Monitoring of microfloor vibrations in a new research building

Dryver R. Huston* , Brian Esser, James O. Plumpton, Xiangdong Zhao Department of Mechanical Engineering, University of Vermont

ABSTRACT
This paper describes efforts at monitoring microfloor vibrations in a newly-constructed research building. This building is intended to house a variety of delicate precision scientific instruments with performances that are deleteriously affected by even small floor vibrations. The building is five stories with welded steel construction. Upon completion of the construction, the initial occupants anecdotally complained of excessive floor vibrations. This resulted in an effort to measure the vibrations and to reduce them at their sources, including the mechanical systems for the building. The measurements are compared with industry standards and with measurements taken at nearby reinforced concrete buildings. The success of efforts at reducing the vibrations due to the mechanical systems of the building are also assessed. Keywords: adaptive sensors, structures, sensor networks, damage detection.

1. INTRODUCTION
Scientific research buildings house vibration sensitive instruments, e.g. AFMs, SEMs, TEM and nano-instruments. These instruments can be very sensitive to floor vibrations, even ones that are imperceptible to humans. A new medical research building was built on the University of Vermont campus in 2001. This building was designed to house a variety of medical research laboratories, some of which hold specialized and vibration-sensitive equipment. Upon moving in, anecdotal reports from new occupants complained of excessive floor vibrations and concerns about laboratory instruments. An investigation was undertaken to determine if the vibration levels were excessive. The are several potential sources for building vibrations. These include external sources, such as traffic, seismic activity, and nearby construction; and internal sources, such as building mechanical systems (HVAC, plumbing), human movement, and machinery. These vibrations can contain a wide range of frequencies, amplitudes, time dependencies, and stochastic characteristics. Upon a review of the literature, it is apparent that there are no widely accepted standards for quantifying and regulating floor vibrations1,2 . ISO 26313 describes methods for assessing the human comfort and perception relative to floor vibrations, but gives little guidance as to the effect of floor vibrations on machinery. ISO 2631 does assess the severity of floor vibrations on a frequency-dependent basis. The frequency dependence concept can be extended to machine vibrations by noting that most sensitive machines are mounted on heavy vibration isolation platforms. These platforms usually employ passive vibration isolation and damping techniques to minimize the vibrations that are transmitted to the machine. The platforms are very good at reducing the effect of high-frequency vibrations, but have difficulty reducing low-frequency vibrations. A very simple frequency-dependent criterion is one that is used at a local semiconductor manufacturing facility in which the requirement is to keep the amplitude of the displacement of vibrations that occur in the 5 -10 Hz range to less 100 microinches. A further extension of the concept that the effect microfloor vibrations can be assessed with frequency dependent criteria is to use the BBN criteria 4 . These criteria recognize that there are several forms of vibration that can cause problems, i.e., periodic, random, single frequency, multiple frequency, stationary and nonstationary. The BBN criteria are based on the 1/3 octave spectrum statistic, which can accommodate and quantify a wide variety of vibration waveforms, with the primary exception being nonstationary signals.

huston@emba.uvm.edu; phone 1 802 656 1922; fax 1 802 658 7409; Mechanical Engineering Department, 201E Votey Bldg., University of Vermont, Burlington, VT 05405-0156

The BBN criteria give five levels (A-E) and suggestions for the use of various machinery: A- Optical microscopes 400x, BOptical microscopes 1000x, C- 1 micron feature lithography, D- TEMs, SEMs, and E-beam writers, E- Difficult to achieve. Since the 1/3 octave spectrum (and other spectral techniques) are not good for quantifying nonstationary vibration statistics, other statistics were also used. These included: 1 minute RMS (acceleration) vs. time, 1 minute maximum absolute acceleration vs. time. Since there is the possibility of frequency dependent effects, the RMS acceleration and maximum acceleration statistics were also calculated on signals that were preprocessed through a 5 -10 Hz bandpass filter. An additional statistic to be calculated was the histogram of the vibrations. The histogram is useful in assessing whether or not the vibration amplitudes are Gaussian.

2. VIBRATION MEASUREMENTS
In order to assess the level of vibrations in the New Medical Research Building, a survey of the vibrations was conducted in three buildings on the University of Vermont campus: Building 1- New Medical Research Bldg with a 4 story, welded steel frame with box columns; Building 2- Given Building (in the Medical Alumni wing) with a 4 story, steel frame with masonry walls; and Building 3- Votey Building with a 3 story reinforced concrete slab frame. The buildings appear in Figure 1.

Figure 1 a. New Medical Research Bldg. b. Given Bldg.

c. Votey Bldg

Vibrations were measured in selected rooms of each building with four piezoelectric accelerometers. The accelerometers were configured in a triaxial layout at one point to measure combined vertical and horizontal vibrations and at a single point in a different location to measure vertical vibrations. The accelerometers had a rated sensitivity of 0.001g. Vibration measurements were taken over a period of approximately 24 hours at each location. The sampling frequency was 100 Hz.

3. MEASURED DATA
Figure 2 shows the maximum vertical accelerations per minute measured in the New Medical Research Building room 407. In order to assess frequency dependent effects, the data were 5-10 Hz band pass filtered. Figure 3 shows the maximum vertical accelerations in the New Medical Research Building room 407 with a 5-10 Hz bandpass filter. Figure 4 shows the corresponding unfiltered RMS acceleration per minute. Figure 5 shows the corresponding filtered RMS acceleration per minute. The acceleration data were processed into a 1/3 octave spectrum and plotted vs. the BBN criteria, Figure 6. A histogram of the acceleration data appear in Figure 7. A similar set of data processing was conducted on the data collected in Given Building room 165. Figure 8 shows the maximum acceleration per minute measured in Given Bldg room 165. An examination of the data indicates that large acceleration spikes occur approximately every seven minutes. In between the spikes, the data appear to be rather quiescent. A closeup examination of the associated time histories revealed that the acceleration spikes were so large that the accelerometer amplifier became saturated, which prevented the acquisition of useful data during the spike. The particular cause of these spikes was tentatively attributed to the cycling of a cooling compressor for a nearby electron microscope. Data in between the spikes appeared to be valid. Figure 9 shows the 1/3 octave spectrum for vibrations between large spikes Given Bldg room 165. Figure 10 is a histogram of vibrations in Given Bldg room 165, including the vibration spikes. Figure 11 is the maximum acceleration per minute for Votey Building room 116. An examination of this data indicated that moderate quantization problems due to the level of vibrations being so small that the lower limits of the measurement

instrumentation are being reached. Figure 12 is the corresponding RMS acceleration per minute. Figure 13 is the 1/3 octave spectrum of the accelerations vs. the BBN criteria. Figure 14 is the corresponding histogram.

Figure 2 Maximum vertical accelerations per minute measured in New Building room 407

Figure 3 Maximum vertical accelerations in New Bldg room 407 with 5-10 Hz bandpass filter

Figure 4 RMS acceleration per minute in New Bldg room 407

Figure 5 RMS acceleration in New Bldg room 407 with 5-10 Hz bandpass filter

C D E

Figure 6 1/3 Octave spectrum with BBN criteria New Bldg room 407

Figure 7 Histogram of vertical floor accelerations New Bldg room 407

Figure 8 Maximum acceleration per minute Given Bldg room 165

C D E

Figure 9 1/3 Octave spectrum for vibrations between large spikes Given Bldg room 165

Figure 10 Histogram of vibrations in Given Bldg room 165

Figure 11 Maximum accelerations per minute Votey Bldg room 116

Figure 12 RMS acceleration per minute Votey Bldg Room 116

C D E

Figure 13 1/3 Octave spectrum with BBN criteria Votey Bldg Room 116

Figure 14 Histogram of vibrations Votey Bldg Room 116

4. CONCLUSIONS
An examination of the data indicated that the reinforced building (Votey) is the quietest slightly larger than BBN level D. The histogram of the data was fairly Gaussian. The steel framed building with masonry walls (Given) had large acceleration spikes. The histogram of the data was distinctly nonGaussian. In between the spikes, the vibrations were relatively small slightly larger than BBN level C. The welded steel building is the liveliest, with a BBN level B. However, scientific instruments with good tables have been able to operate successfully to date.

REFERENCES
1. Kim JJ, Amick H. Active Vibration Control in Fabs Semiconductor International July 1997. 2. Huston DR, Spillman Jr. WB, Sauter W, and Pelczarski N. Measuring Micro Floor Vibrations with Distributed Fiber Optic Sensors SPIE Selected Papers Volume 47-53 Distributed Optical Fiber Sensors and Measuring Networks YN Kulchin, ed., 2000 3.International Standards Organization. ISO 2631/1 1985-05-151/3 4. Amick H, Hardash S, Gillett P, Reavely RJ. Design of Stiff Low-Vibration Floor Structures SPIE Vol. 1619, San Jose CA, 1991.