Gépészet | Gépgyártástechnológia » Kelly-Kent - Gear Shift Quality Improvement In Manual Transmissions Using Dynamic Modelling

Session Home F2000A126 Seoul 2000 FISITA World Automotive Congress June 12-15, 2000, Seoul, Korea Gear Shift Quality Improvement In Manual Transmissions Using Dynamic Modelling David Kelly Christopher Kent Ricardo The importance of the gear shift quality of manual transmissions has increased significantly over the past few years as the refinement of other vehicle systems has increased. The synchroniser is often blamed as the cause of many shift quality issues. This is not always the case The interaction of the entire selector system from the shift fork to the handball, the driveline and the transmission internals all play a part in the overall shift quality. The dynamic interaction of these systems at a component level is difficult to interpret by traditional test methods and virtually impossible at concept stage. To overcome these difficulties a dynamic model of the entire synchroniser, selector mechanism, driveline and transmission has been created. The model predicts the gearshift

quality for a given set of input parameters, which can be correlated against test data. The model can then be used for parameter studies to investigate potential improvements to gearshift quality The model can also be used at the concept stage to indicate suitable specification for synchroniser geometry, component stiffness, mass and inertia. This paper will present experience of gearshift problems using a dynamic model, the structure of the model and the methodology used. Keywords: Shift Quality, Dynamic Modelling, Manual Transmission components. This analysis may be best achieved by using a co-simulation approach where different analysis tools are used for gear shift and suspension modeling. INTRODUCTION Improvement of gearshift quality of manual transmissions has become more prevalent over the past few years as the refinement of other vehicle to driver interfaces has increased. Ricardo has many years experience in investigating gearshift quality problems and has created software

and hardware for the measurement and analysis of gearshift quality. This system, the Ricardo Gear Shift Quality Assessment system (GSQA), is sold commercially and currently has 22 users world-wide ranging from vehicle manufacturers to transmission suppliers. Gearshift quality is made up of several different areas, including gate definition, shift effort, second load, and vehicle response. Gate definition requires careful design of both the internal and external selector mechanism, paying attention to the compliance, static load, backlash, friction, cross gate loads and gear positions. Shift effort requires careful sizing of the synchroniser cones, cone angle, friction material and balancing this with the transmission reflected inertia, clutch inertia and transmission drag. To complement the GSQA system Ricardo has developed a series of dynamic models to investigate the dynamic effects of gearshift quality at the design stage and for specific gearshift quality development. These models

are normally generated in MATLAB / Simulink although a model has also been generated in ADAMS. This paper will concentrate on experience creating models in MATLAB / Simulink for second load issues. GEARSHIFT QUALITY DYNAMIC MODEL Ricardo has created several gearshift quality dynamic models over a five-year period. These have included synchroniser only models, synchroniser and selector models, and synchroniser, selector and driveline. The modelling approach has progressed to the current position where the entire selector, transmission, driveline and synchroniser are modelled in considerable detail. The model is written in MATLAB / Simulink and is built up from low level components such as gains, integrators and other mathematical functions. The system is broken down into relevant degrees of freedom for each of the major subsystems. These sub-systems are the external selector mechanism, internal selector mechanism, transmission, driveline and the synchroniser. The model takes

each of the degrees of freedom and solves for the acceleration of each component either axially or rotationally. The acceleration can then be integrated to determine the velocity and displacement of each component. The velocity and displacement terms are used to generate the Gate definition is a static or quasi-static function and can be defined by careful design and selection of relevant parameters. Shift effort is a quasi-static function, which can be defined by consideration of quasi static equations [1]. Second load however is a dynamic event and depends on the response of the entire system, including the selector mechanism, transmission, driveline, and synchroniser system. Vehicle response requires a full transmission and driveline model coupled with a vehicle model which includes engine and transmission mounts and suspension 1 force or torque acting between components. Backlash and geometrical constraints are also taken into account. The external selector mechanism includes

the shift lever, cables or rod, and any additional selector components outside the transmission. The masses, inertias, mechanical advantage, drag, damping, stiffness and backlash present in the system are modelled. For small angles: Fmesh = K mesh * ( PCR gear 1 theta gear 1 + C mesh * ( PCR gear 1 dtheta − PCR gear 2 * dtheta gear 1 − PCR gear 2 * theta gear 2 ) gear 2 ) Tgear1 = Fmesh * PCRgear1 The internal selector mechanism components typically include a selector rod, selector finger, selector rail, detents, selector fork and sleeve. Again all the relevant parameters that describe the system are used in the model generation. Tgear 2 = Fmesh * PCRgear 2 Where : = Force generated at mesh point (N) Fmesh = Stiffness at mesh point (N/m) Kmesh = Damping at mesh point (Ns/m) Cmesh thetagear1 = Angular displacement of gear1 (rad) dhetagear1 = Angular velocity of gear1 (rad/s) thetagear2 = Angular displacement of gear2 (rad) dhetagear2 = Angular velocity of gear2 (rad/s)

PCRgear1 = Working pitch circle radius (m) PCRgear2 = Working pitch circle radius (m) = Torque applied to gear1 (Nm) Tgear1 = Torque applied to gear1 (Nm) Tgear2 The transmission model varies depending upon the transmission layout. A typical front wheel drive system may include the clutch including torsional damper, input shaft, synchronised gear and output shaft. The driveline model also varies depending upon the application. Models of front wheel drive, rear wheel drive and four wheel drive have previously been developed. A typical front wheel drive system may include differential housing, differential bevel gears, drive shafts and wheels. A vehicle model may also be included. The synchroniser system employed in transmission systems varies greatly in configuration. These systems vary from single to multiple cones, asymmetric teeth, and location of blocker and engagement teeth. The synchroniser model includes provision for up shifts, down shifts asymmetric teeth and non-linear cone

friction coefficients. The degrees of freedom modelled include individual degrees for each of the synchroniser cones, a degree of freedom for the sleeve and hub, and a degree of freedom for the gear. The synchronisation torque is generated from the axial force applied to the blocker ring from a combination of force applied by the pre-energisation strut and the force generated during sleeve to blocker ring tooth contact. This total axial force acts to resist the sleeve axial motion and is also converted into cone and index torque. Once the two sides of the synchroniser approach synchronisation the index torque (derived from the tangential component of the blocker axial force) is greater than the cone torque. This causes the blocker ring to rotate out of the path of the sleeve allowing the sleeve to travel forward and approach the engagement teeth. The blocker force is calculated from the normal displacement of the blocker and sleeve teeth. This normal displacement and the corresponding

normal velocity are used to calculate the normal force which is resolved into its axial and tangential components. Figure 1 Simulink Model Top Level Schematic Two gears in mesh The torque between two gears is obtained by calculating the relative displacement at the gear mesh point. The angular displacement of the gears is converted into an linear displacement and the angular velocity is converted into relative linear velocity at the gear mesh point. The relative angular velocity and displacement are converted into a force at the mesh point. This force can then be converted to a torque and applied to each gear by the multiplication of the mesh force by the pitch circle radius. PCRgear1 PCRgear2 Figure 2 Two gears in mesh. 2 Blocker Force Generation The normal force is resolved into axial and tangential components. Pitch Faxial = Fnormal * (sin γ + sign(dxsleeve ) µ teeth cos λ ) Ftan gential = Fnormal * (cos γ − sign( dxsleeve ) µ teeth sin γ ) θ modified where

θ index X2 γ Blocker dxsleeve = sleeve axial velocity (m/s) µteeth = teeth contact friction coefficient The first contact between the blocker and sleeve teeth is governed primarily by the index angle which is a function of the synchroniser design. The angle through which the sleeve can rotate in relation to the blocker teeth is governed by the axial position of the sleeve. X1 sleeve Neutral Fnormal = K normal * xnormal + C normal dxnormal The impact force between the sleeve and gear engagement teeth is modelled in a similar manner to that of the blocker teeth. However the speed difference between the gear and synchroniser hub are much greater at the start of the shift and as there is no physical constraint between the two components the relative displacement can be very large resulting in incorrect contact displacements. To overcome this a remainder function is used which calculates the position of one sleeve tooth in relation to two adjacent gear engagement teeth. where:

Use of the model Figure 3 Blocker to Sleeve Contacts The normal blocker force is calculated from the relative angular position of the blocker ring and sleeve. Knormal = Normal contact stiffness (N/m) Cnormal = Normal damping coefficient (Ns/m) dxnormal = Normal contact velocity (m/s) and = (( abs ( θ x normal * PCR blocker sl − θ br ) − θ modified 2 * cos γ ) where: θsl = Sleeve angular displacement (rad) θbr = Blocker ring angular displacement (rad) PCRblocker = Pitch circle of blocker ring teeth (m) γ = Blocker to sleeve teeth contact angle (rad) and θ mod ified = θ index − 2 * (( x1 − x 2 ) * tan γ ) PCRbloc ker where: θindex = index angle x1 = sleeve travel from neutral to first blocker teeth contact (m) x2 = sleeve travel from neutral to instantaneous position (m) The model can be used in a number of ways. The first example is to investigate a current gearshift quality problem e.g large second load, nibble, partial clash For this type of

investigation an objective gear shift quality assessment would be performed to analyse the problem ) when the vehicle is driven under normal operating conditions. This would give an experienced transmission engineer an insight into the problem but it can be difficult to pinpoint the causes. Any potential improvement has then to be tested on a vehicle. This can be costly and time consuming. The problem may be related to more than one area and the interaction of different parameters may be overlooked. A dynamic model of the entire system can be used to identify potential problem areas allowing quick and cost-effective investigations of potential solutions both in isolation and their interaction with other parameters. To perform parameter studies it is preferred to start with a correlated model. Correlation of a model poses several problems. The first problem is the driver A driver can shift in a subconscious closed loop manner where he/she modifies the force and velocity profile exerted

on the shift lever based upon the feedback at the lever. A skilled driver may be able to find a problem with every shift or avoid the problem entirely. It is also very difficult to accurately model a driver. For these reasons an open loop approach 3 The entire gearshift process is the combination of several stages. Using dynamic modelling it is possible to understand how each sub-system performs and how a subsystems individual performance affects other systems. to this problem has been adopted. The test method uses a known velocity profile electro-servo actuator which acts on the gear lever inside the vehicle through double acting springs. The velocity of the actuator and the spring rates can be modified to give variable peak input force levels to the system. The vehicle is motored on a chassis dynamometer to give a repeatable vehicle velocity for the test. The actuator shifted data is compared with the hand shifted data to ensure a high degree of correlation. Spring rates and

actuator velocities can be modified to achieve representative shifts. As the shift process is random 50 shifts are typically taken for each shift. The Ricardo GSQA system is used to log the handball position, handball force, transmission input and output speeds. 50 shifts are performed at three force levels and at three different vehicle speeds resulting in a total of 450 gearshifts for each shift type. The model results can then be correlated to the test data. As the model is numerical it will give the same results for a given set of parameters. To overcome this variability is introduced into the model to simulate the effects of randomness. This is achieved by adjusting the relative position of the sleeve and engagement teeth at the start of each simulation. This is performed 15 times for a given peak input force and vehicle speed with the relative position varying between the pitch of the engagement teeth. Correlation allows the tuning of the model to take into account unknown

parameter data such as damping etc. Once correlation has been achieved, problem areas can be investigated in detail, looking at the interactions between components and the causes of specific events. Potential solutions can be assessed for cost, practicality and then simulated. A pre-processor has also been created which allows batch running of multiple parameter changes for investigations of the performance of potential solutions. The following paragraphs refer to figures 4 & 5, and explain what problems can arise at each stage of the gearshift process. This example is for an up-shift Figures 4 & 5 both show three subplots of a 4th to 5th gearshift at 140kph for a front wheel drive car as predicted by a dynamic model . The top display shows the gear and input shaft velocities (rads-1) Vs time (s), the centre display shows the handball force (N) Vs time (s) and the lower display shows synchroniser sleeve axial position (m) Vs time (s). stage 1: Out of gear taking up the

backlash in the system, the sleeve moves forward towards next gear. The sleeve velocity is reduced through drag and friction in the selector system and detent loads. stage 2: The sleeve comes out of gear and drag begins to take effect on the clutch side of the synchroniser and begins to reduce the velocity of the upstream components. stage 3: The sleeve picks up the pre-energisation strut, which has the effect of wiping oil from the cones. The axial force begins generates cone torque, which results in a change in gradient of the gear velocity. The blocker ring is rotated to the indexed position. These events happen prior to blocking to prevent push through clash while the friction coefficient is low due to the oil film between the cones. stage 4: The sleeve and blocker teeth contact and as there is a speed difference between the cones the sleeve cannot push though towards the engagement teeth. The handball force builds up during synchronisation. The output shaft velocity trace shows an

increase in velocity as the driveline is accelerated by the cone torque causing the driveline components to rotate from the coast to the drive flanks. The velocity stabilises but there is a level of torsional windup in the driveline. The level of wind-up in the driveline is dependant upon the level of torque generated during synchronisation. The gearshift quality model can also be used for initial concept design. The basic geometry of the synchroniser, the target selector masses, stiffness and backlash, the driveline stiffness, backlash and inertia can be predicted to give an indication of how a gearshift system may perform. The model can be updated when real information is available, throughout the concept and development phases of a project right through to production intent sign off. Another use of a shift quality model is to predict how an existing transmission may perform in a different vehicle, driveline or selector mechanism application, or to identify potential cost savings.

PROBLEMS DURING THE GEARSHIFT PROCESS 4 stage5: In this particular example the static friction coefficient is lower than that of the dynamic and as synchronisation is approached the cone torque drops below the index torque and blocking release begins. The blocker ring is indexed allowing the sleeve to move through the blocker ring towards the engagement teeth. Once blocking release has occurred, drag can act on the upstream components introducing a speed difference across the synchroniser with the gear velocity dropping below that of the output shaft. The sleeve travels towards the engagement teeth of desired gear. The selector mechanism has been compressed during the synchronisation process storing energy in the system. This is released and the sleeve travels towards the selected gear at a greater velocity than the input to the system. The driveline also has stored energy in the system and begins to unwind. The combination of these three phenomena contribute to second load

problems ranging from double bump (single second load) to nibble (multiple second load with teeth passing). As the sleeve moves forward its stored energy is reducing. It is possible for the sleeve to stop once this energy has been expended. The sleeve remains stationary until the rest of the selector mechanism has moved sufficiently to close the backlash in the system. The effect of drag post synchronisation is a reduction in gear speed below that of the secondary shaft. The extent of desynchronisation depends on the drag and inertia of the system and also the length of time the gear remains unconstrained to the sleeve. Another problem arises postsynchronisation As the driveline unwinds the output shaft velocity reduces and then increases with an oscillatory response. The sleeve and gear engagement teeth impact resulting in the sleeve moving away from the gear, force being transmitted to the handball and an additional speed modification. This particular example shows a single second

load spike. The sleeve is not forced away form the gear engagement teeth past the point of tip to tip contact. The sleeve is then pushed though to final engagement. The relative angular position that the engagement teeth impact each other post synchronisation is random with differing effects depending on which flank is hit and the direction of the relative motion. stage1 stage2 stage4 stage5 stage3 Figure 4 Model results for a 4th to 5th gearshift at 140kph Gear and input shaft veloity Vs time Speed (rad/s) 345 340 335 330 325 0.58 0.6 0.62 0.64 time (s) 0.66 0.68 0.7 0.72 handball force Vs time Sleeve axial position (mm) Handball force (N) 100 There are therefore several potential problem areas, which must be overcome. The driveline unwind as the driveline is unloaded must not happen at a point where the engagement teeth can contact. The driveline torsional effect must also be small. The sleeve should travel towards the engagement teeth as fast as possible so as

to minimise the effect of transmission drag. The sleeve must not stop part-way into engagement while the selector mechanism catches up with the sleeve. 80 60 40 20 stage5 0 0.58 0.6 0.58 0.6 0.62 0.64 0.66 time (s) Sleeve axial position Vs time 0.68 0.7 0.72 0.68 0.7 0.72 20 15 10 5 0 0.62 0.64 time (s) 0.66 Figure 5 Zoom on stages 4 and 5 of Figure4 5 Test data 4th to 5th gearshift at 100kph 60 Force (N) 40 20 Input shaft and gear velocity (rpm) 0 0 0.1 0.2 0.3 0.4 0.5 time (s) 0.6 0.7 0.8 0.9 0 0.1 0.2 0.3 0.4 0.5 time (s) 0.6 0.7 0.8 0.9 3000 2800 2600 2400 2200 Figure 6 Example of test data CONCLUSIONS Past experience has shown good correlation between model parameter modifications and vehicle tests. Modifications have ranged from increasing sleeve mass, reducing shift fork stiffness and improvements of spline qualities. Additional improvements can be made if improvements to the driveline are performed. Gearshift dynamic modelling

can give a much clearer picture of how the physical interactions of transmission components influence the shift quality. ACKNOWLEDGMENT The author wishes to thank the directors of Ricardo for allowing the publishing of this paper. REFERENCES [1] Socin R.J and Walters LK 1968 Manual Transmission Synchronizers SAE 680008 [2] Sykes L.M 1994 The Jaguar XJ220 Triple Cone Synchroniser - A case study SAE 940737 6