In recent years, due to a need to reduce emissions, the automotive industry has focused on increasing vehicle efficiency. One of the areas being examined for potential improvement is the automatic transmission; specifically, the torque converter clutch damper. The better the performance of the damper, the more time the torque converter can be kept in the optimum locked position, thus increasing vehicle efficiency. Currently a large number of vehicle manufacturers use transmission technology sourced from external OEMs; due to a lack of available performance data or validated simulations, sometimes vehicle manufacturers are not able to fully understand the behaviour of the damper. If damper performance (or interactions with other components) cannot be fully assessed during the design development phase, key issues may become known too late in the development process. Thus a deeper understanding of the processes of experimentally characterising and simulating torque converter dampers is required. This thesis describes the development of an arc spring torque converter damper simulation, including the gathering of the experimental data required to validate the simulation. The simulation is used to draw conclusions on the impact of excitation signal form on damper behaviour, leading to new knowledge on the signals required to experimentally characterise a damper.
In this thesis a methodology for (and implementation of) the characterisation of torque converter dampers is detailed. It was found that existing available technologies (e.g. fired engines, electric dynamometers) were either too inflexible or prohibitively expensive; thus a novel high frequency mechanical pulsation generator was developed. This solution was developed from a 4 cylinder motored diesel engine; the cylinders are filled with compressed air and the crankshaft driven using an electric dynamometer. Simulation and experimental data has confirmed that mean torque can be controlled using the input dynamometer, with the compressed air producing fluctuations of up to 900Nm amplitude. However, it was found that the frequency of the output pulsations varied from a fired engine; this is due to reactions between the pulsation generator and the stiffness and inertias of other components on the rig. A review of the performance of the novel pulsation generation concept against other damper excitation methods was also conducted. It was determined that fired engines and electric motors are more suitable for durability testing; the flexibility of the electric motors and the low running costs of the pulsation generator suit damper performance tests.
The second phase of this project was to develop a simulation of a two-stage arc spring turbine damper. This damper consists of three inertias, separated by two spring sets; the outer spring set has 3 individual arc springs, while the inner spring set has 5 nested pairs. The principle of conservation of angular momentum is applied to each of the three inertias in order to calculate their individual accelerations. This method is also applied when calculating the acceleration and movement of the springs; the arc springs are discretised into mass and (massless) spring segments. Two features not previously seen in literature are included in the simulation; hardstops and nested springs. The physical hardstops limit the movement of the spring sets (relative movements of the inertias). In this study, the nested springs were simulated as a pair of parallel springs, rather than as a single stiffer arc spring; this is due to the friction that occurs between the springs (the inner race of the larger spring forms the housing for the inner spring). These two features highlight the need for hardware examination before simulation development; disassembling the hardware also allows the location of hardstops (and other features) to be measured rather than relying on the test data.
Once a damper simulation was designed, a methodology for simulation parameterisation was required; parameterisation is the process of improving simulation performance through iterations of estimated parameters. The simulated damper was excited using sampled experimental data; to maximise parameterisation process efficiency, each time a parameter change was made, a set of key test points were selected in order to assess simulation performance change. It is not recommended that single test points be examined individually; parameter changes may improve simulation performance at one test point but have an adverse reaction at another.
A clear causal relationship between simulation timestep and accuracy (as well as simulation run time) was found; a link between the number of discretised segments and simulation accuracy (and run time) was also confirmed. It was determined that 8 segments was optimal for the inner springs and 18 outer segments offered the best balance between computing power and simulation time. A variety of methods for analysing damper (and simulation) performance are presented in this thesis; it was found that for the 2.5 bar torque curve experimental data set the simulation performs excellently, with on average less than 5% error. Overall torque error is less than 10% across the tested speed range (900 to 2800rpm), with mean torque differences between simulated and tested order magnitudes of less than 5Nm. It has been determined that hysteresis loops are not an accurate predictor of real-world damper performance; while they can approximate general trends, they do not cover the normal operating condition.
In the final phase of this thesis, the validated simulation has been used to investigate excitation signal, areas of poor damper performance and the link between speed and damper stiffness. By subjecting the simulation to a variety of sinusoidal input signals, it was established that if a sinusoidal signal approximates the 3 most dominant frequencies in a real signal, the damper will behave in a representative manner. Additional orders that have lower frequencies than the dominant order will have a greater impact on the attenuation behaviour of the damper; the effect of additional orders on attenuation behaviour is also linked to their magnitude (relative to the dominant order).
A methodology for efficient damper mapping is proposed; the key aim is to produce a dataset that will minimise the length of the parameterisation process while capturing key damper behaviours. It was found that the magnitude of the torque oscillations used to excite the damper is linked to parameter adjustment impact, though this relationship is not linear for all parameters; an approximate level of 300Nm should be used for excitation. Parameters such as spring stiffness and plate inertias are more likely to have a substantial impact on damper performance at frequencies below 70Hz; friction tuning factors are impacted more by magnitude changes at frequencies above 150Hz. It has been demonstrated that while speed can have an effect on magnification ratio, this effect is far less significant at mean torques above the knee point and when sinusoidal input magnitude is kept at or above 300Nm. It was concluded that neither engine speed nor precise excitation magnitude must be replicated in order to predict approximate performance.
During the investigation into areas of poor damper performance, it was confirmed that the trend of increasing magnification ratio with lower frequencies (<30Hz) seen in experimental data continued. Simulation testing above 140Hz revealed that there is not a linear relationship between increased frequency and increased magnification ratio; these areas of magnification ratio spikes are likely due to system resonances. It has been confirmed that while fluctuation magnitude does impact magnification ratio, fluctuation frequency has the most significant (dominant) impact. Finally, the effect of speed on apparent damper stiffness was investigated for both hysteresis loop testing and across a range of outer spring vibration angles; it was confirmed that increasing speed does result in non-homogeneous compression of the springs. It was established that while speed can have an effect on spring stiffness, this effect will vary significantly depending on the movement range (vibration angle) of the spring. The largest increase in spring stiffness with speed is seen when segments of the spring become inactive (cease to move), hence why the effect of speed is more substantial at vibration angle of <10°. The simulation was used to confirm the theories linking speed and stiffness found in the literature; higher speeds increase frictional forces, slowing damper segments, resulting in reduced movement.
The findings of this thesis are relevant to damper simulation and testing engineers; by expanding knowledge of damper behavioural responses to high frequency excitation signals, as well as demonstrating an effective method for producing validated damper simulations, it is hoped that the vehicle design process will be more efficient and damper modifications more effective.
|Date of Award
|27 Jun 2017
|Chris Brace (Supervisor) & Sam Akehurst (Supervisor)