Nipping valve-train noise in the virtual bud

The valve train is probably one of the most complex and critical assemblies in a modern automobile engine. Its operation directly impacts engine efficiency and emissions. Engineers with the General Motors Power Train Group, Pontiac, Mich., used to model valve-train kinematics and dynamics with homegrown, 2D programs, which was OK for simple, planar motion. Obviously, such programs won't work for 3D lateral forces associated with more advanced designs, nor can they predict nonlinear behavior in helical springs.

Overcoming these limitations is the 3D multibody dynamic code, LMS DADS from LMS International, Leuven, Belgium (www.lmscadsi.com). GM engineers used the program to eliminate socket hop in a concept V6 engine valve train before building prototype engines, saving considerable development time and the associated costs. Socket hop is defined here as a momentary loss of contact between a rocker arm and lash adjuster. Such disturbances generate noise and vibration and can rob power, especially at high engine speeds.

The concept-turned-production 3.5-liter Premium V6 (PV6) 24-valve, dual-overhead cam engine powers the Oldsmobile Intrigue. Valves actuate with separate finger follower rocker arm assemblies. Each finger follower has a socket at one end mated to a spherical head on a self-adjusting hydraulic lash adjuster. The socket also acts as a pivot point. A roller shaft/needle bearing assembly located about midspan rolls on the cam lobe surface. On the opposite end of the follower is a valve pallet. It contacts the stem top of an intake or exhaust valve. The valve pallet is radiused to prevent edge contact with the valve tip.

The arrangement generates less noise than earlier direct-acting valve trains but also introduces several complexities. Of these, and probably of most concern, is socket hop. Creating an accurate model of the hardware was the first step to eliminating the problem.

A MODELING JOB
The DADS software models 3D systems or imports them from major CAD packages, such as Catia, Pro/Engineer, and I-DEAS. Engineers define joints, constraints, and forces on the system. Then, DADS automatically derives and solves the nonlinear equations of motion and reports loads, positions, velocities, and accelerations at each time step. Results are graphed or output to photorealistic 3D animations that help visualize flexible deformation of engine components in motion.

The PV6 DADS model simulates all three cylinders on the right bank. A total of 38 flexible bodies are modeled including camshafts, intake and exhaust valves, rocker arms, and helical valve springs, each with multiple mode shapes.

Cam lobe profiles are represented by a series of fifthorder spline functions fit to a table of points of lift-versuscam angle. The spline-fitted surfaces determine penetration depth, follower linear velocity, and cam/follower sliding velocity. These values feed to a contact algorithm that applies forces to the cam surface and roller bodies based on material stiffness and damping properties and tangential friction coefficients. These calculations run simultaneously for all cam lobes and update each time step.

The model can also predict valve float because the cam and follower are allowed to separate and reimpact. In addition, contact force elements model the interaction between the rocker pallet and valve tip and between the valve head and valve seat insert. A Hertzian contact model computes the surface radii of curvatures and related stresses at the contact points by assuming sufficiently small local surface deformations. The normal contact force is a function of Young's modulus, restitution coefficient, and transition velocity. Tangential contact force uses a friction coefficient or nonlinear function of friction versus tangential velocity.

Valve spring inertial surge and resonance are other important metrics to model, particularly for severe vibrations that can trigger coil-to-coil contact. A helical-spring element within the software accounts for these behaviors, including nonlinear effects. First, a 3D FE mesh is solved in the modal domain and inserted into the larger valve-train simulation.

An optional reduction procedure turns the 3D spring model into a dynamically equivalent and more computationally efficient one-dimensional model. A quasi-static compression procedure retains spring mass distribution, needed to capture surge, while a series of nonlinear coilcontact force relationships emulate contact algorithms of the 3D model. The model also simulates combustion loads acting on valve heads, cam-bearing loads with a hydrodynamic oil film, and dynamic hydraulic pressure inside lash adjuster chambers.

RUNNING THE SIMULATION
The simulation applies a constant angular velocity to a camshaft end. Changing angular velocity simulates different engine speeds. Camshaft rotation velocity, however, is nonuniform because combustion loads on valve heads torsionally flex the camshaft along its length. Plots of hydraulic adjuster force versus cam angle reveal a skewed double-peak feature from a changing rocker-arm ratio. This is because adjuster motion changes the pivot position which, in turn, moves the cam/follower contact point.

Socket hop happens when adjuster force drops to zero along the translation axis. This can cause loss of contact and valve bounce at high engine speeds, setting up vibrations that transmit to the valve head through the valve seat, spring seat, and journal bearings. The culprit, it turned out, was the valve springs. Though details are proprietary, these high-fidelity simulations helped GM engineers tweak the spring design to eliminate socket hop, leading to successful prototypes, and ultimately, production engines.