In an effort to model a complete engine cooling module,. Valeo has developed static and dynamic FEA models for its engine cooling fans. The static FEA ...
2000-01-0968
FEA Computations Applied to Engine Cooling Fans S. Moreau Valeo Engine Cooling, La Verrière, France
P. Boulanger Valeo Electronics, Créteil, France Copyright © 2000 Society of Automotive Engineers, Inc.
ABSTRACT In an effort to model a complete engine cooling module, Valeo has developed static and dynamic FEA models for its engine cooling fans. The static FEA analysis has led to a significant fan weight optimization for constant Van Mises stresses and ring deflection (a 15% reduction). Both experimental and numerical results have stressed the importance of the model of the interface between the motor and the fan in the dynamic FEA modal analysis. Moreover, the main engine cooling fan modes have been identified for the first time.
subjected to centrifugal loading. In all cases, we have found that the aerodynamic loading can be neglected compared to the centrifugal inertial loading. The material, mainly nylon 6.6, has been assumed linear.
INTRODUCTION The development of new engine cooling fans is now requiring more and more mechanical computations. Indeed, the fan structural analysis will directly impact several customer needs such as : ♦ Improvement of comfort ; ♦ Noise reduction ; ♦ Energy and gas consumption reduction. The first two needs are related to the filtering of vibrations by either the knowledge and the modification of the fan natural frequencies, or the use of vibrational dampers. The last objective is referring to the mass optimization of the fan system that involves the reduction of its thickness. Finally, as shown in [1], Valeo wants to be able to predict the global dynamic properties of its engine cooling module: the present fan models will be building blocks in the knowledge of the dynamic behavior of the module separate components. NUMERICAL MODELS Linear and geometrical non-linear FEA analysis have been performed on several fans subjected to their design aerodynamic loading and centrifugal inertial loading at their design rotational speed. We have presently focused on a 9 blades fan termed fan1 and a 5 blades fan termed fan 2. Several commercial codes including ABAQUS, NASTRAN and ANSYS have been used [2]. Figure 1 displays a typical stress field of fan 1
Figure 1: Stress Field at the design conditions for Fan 1 The influence of the grid on the results has been carefully studied, which provided us with recommendations on the grid type and resolution. The comparison was done on two types of grids with the same loading and boundary conditions: 1) The fan grid is done with only quadratic tetrahedral elements (10 nodes) yielding about 400,000 degrees of freedom (dof). 2) The fan grid is done with quadratic shell elements (8 nodes) for the ring, the blades and a part of the hub and quadratic tetrahedral element (10 nodes) for the insert and the central part of the hub; the interface between the two types of elements is assumed by shell elements that preserve all dof (rotation and translation). The fan blades have been simulated with shell elements with variable thickness, whose distribution is given around its neutral fiber. The comparisons in term of grid time, calculating time and validity of the results show that the second grid is the most efficient of both. It has been used for both static and dynamic FEA analysis presented below.
STATIC FEA COMPUTATIONS OBJECTIVES We have first focused on the fan weight optimization of fan 1 to make direct progress in the fuel consumption reduction. Efficient control over the fan deflection will also contribute to a more efficient system by limiting flow leakage at the tip clearance and can yield a possible noise reduction. Within this framework we have also studied the impact of the fan sweep on this deflection by comparing three different sweep angles for fan 1: EC1 version: 18°, EC2 version: 23° and EC3 version: 42°. EXPERIMENTAL VALIDATION The correlation with the FE model is done by measurements of stresses and displacements with strain gauges and fast speed camera. For the first conditions shown in figure 2, camera measurements have shown a maximum forward deflection of about 2 mm. A precise measurement is made difficult by the fan wobble. Numerical results yield the same forward movement of 1.5 mm, which is in good agreement with experiment.
significant decrease of fan rigidity and a much larger ring deflection. Interestingly, the location of the maximum constraint will change from the hub EC1, to the ring EC2 and to the blade EC3. Fan EC1 EC2 EC3 Table
Max. Constraint (MPa) Max. Displacement (mm) 16.1 4.17 16.51 4.07 21.83 4.83 1: Effect of Sweep on the Static Analysis of Fan 1
DYNAMIC FEA COMPUTATIONS OBJECTIVES We have then focused on the dynamic behavior of fans. Several modal analyses have been performed to yield the fan natural frequencies and its corresponding eigenmodes at the design conditions (fan rotating) or at rest (no speed). EXPERIMENTAL VALIDATION The analysis was intended to be on a rotating fan mounted on an electric motor. Yet, for validation purposes, experimental static modal analyses have been performed with two types of boundary conditions: a “free” system (no motor) and a fan attached to its production motor. Two types of experiments have been performed and compared. On the one hand (exp 1), vibration measurements were measured with an accelerometer set at each grid node of the fan defined in figure 3 by colored stickers (about 70 points). Because the fan is very flexible, the excitation could not be done with a hammer but instead at a single point with a shaker. The grid was fine enough to yield the first four modes below 200 Hz.
Figure 2: Aero-acoustic properties before and after mass optimization for Fan 1 RESULTS The fans are optimized on the one hand by reducing the thickness of the hub, the blade sections and the rotating ring, on the other hand by improving the position of the ribs inside the hub, while keeping the same level of Von Mises stresses and deflections. The original aerodynamic, acoustic and endurance properties will then be preserved as shown above in figure 2. A 15% mass reduction was achieved this way. It should be noted that the mass gain in these conditions has also a positive influence on the dynamic behavior of the fan. As shown in table 1, blade sweep will slightly decrease the deflection of the EC2 fan while keeping similar maximum constraint. Yet a larger sweep triggers a
Figure 3: Classical modal analysis fan grid.
On the other hand (exp 2), a laser vibrometer has been used to locally measure displacement and velocity to yield natural frequencies. Moreover, at the detected critical speeds, holograms have been registered with a continuous laser to visualize the nodal lines: in the images, white zones correspond to nodal lines and black zones to moving parts. The excitation is performed with a piezoelectric shaker aligned with the rotational shaft as shown in figure 4. For the fan alone (4.a), the excitation is directly on the fan shaft, whereas for the fan system it goes through the electrical motor (4.b). RESULTS For the “free” system, good agreement was achieved between the different experiments and the FEA results as shown in Table 2. The only mode missed by the laser vibrometer set-up is the second flexion mode. Modes 1 2 3 4
Exp.1 (±5%) ABAQUS ∆(exp1-cal) Exp.2 (±1%) 61.5 53 -13.8% 59.3 89.8 67 -25.4% 88.3 112.6 82 -27% 136.8 139 +1.6% 131 Table 2: Static Modal Analysis on Fan 2
The mode shapes for the first flex mode and the axial or “pumping” mode (the rotating ring and the blades are moving back and forth along the rotational axis w.r.t. hub), as computed by ABAQUS and the ones measured by interferometry, are compared in figures 5 and 6. A single nodal line can be seen along the fan diameter for st nd the 1 flex mode. In the 2 flex mode, two nodal lines are found along two perpendicular diameters. The bending, torsion and “pumping” modes are therefore well predicted. Moreover, in the numerical simulations, a strong influence of the boundary conditions at the attachment between the fan insert and the shaft has been found as shown in Table 3 (columns 2 and 4); BC1 only blocks the nodes in the cylindrical part. BC2 adds the nodes of the insert in contact with the nut and the flats. It should first be noted that all codes yield very similar results as shown in columns 1,3 and 4 of Table 2 (difference less than 1% between ABAQUS and ANSYS for instance). The (D) stands for a double mode. The more realistic rd BC2 suppresses the 3 mode at 69.5 Hz. Mode 1 2 3 4 5 6 7 8 9 10
ABAQUS ANSYS (BC1) NASTRAN ANSYS (BC2) 46.7 (D) 44.0 (D) 45.3 (D) 46.7 (D) 46.7 (D) 44.0 (D) 45.3 (D) 46.7 (D) 84.3 (D) 69.5 83.6 84.3 (D) 84.3 (D) 84.2 (D) 84.8 (D) 84.3 (D) 84.5 84.2 (D) 84.8 (D) 84.5 124.1 90.6 125.8 124.3 183.3 (D) 182.8 (D) 184.9 (D) 183.3 (D) 183.3 (D) 182.8 (D) 184.9 (D) 183.3 (D) 196.1 (D) 192.0 (D) 194.9 (D) 196.1 (D) 196.1 (D) 192.0 (D) 194.9 (D) 196.1 (D) Table 3: Dynamic Modal Analysis on Fan 1
In the second experimental set-up we have compared the fan system with the fan alone. Table 4 clearly shows that the natural frequencies corresponding to the flexion modes and the torsion modes are no longer detected. The “pumping” mode is now at a higher frequency, about 98 Hz and the only remaining natural frequency at about 177 Hz is a much more complicated bending mode presenting a nodal diameter at about 60% of the blade span, as seen in figure 7. The double-D interface between the fan insert and the motor shaft is most likely the main cause of the observed differences. Modes Fan Alone Fan+Motor Type st 1 59.3 1 Flex 2 88.3 98.6 Pumping 3 131 177.3 Bending Table 4: Modal Analysis on Fan system 2 Finally, the effect of blade sweep on the natural frequencies of fan 1 has been studied (table 5). Fans EC1 and EC3 have the same modes in the same order. st For the EC2, the 1 flex mode increases and becomes the third mode: in this particular mode, the rigidity increases due to the blade shape. But overall, all natural frequencies decrease with increasing blade sweep and the mode shapes are preserved. Modes 1 (D) 2 (D) 3 4 5 (D) Table
Type EC1 EC2 Wobbling 49.9 42 st 1 Flex 76.5 79.4 Axial 79.4 76 Bending 112.9 100.3 nd 2 Flex 173.8 170.8 5: Effect of Sweep on Modal Analysis
EC3 41.7 69.8 71.2 98.6 151.9 of Fan 1
CURRENT LIMITATIONS AND FUTURE DEVELOPMENTS The strong anisotropy of the fan material (typically a glass-filled nylon with long fibers) due to the injection process is not yet well known and has not been taken into account in the present simulation: we have always assumed linear properties with averaged Young’s modulus and Poisson’s coefficients for a given fan (experimentally measured on the molded part) and a given temperature. Future developments will include taking into account the non-linearity of the material due to injection molding and the viscoelastic properties of the resin with temperature to simulate the underhood environment more closely. Further grid study will include putting at least three layers of tetra grid elements in the blade thickness: to solve the resulting huge problem (> 1 million dof) some condensation method as described in [1] might be useful. Finally, experimental modal analysis will be further pursued to reduce the discrepancy between the different
experimental and numerical results. The absence of the second flex in fan 1 with the vibrometer laser will also be investigated.
found both experimentally and numerically. Only a proper and accurate model of the former can yield meaningful results in terms of natural frequencies and mode shapes.
CONCLUSION ACKNOWLEDGMENTS Engine cooling fan structures have been characterized both statically and dynamically. A shell model is currently the most reliable model for the fan blade. Centrifugal forces have been found to be the dominant load on the blade whereas the aerodynamic loading can be neglected.
Special thanks go to HOLO3 for performing the modal analysis laser measurements and to Teuchos and Calass for performing some of the numerical computations. REFERENCES
Static studies have yielded scaling laws between the fan ring deflection and the fan diameter, which needs to take into account the blade sweep. Indeed, large increase of sweep angle has been found to trigger larger constraints and axial displacements. A first mass optimization at constant maximum Van Mises constraints and fan ring deflection has provided a 15% weight gain.
[1] “Virtual Mock-up of Heat Exchangers in Mechanics,” D. Le Hegarat, SAE 1999-01-0581, Detroit 1999. [2] ANSYS reference manual.
The modal analysis has shown three main modes below 200 Hz for engine cooling fans: an axial or “pumping” st mode, a 1 flexion mode with a fan diameter being a nodal line and a bending or torsion mode. Yet, the shape and the tolerances of the interface between the fan and the electrical motor strongly modify the modal response of the fan system compared to the fan alone, as it was
(4.b)
(4 a)
Mirror
Mirror
Fan Fan
Fan System
Axis Piezoelectric Shaker Magnetic Base
Piezo Shaker Magnetic Base
Figure 4: Holographic Modal Analysis: (4.a) Fan Alone; (4.b) Fan + Motor
Figure 5: Experimental and Numerical Flexion Mode Shapes for Fan 1
Figure 6: Experimental and Numerical Pumping Mode Shapes for Fan 1
Figure 7: Fan System First (98 Hz) and Second Modes (177 Hz)
