotor rotations, a double epicyclic set is used at each shaft. The system is designed, such that one (primary) actuator accounts for the ratio changes and one (secondary) actuator sets the clamping forces in the variator. To optimally use the efficiency potential of the Empact system, the slip in the variator must be controlled. In this way, the clamping forces reduce to small values, thereby reducing the friction forces in the gears, spindles and bearings. Efficiency improvements of up to 20 [%] can then be reached at partial load (during 75 [%] of the duration of the FTP72 cycle) compared to a conventionally controlled CK2 147 transmission and efficiency gains of up to 10 [%] compared to an optimally, slip controlled CK2. To gain insight in the physical behavior of the Empact CVT, a multi-body model of the system has been developed, which incorporates a dynamical description of all major components of the test setup. Results show a realistic behavior of the system for both stationary and transient situations. Although this nonlinear simulation model gives a basis for control design and yields a realistic description of the closed loop system, for the actual control design an approximate, linear plant model that describes the frequency domain behavior of the system is estimated. These linearized descriptions are obtained from the simulation model using approximate realization from pulse response data. An iterative model identification and control design procedure is used, such that the plant is estimated in closed loop. In this way, the uncertainty in the frequency range of importance for the design of the controllers is reduced, which leads to less conservative control designs. Parallel to the identification and control design with the simulation model, this procedure is also applied for the test setup. Due to high measurement noise and excessive friction in the system, the quality of the approximated plants at the test setup is relatively low. The time responses are however comparable to the results from the simulation model. An important constraint for the controlled system is that slip cannot be controlled under all operating conditions. At low variator speeds and low loads, the slip controller must be switched off. A decentralized control structure is chosen. Pairing of the in- and outputs is primarily based on the mechanical design of the Empact CVT and are supported by a interaction analysis. The controllers are designed using a sequential loop closing procedure, in which the slip loop is closed last, such that stability of other loops is guaranteed independent of the switching of the slip controller. Using manual loop-shaping, decentralized lead-lag controllers are designed. Nominal stability and performance can be guaranteed. To obtain robust performance, gain scheduling of the slip controller is implemented. Resulting closed loop bandwidth is 8-10 [Hz] for both the ratio and slip control loops. Because the slip dynamics is not well defined at low or zero variator speeds, the slip controller is partly switched off below 2 [km/h]. Both the simulation model and the experimental setup show very good results for disturbance rejection and tracking performance. Torque disturbances of up to 100 [Nm], applied at the secondary variator shaft, can be suppressed within 0.2 [sec] for all ratios. The ratio tracking error is very small compared to conventional CVT systems. Experimental evaluation of the Empact CVT at the test rig showed that the average power consumption of the Empact CVT on the FTP72 cycle is 155 [W], whereas conventional hydraulically actuated CVTs consume over 400 [W] on the average at this drive cycle. Efficiencies of 90 [%], which is close to the maximum efficiency of the Empact CVT, are reached during these experiments. Evaluation of the Empact CVT in an Audi A3 2.0 FSI shows similar performance. Overall, an efficiency improvement of up to 10 [%] is obtained with the Empact CVT compared to a comparable size hydraulically actuated CVT.
Original languageEnglish
QualificationDoctor of Philosophy
Awarding Institution
  • Mechanical Engineering
  • Steinbuch, Maarten, Promotor
  • Vroemen, Bas, Copromotor
Award date12 Apr 2007
Place of PublicationEindhoven
Print ISBNs978-90-386-0939-3
Publication statusPublished - 2007


Dive into the research topics of 'The empact CVT : dynamics and control of an electromechanically actuated CVT'. Together they form a unique fingerprint.

Cite this