Industrial Sponsorship

Dynamic Performance Investigation of Base Isolated Structures

By Ather K. Sharif

1.1 Introduction

This thesis concerns an investigation into the dynamic performance of Base Isolated structures designed to attenuate the effects of groundborne noise and vibration caused by railways. Base Isolation involves the introduction of isolators, typically between the foundations bearing upon the soil and the superstructure.

Diagram 1.1

The earliest known application of Base Isolation arose from recommendations in 1916 by the New York Central Railroad Company, acting on the advice of Professor Miller of Columbia University, suggesting that all buildings constructed over their tracks should be built upon lead-asbestos pads. These recommendations followed from tests which involved holding two 15 foot steel rails on end, below one of which different isolators were in turn installed, and the vibration of the rails (due to a train source) was judged by one person using a sense of touch. Of the materials tested, it was concluded that cork would be sufficient to eliminate a considerable part of the vibration conveyed through the rock. Given the concerns over the long term durability of cork, another material which was tested, a combination of lead-asbestos material was finally recommended. Many developments in New York’s Park Avenue thus proceeded with the use of lead-asbestos pads, of a type shown in Diagram 1.1 (Goodfriend et al, 1962).

The first serious investigation into the performance of lead-asbestos pads was undertaken by Bolt Beranek and Newman Inc. in 1960, when evaluating a large development site known as the ‘Place Ville Marie’, over the Central Station of the Canadian National Railways in Montreal. An existing building (Queen Elizabeth Hotel) on a neighbouring site had already been isolated on lead-asbestos pads directly over the railway, and immediately alongside there was a structure incorporating car parks and a road bridge which was un-isolated.

Measurements of vibration on adjacent isolated and un-isolated columns showed a benefit of 5 to 12dB in the vertical direction and a benefit of about 4 to 6dB in terms of groundborne noise. However the measurements were processed using octave band-pass filters, which would not provide adequate frequency resolution. These measurements were used as the basis to proceed with the use of 1" thick lead-asbestos pads for the Place Ville Marie development (BBN, 1960; Brett, 1962). Subsequent reports about the performance of the completed building were only subjective, stating that a good result had been achieved (Miller, 1963).

Another development where performance was investigated occurred in 1961 at the New York Philharmonic Hall, which utilised 3" thick lead-asbestos pads in areas close to the subway source, reducing in thickness further away. Comparisons were made between measurements in the basement of the isolated building with those taken in the basement of a former un-isolated building which occupied the site, with both measurements taken at about the same distance from the subway (17m). Miller (1963) reports a worthwhile improvement of the order of 15 to 30dB in terms of vibration, but notes that part may be attributed to the different/heavier form of construction used in the Philharmonic Hall.

The International Lead-Zinc Research Organization were interested in promoting the continued use of lead-asbestos pads, and commissioned tests by Goodfriend et al in 1962. Their report provided laboratory test results for a mass supported on the lead-asbestos pad, simulating typical bearing pressures under a building column. This showed that vibration isolation could begin to be achieved above 50-70Hz, reaching 8-10dB reduction between 90 and 150Hz. Field tests were also commissioned below and above the lead-asbestos isolators in a 50 storey building, but the crucial data for measurements above the isolator were omitted, and put down to an instrumental fault.

Adverse remarks concerning the performance of lead-asbestos pads have been made by Crockett (1968;1973). He quotes that the Massachusetts Institute of Technology stated back in 1935 that the use of lead-asbestos pads for vibration isolation was questionable given the high stiffness and also cites examples of its limited unsuccessful use in the UK. Crockett reports that in 1937 a six-story office building in London was mounted on 100mm thick cork, but that this material was too stiff and in fact failed due to dry rot.

But the application of lead-asbestos pads in New York, with a vertical rigid body natural frequency of the order of 40Hz, may well have been to a degree beneficial in reducing groundborne noise. This is because ground vibration from trains on the rock site led to predominantly high frequencies in the spectrum and at least the method introduced a discontinuity in the construction. The translation of the lead-asbestos application from the USA to the UK was however particularly ill conceived, given that vibration from trains in the London Clay would lead to predominantly lower frequencies.

Recognizing the lower dominant peaks in the spectrum for train induced vibration in London, a softer Base Isolation system proceeded in the UK from 1965 using laminated natural rubber bearings (Diagram 1.2) for a target vertical natural frequency of 7Hz. It was intended to attenuate both low frequency perceptible vibration and groundborne noise. Waller (1969) is credited for the first building in the UK to be isolated in this way, known as Albany Court in London, built directly over an underground railway.

Diagram 1.2

The natural rubber used in these bearings is made from the latex produced beneath the bark of the rubber tree. Compounds are added to secure certain properties during the vulcanization. In particular, carbon black is added to improve the durability of the material and has the effect of increasing stiffness and damping, but also increases creep (Payne and Scott, 1960). Vulcanization is undertaken under high temperature and pressure in a mould (Lindley,1974). The bearing comprises the vulcanite typically 10-20mm thick interleaved with steel plates typically 3mm thick.

The size of the bearing in plan, the number of steel plates and the thickness of rubber are dictated by the engineering requirements for stiffness. A key parameter in the design of these bearings is the shape factor, which is the ratio of one loaded area to force free area, that is, the area that is free to bulge. The dynamic shear modulus and therefore dynamic stiffness have been shown to increase as strain levels are reduced, and depends upon the amount of carbon black filler that is added (Harris and Stevenson, 1986). This has a significant implication for the performance at high frequencies, which are associated with much smaller strains. The dynamic stiffness of these materials is greater than the static tangent stiffness at the working load, and depends upon the compound, but typically a ratio of 1.7 to 2 is quoted in Base Isolation applications (Grootenhuis, 1989). The horizontal stiffness of the bearing can also be controlled by design and is invariably less than the vertical stiffness.

The vertical rigid body natural frequency that can in principle be achieved using natural rubber in a Base Isolation situation is from 4Hz upwards (Muhr, 1992), though in earlier schemes was limited to 7.5Hz. In practice the lowest natural frequency is limited to 6Hz given the deflection that is needed. There are a number of companies supplying steel laminated natural rubber bearings, but the principle source in the UK is Andre Rubber, which is part of the BTR group.

An alternative to natural rubber has been the use of a synthetic rubber known as neoprene, which can be used in two ways, shown in Diagram 1.3 and 1.4.

Diagram 1.3

Diagram 1.4

In one version (Diagram 1.3) the material is composed of layers of high tensile strength fabric bonded to layers of neoprene modified by the inclusion of cellular particles. Under load, the neoprene bulges and compresses the cellular particles, which is a method of reducing the stiffness to achieve lower natural frequencies. An alternative to the use of cellular particles in the neoprene is to create air voids into which the rubber may bulge. These air voids are formed as cylindrical holes within the bearing (Diagram 1.4).

The rigid body vertical natural frequency that can now be achieved in a building using neoprene is 7.5Hz, although are typically in excess of 10Hz. These materials also exhibit a higher dynamic stiffness, by a factor of 2 to 3 times, which can be significantly greater (BS6177:1982).

TICO in their promotional literature provide graphs of target natural frequencies as a function of compressive stress for various product types and thickness. The user is directed to select a desired natural frequency, and deploy the bearings, of a given type, with a plan size and quantity that meets the required compressive stress (see TICO prom lit). The principle supplier in the UK for these synthetic rubber bearings is TICO who are now known as TIFLEX.

Many developments using such isolators have continued, with numerous papers (Waller,1969; Crockett,1968,1973,1982; Reed, 1966; Grootenhuis, 1979,1982,1987, 1989; Liquorish and Green, 1987; Cowell,1991; Makovicka, 1992; Anderson, 1994) describing the concept of Base Isolation and how it had been implemented at a number of sites, with vertical natural frequencies in the range 7Hz to 20Hz. The lower vertical natural frequencies being selected to cater for low frequency perceptible vibration and higher frequencies where groundborne noise alone was the main concern.

A British Standard was published in 1975 as a draft for development (DD 47), being revised and issued in 1982 as BS6177 ‘Guide to selection and use of elastomeric bearings for vibration isolation of buildings’. It deals principally with aspects of safety concerning the use of elastomeric bearings. It details the different types of elastomeric bearings, and outlines considerations for their design and testing, along with some implications on the design and construction of the building itself. It did not deal with the issue of Base Isolation performance.

There has been a trend to aim for lower rigid body mount frequencies, ever since the concept of Base Isolation was introduced in 1916. Limiting factors are a concern over increased susceptibility to wind induced motion, a desire to limit the magnitude of deflection that can arise during construction, which can range from 5-30mm, and the increased cost for softer mounts.

Heiland (1992) reports upon an investigation undertaken in Germany around 1979, to investigate how low the natural frequency for Base Isolation could be taken with due regard to the increasing sensitivity to wind induced motion. For this investigation, a 6 storey building over an underground railway in Berlin, was mounted on steel helical springs for a vertical natural frequency of 7Hz. Given that there were no adverse effects due to wind loading, the steel helical spring system was altered to reduce the rigid body mount frequency down to 3.5Hz, with successful results.

Diagram 1.5

The vertical and horizontal stiffness of these steel helical units, shown in Diagram 1.5, are a function of the wire diameter, mean diameter of windings, number of effective windings, the shear modulus of the material and the free height of the spring.

The lower natural frequencies in buildings can be achieved but stability of the spring is important. Often a number of steel helical springs are used in a capsule. Because the springs are made from steel there is no inherent damping in the material. Additional viscous dampers may be used in tandem, or the damping can be achieved by submerging part of the coil in a viscous fluid. The steel coils can however transmit audio frequencies and therefore their design usually incorporates a rubber mat above or below the unit.

Base Isolation systems with vertical natural frequencies down to 4Hz using such steel helical springs have also been used in the UK since 1992 (Manning, 1996; Anderson, 1996). The principle supplier for these isolators in the UK is GERB. Muhr (1992) shows that laminated natural rubber bearings could in principle be designed to achieve similar low frequencies, although to date there are no known projects in the UK adopting such low frequency rubber bearings.

Appendix 1.1 gives a list of large structures that have been Base Isolated in the UK using the common types of isolators in use.

The performance expectations in Industry for a Base Isolation system are high, as seen by some examples quoted in the literature. Waller (1971) indicates that a reduction might approach 20dB in good circumstances. Crockett (1982) quotes a benefit of 14dB. In the discussion section of a paper by Grootenhuis (1982) it is remarked that a hoped for benefit between isolation and no isolation is 20 to 30dB. Wilson (1988) justifies a design based on an expected attenuation of 17dB to 34dB. In another paper, Krüger (1990) suggests a benefit of 20dB. Bines (1993) uses the transmissibility curve for a SDOF model to anticipate 14dB attenuation at a given frequency ratio. Such an approach is also inadvertently illustrated in the promotional literature of some isolator suppliers as well as in some technical papers, reproduced here in Diagram 1.6.

Diagram 1.6

Most buildings are believed to have been Base Isolated on the performance predicted by a SDOF model, as no other model has ever been proposed. Where there are papers that describe the application, they rarely ever show measurements to demonstrate the actual performance achieved. Yet the response of a SDOF model, shown in Diagram 1.6, would not provide the insertion loss of a Base Isolation system, which is the benefit or disbenefit that arises between a building with Base Isolation and that which might have arisen had traditional foundations been used. Clearly a SDOF model would also not be able to describe any resonance modes of the supported structure itself.

The first more appropriate theoretical examination of the performance of Base Isolated structures was undertaken by Newland (1989) and Grootenhuis (1989), who examined the axial behaviour of a column (simple bar) mounted on isolators. They concluded that axial resonances of the column could negate the performance predicted by a SDOF model, and depended upon the damping that was assumed. The axial modes of the simple bar were shown to be a function of the material properties and bar length, and the resonance frequencies for a typical tall building (30m column length) lay in the low audio range (see Figure 5.9 of Chapter 5). Rücker et al (1986) have also theoretically examined the axial column and beam resonances of a 2-dimensional portal frame for a variety of storey heights, but in the context of traditional foundations.

Walker and Mathers (1992) investigated the use of isolators for mounting box-within-box type constructions for recording studios, to deal with groundborne noise. They showed that performance could be limited by the modal characteristics of the isolator itself where many of the axial modes of a coil were not adequately damped using the thin ribbed rubber pad (noise stop) which is often placed at one end of the isolator. This emphasises the need to adequately damp all resonance modes within an isolator, which arise within the frequency range of interest. They showed that the manner in which this damping is introduced in the isolator is critical. They demonstrated this with a small section of writing paper placed inside on the circumference of the steel coil (50mm dia.) and short of the ends, which created friction damping. This method of damping raised the effective stiffness of the isolator at low vibration levels, as the motion was unable to overcome the friction force. Submerging the coils partially in a viscous fluid can damp the axial modes (circa 300Hz) of the coil more effectively. Alternatively the problem can be eliminated from the frequency range of interest by raising the resonance frequency, by splitting the coil into two lengths separated by a heavy steel plate (Heiland, 1992;NVW, 1992). A small test floor resiliently mounted in the basement of a building subject to vibration and groundborne noise from an underground railway was used to demonstrate that resonance of the floor would limit performance. They also used this experiment to show that an airborne acoustic path, such as within an air cavity, could cause the isolated structure to vibrate, thereby limiting performance to less than 30dB, less than might be expected from a SDOF model at the relevant frequency.

Cryer (1994) undertook a more detailed theoretical and experimental investigation. He showed using a two dimensional infinite model of a building and its piled foundations that an improvement of the order of 10dB was possible with Base Isolation. These predictions showed similarities with measurements taken in a building over an underground railway in London. A parametric study was used to show that performance was very sensitive to the natural frequency of the mount system, improving with lower frequency. Performance was shown to be less sensitive to the damping of the isolators, and damping in the structure was relevant in that it would help to attenuate resonance modes of the frame. He showed that the position of the isolators within the building was important, with generally better performance predicted when isolators are placed near the foundation, as opposed to higher up in the building. This is because resonance of the structure below the isolators could adversely effect the response of the isolated building. He theoretically demonstrated that a building with Base Isolation could respond more adversely than an un-isolated building, due to sources of vibration that were internally generated.

Maillard (1997) reports vibration measurements in a 5 storey building in France immediately adjoining an at-grade railway, and Base Isolated using steel helical springs for a target natural frequency of 3.7Hz. The isolators were placed upon the top of enlarged columns, rising through a car park 2 storey in height. He shows measurements across the isolators to the ground floor parts of the structure, indicating a significant reduction, with a mean of 15dB at 63Hz and 19dB at 125Hz, in line with the required attenuation at these frequencies. The measurements did not strongly indicate any modal characteristics of the structure below or above the isolators, but the analysis was based upon the resolution of 1/3rd octaves.

Kebe (1998) shows measurements in 4 storey houses Base Isolated using steel helical springs, immediately adjoining an at-grade railway in Germany. The resulting environment met German standards, except in one house, where transmissibility measurements showed areas of improvement in the spectrum, accompanied by areas of reduced attenuation / amplified levels, above the rigid body mount frequency of 5Hz. The poor performance of this property was attributed to a suspected short circuit.

Despite the recent knowledge (Newland,1989; Grootenhuis, 1989; Newland and Hunt, 1991; Walker and Mathers, 1992; Cryer, 1994), Industrial practice for the selection and implementation of Base Isolation appears to proceed on the basis that it is at least conservative to Base Isolate, even if it is not entirely appropriate or justified.

The entire Bridgewater Concert Hall structure in Manchester was Base Isolated on steel helical springs with a target vertical natural frequency of 3.5Hz to deal with a light metro operating at grade at 30m distance (Anderson, 1996). The need for Base Isolation, or for Base Isolation to such a low natural frequency might be questioned, when a simpler form of discontinuous construction may have sufficed in this case.

A legal case arose in the UK in 1993, resulting from an inadequate design and specification for a Base Isolation system. A residential development was to have been built adjacent to a factory of forging presses, and the original design for the site detailed a scheme using steel helical springs for a low natural frequency (Crockett & Associate, 1988). Owing to the high cost of the steel helical isolators, the developer elected to adopt a scheme prepared by a different designer, involving the use of less expensive elastomeric bearings. Following occupation of the site, a claim for nuisance was pursued by the residents, and supported by the Council. The Court ruled that a statutory nuisance existed as a result of a poorly designed and specified Base Isolation system, with costs and damages awarded against the developer (see Sharif, 1993).

It is notable that the design for Base Isolation systems in the UK, are never accompanied by a formal performance specification that could form the basis of some contractual agreement between the Designer of the Base Isolation system and the Developer, and be subsequently checked for compliance. Such a lack of formalism in this area of design has led to cases where Base Isolation will not always provide a cost effective means of isolation and may in some cases have a detrimental effect (Sharif, 1987; Sharif, 1993).

The additional cost of the Base Isolation 'system' and its 'knock on effects' on the structure are typically 2% to 5% of project value, where the larger ratio may arise in smaller projects (the cost of the isolators themselves being a much smaller percentage). This additional cost may therefore in some cases be unjustified, or may be better spent on alternative vibration control measures.

Where Base Isolation is deemed appropriate, it may be possible to optimise performance, using a greater understanding of the performance related issues. Where good performance can be achieved, the additional cost of the Base Isolation system is in fact a small proportion of project value, given the scale of benefit that can be achieved.

1.2 Objectives

The objectives of this thesis are to develop a deeper understanding of the performance of Base Isolated structures, and the wider implications of Base Isolation, with a view to preparing an International Standard that could form the basis of a contractual agreement between the Designer and the Developer. This is to ensure that Base Isolation is implemented in appropriate circumstances, bearing in mind its cost effectiveness in relation to possible alternative vibration control options, and that it is modeled appropriately, with a greater focus of attention on expected and achieved performance.

1.3 Outline of Thesis

Chapter 2 describes railway-induced vibration source and propagation characteristics, and Chapter 3 deals with the effects on buildings. Chapter 4 describes alternative vibration control measures, with Chapter 5 covering theoretical aspects of isolation. Chapter 6 covers instrumentation and data processing. Case studies in Chapter 7, 8 and 9 are used to understand the dynamic behaviour of Base Isolated structures. Impact, shaker and ambient testing have also been used to establish dynamic characteristics, and simple theoretical models are used to aid an understanding. An International Standard is drafted in Chapter 10. Conclusions and recommendations are given in Chapter 11