Initial Invesitgation

An overhung fan at a chemical plant that sits five stories up on a mezzanine was shaking the whole tower structure. The initial investigation was undertaken because the customer felt that excessive vibration of the fan was not only hard on the fan assembly but also detrimental to the integrity of the structure.

Operating Deflection Shape

Since the fan could not be shut down to attach a shaft reference, three reference readings were collected on the inboard fan bearing. The remaining five channels on the tape recorder were used to collect data at other points in the system. A cross-spectrum was calculated for each roving reading with respect to each fixed reading and then the phase angle of each cross-spectrum was computed to ascertain relative motion. This phase information was then used to formulate a frequency domain function. This new function was downloaded to Vibrant Technologies ME’Scope software for ODS animation of one, two, three and six times the fan shaft frequency. Significant bending modes were discovered for the channels, cross beams, bearing support box and motor supports at almost all of these frequencies. Figure 1 below illustrates the significant system displacement at twice the fan shaft frequency.

Modal Analysis

Modal analysis was performed with impacts made on the inboard fan bearing, bearing support box and skid beams. Response data were collected for all specified collection points in the system with respect to each impact strike. After processing this data they were downloaded to ME’Scope to observe the mode shapes. Impact testing revealed the same modal frequencies and shapes for the channels, cross beams and bearing support box as were expected from the operating deflection shape analysis. The system displacement at twice the fan shaft frequency is presented below in Figure 2. Since there were so many different modes on most of the major components, a redesign to move just one or two natural frequencies would have been insufficient. Therefore, the goal of the redesign was to move all modes above the operating frequency range.

Finite Element Analysis

Finite element analysis (FEA) was utilized as a redesign tool for the bearing support box and skid. After concluding that the best redesign for the skid was to fill it with epoxy grout, a solid model was created and analyzed to ensure that it was well out of danger for a resonance problem. This was found to be the case as the first mode frequency was above three times the fan shaft frequency. The results of FEA for the bearing support box showed predicted mode shapes in the operating frequency range that matched the modal results well enough to validate the FEA model. The redesign of the bearing support box consisted of a four piece grid installed within the box to make nine cells. After a few iterations to determine the correct placement of the grid walls with respect to the outer box walls, the first mode frequency was increased by more than 300%. However, this was still judged to be unacceptable as it was too close to the blade passing frequency. To further increase the first mode frequency, well above any problem frequency, a piece of angle iron was added to the front and back of the box. The final design of the bearing support box in its first bending mode is shown in Figure 3.

Isolation System

An isolation system for this assembly was designed to provide 97% isolation. Using the system mass and desired isolation system natural frequency, the maximum isolation spring stiffness was calculated. With this information it was determined that only four springs were needed for the isolation system which conformed to the stiffness constraint and had adequate static load capacity. The springs of the isolation system were mounted on the inside corners of the skid with four triangular brackets fully welded to the skid frame. The existing skid system had an isolation rate of only 30%. Considering the mass of the skid, the dynamic load on the building was determined to be 18,000 lbf. Combined with the static load this resulted in a total load of 22,000 lbf on the building. With the isolation system, the dynamic load dropped by 90%. The vibration to be isolated also dropped with redesign, so the new dynamic load was lowered to approximately 1% of that before redesign. This design created a dramatic, 400% reduction in total loading on the building.

Conclusion

In the case presented here, excessive vibration of an overhung fan required the redesign of several aspects of the system. The skid and bearing support box were redesigned to increase their lowest natural frequencies to an acceptable level, thereby avoiding resonance problems. Operating deflection shape, modal and finite element analysis were all utilized as tools in the investigation and redesign process. An isolation system was also designed and provided a dramatic reduction in total system loading. In the end, this fan went from the worst in