Industrial Sponsorship

Dynamic Performance Investigation of Base Isolated Structures

By Ather K. Sharif

3.1 Introduction

Vibration from railways may need to be evaluated in terms of effects on the building structure, response of occupants, or any possible adverse effects on the operation of equipment. This Chapter describes an evaluation in terms of these distinct areas.

3.2 Effects of Vibration on the Building Structure

Public opinion, brought about primarily by the judgement of levels based upon human perception is that vibration which is strongly perceived poses a risk of damage to the structure. An inspection of the property often reveals cracks, which are readily attributed to vibration, without any consideration that such damage may have existed for some time due to other causes. This became a typical situation for homeowners who have been exposed to vibration from blasting or construction activities, and led to demands to measure vibration, often a specific requirement of the Contractors insurers.

The growth in instrumentation has led to ever more detailed measurements of vibration, often in terms of velocity, yet very little effort is directed at carefully monitoring the structures for evidence of damage which can be directly linked to the vibration.

Whilst Peak Particle Velocity (PPV) is often used to assess vibration effects on a building, velocity in itself cannot create damaging forces. New (1986) pictorially showed (Diagram 3.1) that such forces are generated by :

Differential displacements, which give rise to distortion of the structure as it follows movement of the ground.

Change in ground particle velocity vector (magnitude or direction) which produces inertial forces upon the structure.

Distortion and inertial mechanisms arise at the same time. These are superimposed upon pre-existing stresses and strains from other causes. Damage occurs when the combined stress exceeds a limiting (trigger) level, or there may be fatigue failure (Crockett, 1979).

A misconception that vibration measurements in some way provide an accurate measure of damage still however exists, whereas in fact it does not measure damage directly at all, but is merely a parameter to gauge its possible effects, based upon reference to a data base of experience. Statistical analysis of a database leads to guide values below which probability of damage may be small. A database has grown within many specialist organisations, and this has sometimes been collated to provide guide values in National standards (UK BS7385 pt2, 1993; German DIN 4150 pt3, 1986; French Regulation 23, 1986; Swiss SN640312a, 1992; Sweden SS 460 4866, 1989).

Peak particle velocity (PPV) is used to characterise vibration, because it is the best single descriptor for correlating case history data with vibration-induced damage, and has a reasonable theoretical basis (Gasch, 1968; Steffens, 1974; New, 1986). This may be reported as either the 'peak component particle velocity' (the maximum value in any direction), the 'peak true resultant particle velocity' (summing the three orthogonal components coincident with time), or the 'maximum vector sum' (vector sum using maximum of each component regardless of time). The latter format is now discouraged as it includes an unknown factor of safety (BS7385 part 2, 1993). The most prominent and credible research to identify safe vibration levels from blasting was undertaken by the United States Bureau of Mines (Siskind et al, 1980). This reports from case history data that a limit of 12mm/sec will provide protection from blast damage in greater than 95 % of cases (less than 5% probability of very superficial cracking), which is summarised in terms of frequency and house type in Table 3.1.

Table 3.1 Safe levels of blasting vibrations (transient) for residential type structures (2 storeys), sited on firm foundations (USBM:Siskind et al, 1980)

Type of Structure
Ground vibration PPV in/sec (mm/sec)
> 40Hz
Modern homes, drywall interiors
0.75 (19mm/sec)
2.0 (50mm/sec)
Older homes, plaster on wood lath construction for interior walls
0.50 (12.7mm/sec)
2.0 (50mm/sec)

The above criteria has a sharp transition at 40Hz, and therefore a more detailed criteria in terms of displacement and PPV limits was also put forward and shown in Figure 3.1.

Other systematic studies include work by the Transport Road Research Laboratory in the UK (Hood and Marshall, 1987) who exposed a ninety year old pair of semi-detached houses on loose to medium dense sand to prolonged exposure of simulated road traffic. Airborne noise was simulated by a loudspeaker mounted on the side of a lorry and vibration was simulated using a geophysical vibrator. The vibrator produced an impulse on the ground (on a concrete pad 2.6m to the side of the house) with a period of 1 second, leading the building foundation to respond in the vertical direction at 2.5mm/sec with a frequency of 12Hz. The 1 second impulse, including the rise and fall time, led to 8 reversals at peak velocity, which is broadly equivalent to 4 (heavy goods) axles. Detailed vibration, structural and visual examination was undertaken, during a four month period in which 888,000 vibrator single second pulses were completed, equivalent to over 3.5 million (heavy goods) vehicle axles. No foundation movement was detected, although new plaster cracking and extension of existing cracks did occur, but damage was slight and small compared to cracking that already existed prior to the experiment. It was also said that hairline cracking in some plaster finishes would ordinarily go unnoticed in a normally decorated house.

Studies have also been commissioned to examine the possible effects of rail-induced vibration on buildings. A relatively simple experiment was undertaken by the Office for Research and Experiments of the International Union of Railways ( ORE D151 Rep 7, 1984). In this experiment a small brick wall, "L" shape in plan, measuring 3m long on one side and 1.15 m on the short side, was built off a solid concrete floor, to a height of 1.87m. The inside face of brickwork was plastered to a thickness of 13mm, giving an overall wall thickness of 113mm. This single leaf unloaded structure with a free edge was typical of a top floor interior wall found in buildings. The mortar and plaster were specified to be hard and brittle to represent a worst case structure sensitive to vibration. Six such identical walls were built inside an unheated, ventilated single storey building with concrete floors. No testing was undertaken for 30 days after construction following which a careful inspection of the walls was made. One test wall was used to identify the best method of attaching the vibrator and one wall remained as a control, not subjected to vibration, other than by indirect effects through the ground. The vibrator was suspended independently, and attached on the centre of the largest area of test wall from the plastered side, to apply only horizontal forces. The test was undertaken at a driving frequency of about 13Hz (identified as a first resonant mode 'transverse flapping'), to keep peak particle velocity on the wall to about 14mm/sec and continued for 60 days. The results concluded that the natural frequency was generally reduced (but also slightly increased) by up to 25%. During the course of the tests, some trivial damage was observed. This consisted of minor cracking in the joints on the exposed brick face side, with surprisingly no cracking on the plaster face at all. The report concluded that the sinusoidal vibration at a peak value of 14mm/sec for the period of sixty days, was for example representative of the passage of 100 trains/day for 15 years, with each passage lasting 10 seconds. This exposure did not lead to any serious changes, where even the minor damage noted was not entirely attributed to vibration. The report omitted to give proper account of the change in natural frequency, which amounted to a significant change in stiffness.

This experiment represents the type of study which needs to be pursued in much greater detail, isolating all the variables that arise in a real life structure (BRE digests 75:1975 and digests 227, 228, 229:1979), where damage can occur for reasons both unrelated and affected by vibration. Such as thermal, moisture, shrinkage, pre-strain, creep, static overloading and differential settlement, notwithstanding chemical changes in materials, normal ageing effects or even poor workmanship, which can all cause signs of damage when there is no external source of vibration.

There remains some debate about the question of long term effects of low level vibration, which were raised by Crockett (1979), which can only be addressed by well conceived controlled experiments. A more recent review of the possible effect of road traffic vibration on heritage buildings was undertaken by Morton and Fox (1989), which concluded that there are many other plausible reasons for the damage observed, related to factors such as poor ground, weathering effect and decay. Another factor not specifically mentioned relates to the effect that the road surface itself introduces on the moisture content of the soil, such as drainage of storm water from road surfaces.

There is now an international standard (ISO 4866, 1990) which provides general procedures for the measurement and evaluation of vibration in buildings. It classifies damage to structures, as 'cosmetic' (formation of hairline cracks or growth of existing cracks), 'minor' (formation of large cracks or loosening and falling plaster) and 'major' (damage to structural elements). It does not provide levels of permissible vibration to prevent onset of cosmetic damage. This is left to National standards bodies, but does indicate factors which increase sensitivity of a structure to vibration damage such as; category of structure (elderly/modern), foundation types (from piled to no foundation at all), soil type (from rock to fill). It indicates that limits should be approached in a probabilistic way, where minimal risk for a named effect (e.g. cosmetic damage) is usually taken as a 95% probability of no effect.

In the UK provisions of ISO 4866 were accepted, and it was therefore issued as a dual numbered document, BS7385 part 1(1990). Guidance on safe levels of vibration was provided in BS7385 part 2(1993). In arriving at the guide values, BSI oversaw the compilation of a UK database, to identify case histories where vibration was credibly demonstrated as the cause of damage. The study involved consultation with 453 organisations in the UK by Moore and Malam (1987). They showed that in the data blasting and piling were the most common sources of vibration to be measured and a prevalence of two storey domestic buildings. From this database only few cases of damage could be directly related to vibration and even then with a degree of uncertainty (Malam, 1994). There was therefore a lack of reliable damage data in the UK to form a basis of guide values. Enquiries by BSI panel GME/21/3/2 of other European standards organisations revealed no more substantial data supporting definitive guidance given elsewhere in Europe at the time. Guide values in BS7385 part 2 (1993) were therefore largely based upon a review of systematic studies, such as the small sample described.

The standard provides guide values shown in Figure 3.2 for transient vibration in terms of peak component particle velocity, measured at the base of a building, to prevent onset of cosmetic damage with minimal risk (95% confidence). Strains imposed in a building by ground motion depend upon frequency, being larger if low frequencies predominate, and smaller at higher frequencies. The limits are therefore frequency dependent. At frequencies below 4Hz, a constant displacement limit is applied. These limits relate to transient vibration, for which it is assumed that resonant response in the structure does not arise. For continuous vibration which will cause magnification due to resonance, it is suggested that the guide values may need to be reduced by up to 50%, a factor based more upon current practice than is substantiated by case data.

The standard does not recommend reduction of limits due to fatigue, unless demonstrated as relevant by calculation. There is a great deal of knowledge concerning fatigue effects on steel (BS7608, 1993), and a reasonable knowledge on concrete (ACI, 1982), but very little is known about fatigue effects on other more traditional building materials such as brick, mortar and plaster. The standard does not advocate lower levels for a historic building, unless it is structurally unsound, although notes that special consideration may be required if an important building is difficult to repair. The possibility of building damage due to soil compaction (consolidation and densification by rearrangement of constituent particles) is guided to be an issue at shear strains of 0.0001, which becomes marked at 0.001. This shear strain can be estimated as ratio of PPV to shear wave propagation velocity for the soil (New, 1986). Vulnerability of loose and water saturated cohesionless soils to liquefaction is also noted (Richart et al, 1970). Whilst there are no suggested modifications to the guide values to cover these issues, it is recommended that these concerns be dealt with on a site specific basis.

The limits provided in this UK standard (BS7385, part2:1993) are somewhat different to those discussed in other UK guidance such as; DoE Mineral Planning Policy Guidance Note (DoE MPG9, 1992) which states that blasting associated with the minerals industry should be limited to 12mm/sec, although this limit considers environmental impact as well. BS5228 part 4 (1992), which is a code of practice for noise and vibration applicable to piling operations sets out more onerous limits for intermittent vibration to avoid cosmetic damage. These are given as 10mm/sec for residential properties, increasing to 20mm/sec for light industrial buildings and 30mm/sec for heavy industrial buildings. It indicates that, at frequencies below 10Hz the limits may need to be reduced by 50%, and above 50Hz the limits may be doubled. It deals with continuous vibration by recommending that limits be halved. The greatest disparity occurs with the tentative guidance given in the Construction Industry Research and Information Association (Head and Jardine, CIRIA Tech note 142, 1992). This document states that 'for PPV values between 2 to 10mm/sec generated by impact piling there is an increasing possibility of plaster cracking, or enlargement of existing defects'. Such onerous limits are in fact taken primarily from the German Standard DIN 4150 part 3 (1986), which are shown in Figure 3.3.

There is therefore some significant variation in guide values both within a Nation and Internationally. Figure 3.4 compares a set of frequency based guide values from the UK, USA and Germany. Such differences between countries may however be anticipated by the variation in structural types and ground conditions they cater for. For example, much of the data from the USA relates to timber framed houses, although brick structures are included (Siskind et al, 1980). Whereas guidelines in Sweden, which are for buildings mainly founded on rock, adopt a safe limit for blasting as high as 70mm/sec (vertical velocity) for normal residential buildings in masonry (Langefors and Kihlström, 1959; Swedish Standard SS 460 4866, 1989).

Vibration levels in buildings adjoining railways can be assessed in relation to a chosen set of guide values, although the key issue is whether to take train vibration as a transient/intermittent or continuous source. Individual events may be classed as transient/intermittent, a fast train for example. Whereas a freight train pass-by can persist for a long time, and railways also cause long term exposure, which could justify the classification of a continuous source. There is as yet no consensus on this point.

British Rail (Report to GME/21/3/2, 1993) undertook measurements at 19 residential type properties of brick construction, which were all 1 or 2 storey. Their distances from lineside varied from 1.5m to 25m with a mean distance of 4.8m. The PPV in the vertical direction ranged from 0.7 to 8mm/sec, with a mean value of 3.2mm/sec and in the horizontal direction varied from 0.8 to 10.5mm/sec with a mean value of 3.6mm/sec. These levels are clearly below the values given in BS7385 part 2 (1993) (see Figure 3.2), and therefore would not be expected to cause cosmetic damage, even if some allowance were made for the continuous nature of the source. Vibration from railways at grade, and certainly for railways in tunnels, which produce less vibration, are therefore not considered to damage structures which are in a good state of repair.

However, it must be noted that any structure, which is highly strained and therefore at the point of cracking, may be triggered to form a crack, and this crack would in due course have occurred anyway. This is particularly so at points of high stress, (stress raisers at window and door openings). Some ceilings of the old 'lath and plaster' type can have large areas of heavy ceiling inadequately supported due to failure of the tails. Any level of vibration can cause the ceiling to come down, but such a situation could equally have occurred from vibration generated within a building, such as due to footfalls, door slamming and moving furniture which can cause much higher levels of vibration than can possibly arise from trains.

The variation in guidance and lack of firm experimental evidence concerning long term fatigue effects on traditional building materials is likely to leave the railway industry open to allegations of vibration induced building damage. Interestingly however, railways in the UK are a 'Statutory Undertaker' and therefore are obliged in law to provide a rail service. They therefore cannot be prevented from doing so by a private prosecution for nuisance, and if they were ever challenged would have recourse to demonstrate 'best practicable means' (BPM).

3.3 Evaluation of Human Response to Vibration in Buildings

The earliest notable work to establish human annoyance and comfort levels related to laboratory studies conducted in Germany by Reiher and Meister (1931). They obtained judgements from 10 subjects to identify boundaries between not perceptible, weakly perceptible, easily perceptible, strongly perceptible and unpleasant. One of the findings was a vertical vibration perception threshold of 0.3mm/sec for a standing subject, exposed to continuous vibration.

Many better controlled experiments have been conducted since, summarised by Griffin_(1990), which show that perception thresholds are difficult to establish, and exhibit a great deal of scatter. This is because determination of perception threshold requires many other factors to be considered or isolated. Such as frequency and duration of vibration, effect of visual and audible cues, foot vibration of seated subjects (stationary or moving footrest), subject attention (reading or concentrating), presence of background vibration, subjects willingness to guess or wait until they are sure that they could feel the vibration, subjects confusion between a required judgement for f perception or comfort, and lack ofrt repeatability of results for a given subject's response. The psychophysical and statistical methods used for determining thresholds also have a significant bearing on the results.

Bearing these issues in mind, a lower perception threshold of 0.15mm/sec for a standing subject exposed to vertical continuous vibration is recognised, although it can seldom be quoted so simply, without qualification. A more qualified statement can be taken from Griffin (1990), who quotes that for approximate purposes a weighted peak acceleration of 0.015m/s2 may be expected to be perceived by 50% of persons.

Within residential areas people exhibit a wide variation of vibration tolerance. For example, Parsons and Griffin (1988) report that there may be a 2:1 difference between perceptible and unacceptable vibration in the home, and a 5:1 difference between the most and least sensitive subject. Acceptable values depend upon factors such as; the sensation the vibration produces, what they hear and see, what they expect, a concern about damage to the building, whether they believe anything could be done to reduce the vibration and whether they anticipate that expressing their dissatisfaction is likely to achieve a reduction or some financial compensation.

Social surveys have been conducted to identify an objective measure of subjective response and severity of vibration. Woodroof and Griffin (1987) undertook one such study involving a questionnaire to 720 potential respondents drawn from the adult population of Scotland who live within 100 metres of a railway line. Of the 720 potential respondents, 459 were interviewed, of which 160 reported noticing railway-induced building vibration. Vibration measurements were made in 56 different dwellings where occupants had said they noticed railway-induced building vibration. The study found that 34.8% of respondents to the social survey noticed railway-induced building vibration. Subjective judgements were found to be primarily related to the number of trains passing in 24 hours. Response to questions about annoyance were influenced by other factors such as noise, and it was said that building vibration was amongst the least annoying aspect of a railway's presence in a neighbourhood, whereas (Fields, 1979) in another social survey dealing primarily with noise, indicates that vibration was rated as the most important non-noise disturbance associated with the railway. There was no clear relationship of a variety of objective measures to rate severity of vibration. It was concluded that this is best done in laboratory experiments.

A great deal of interest in the area of human perception, and annoyance has led to both national and international guidance on the subject.

3.3.1 UK Guidance

BS6472 (1992) provides a means to evaluate the effects of vibration on Humans in buildings. Although not specifically developed for evaluating railway vibration in buildings, a train vibration evaluation is given as an example in the appendix to this standard. The standard indicates that acceleration is the preferred measurement, unless impulsive events are of interest, for which case velocity is said to be more convenient to identify peak values. The procedure involves recording the acceleration time history of the event, a train pass-by for example. This should be recorded on the building structural surface supporting the human body, on say the middle of a suspended floor, or allowances made for a transfer function between the actual measurement point, say at ground level to the point of entry to the human body.

The standard provides base curves (contours), shown in Figure 3.5(a) that represent approximate equal human response with respect to annoyance and/or complaint. These curves are not necessarily the same shape as vibration perception thresholds. However, at vibration magnitudes below the base curves, adverse comments/complaints are rare.

The standard indicates that doubling of suggested vibration magnitudes (e.g. in terms of r.m.s.) may result in adverse comment, and this may increase significantly if the magnitudes are quadrupled. It recognises that comments are likely to arise from occupants of buildings when vibration levels are only slightly in excess of thresholds of perception. However, it indicates that in the case of long term vibration exposure, adverse comment may be modified by familiarisation.

The sensitivity of a person to vibration depends, among other factors, to the direction of the vibration input in relation to the posture, for example vertical vibration from a floor is perceived differently to a person standing or lying down. The measured acceleration time history in the vertical and horizontal axes is therefore subject to a frequency weighting appropriate to the direction of vibration input and posture, and each direction is evaluated separately, for each occupied room. The modulus of the frequency weighting curve (BS6841, 1987) is in fact the inverse of the base curves and shown in Figure 3.5(b) for vertical foot to head axis, and horizontal back to chest, or side to side.

The standard indicates that summation effects arise due to vibration at different frequencies, which is why overall frequency weighted values are preferred. The frequency weighted acceleration time history for an event can be evaluated in various ways. The r.m.s. level of the frequency weighted acceleration can be a useful indicator of some vibration environments, but for a transient or intermittent event the r.m.s. level is affected by the time period over which the r.m.s. is calculated, and tends to underestimate the effect of the peak. An alternative is to use the r.m.q. which involves squaring the value twice, averaging and then taking the square root twice. An extension of this is known as vibration dose value (VDV) given in eqn. 3.1.

(aw = frequency weighted acceleration)

The vibration dose value time dependency means a twofold decrease in vibration magnitude (a) is equivalent to a 16 fold decrease in duration of vibration (t), where a4t=constant. This means that the VDV parameter is more sensitive to level, than duration. VDV is a measure of the total quantity of vibration perceived, and its value in the case of a continuous source of fixed magnitude increases as time passes. It is therefore essential to quote the time period over which the vibration dose is referred to.

BS6472 (1992) provides guidance on a range of VDV for both daytime and night time for ‘low probability of adverse comment’, reproduced in Table 3.2. This table does not consider the effects that groundborne noise, which may arise from structural vibration, can have on perception/annoyance.

Table 3.2 Vibration dose values (m/s1.75) above which various degrees of

adverse comment may be expected in residential buildings (BS6472, 1992)

Low probability of adverse comment
Adverse comment possible
Adverse comment probable
Residential buildings 16hr day
0.2 to 0.4
0.4 to 0.8
0.8 to 1.6
Residential buildings 8hr night

These guide values provide VDV at which there is low probability of adverse comment, and therefore by implication are used in Industry as acceptable values. This standard has been used to evaluate vibration impacts from the Jubilee Line Extension, High Speed Channel Tunnel Rail Link (CTRL) and proposed Crossrail scheme (Taylor, 1994; Ashdown Env, 1996). It is used by some consultants to evaluate development sites adjoining railways, and is sometimes referred to by planning authorities.

3.3.2 Validity of VDV guidance in BS6472 (1992) for evaluating Train Vibration

Given that the VDV method of BS6472 (1992) might be used to evaluate sites and influence a decision to adopt base isolation, or be used to judge an end result and thereby infer performance, it was interesting to examine this evaluation parameter in more detail.

When dealing with a new parameter, such as VDV, it is important to be able to gauge what VDV actually implies in certain environments. In an attempt to put the VDV for low probability of adverse comment into context, the Author undertook vibration measurements on BR platforms for railways at grade, and on the platforms in stations to the London Underground. Appendix 3.1 describes these surveys, the procedures of BS6472 (1992) used to evaluate these environments, and the results. Table 3.3 summarises the vibration dose value for train events recorded on the platforms and summed over a daytime and night time period.

Table 3.3 Vibration Dose Value on Railway Platforms

Measurement Location
16hr Day VDV
8 hr Night VDV
Purley Oaks Station (BR)

4.2m from edge of platform

(local passenger/non-stop Gatwick express)

(5 trains/hr)*
(2 trains/hr)
Carshalton Station (BR)

0.85m from edge of platform (local and non-stop passenger)

(5 trains/hr)
(2 trains/hr)
Northern and Victoria lines (LU)

1m from edge of platform

(20 trains/hr)
(2 trains/hr)

* assumed frequency of trains, VDV calculated for vertical direction using mean result from 5 events exceeds night time VDV column 2 of Table 3.2 by negligible amount

These results can be compared with the VDV guide values in Table 3.2, from which vibration levels on the railway platforms would according to BS6472 (1992) lead to low probability of adverse comment in all cases.

Yet these levels were recorded in some cases within 1m of a passing train, either arriving, departing or rushing through the station at speed. By obvious implication such levels would be indicated as appropriate on the floor of a bedroom or living room.

This would come as a surprise to most people who have experienced and can recall the sensation of vibration that arises from trains passing through a station. We should however bear in mind that our attitude when judging vibration in this environment would be affected by the airborne noise, air overpressure, or a prejudice that such vibration is bound to be unacceptable.

Even if the vibration level on the platforms were doubled in magnitude, this would produce a predicted daytime VDV at the upper range of values in BS6472:1992 (0.2 to 0.4 m/s1.75) for which low probability of adverse comment is again expected.

The standard having presented a dose approach, does not clarify whether the dose should only be measured for the source under study, the trains for example, or whether the dose should include the background vibration that occurs in the building. A further problem would then arise, where it could be argued that the background legitimately included events such as footfalls, door slamming, etc., which would leave the VDV measurement open to some corruption.

It is unfortunate that engineers do not as yet widely have a feel for this new parameter. It is also unfortunate that the current British Standard guidance has formed the basis of passing major rail projects (Jubilee Line extension, CTRL, Crossrail,) through Parliament, which has given it a further degree of credibility which makes it more difficult to challenge.

Whilst the frequency weighting approach has wide acceptance and fourth power dose relationship some acceptance (Griffin, 1990), there is clearly a need to arrive at a more realistic set of VDV guide values for low probability of adverse comment. Individual strong events at night time can cause sleep disturbance, but cannot be adequately dealt with by the dose approach. This might be better addressed by a supplementary limit in terms of overall maximum level (e.g. frequency weighted acceleration level, defined as the highest running r.m.s level with an integration time constant of 1_second for the event). This approach would be similar to the way in which airborne noise is dealt with, for example the UK Department of Environment Planning Policy Guidance Note PPG24 (1994) which suggests LAeq guide values for the daytime and night time period to deal with overall exposure, with LAmax,S limits on individual train events at night time in an attempt to reduce sleep disturbance (Section 3.4 for acoustic terminology).

It is therefore evident that there is a reasonable basis to review the current British Standard (BS6472) guidance. The results of this study have been used by the Author to present a case to the relevant BSI panel GME/21/6/4 (1998) to seek a review of current guidance. This resulted in the BSI panel approving a new work item to begin a review of BS6472 in 1999.

The VDV guidance of BS6472 (1992) cannot at present be used in the decision making process to adopt base isolation at a site, and compliance with limit values from measurements in a completed base isolated building should not be used to infer adequate performance.

3.3.3 Review of International Experience

International Standard ISO 2631 part1 (1997) provides general requirements for an evaluation in respect of health, comfort and perception, and motion sickness. Such evaluation may be relevant in vehicles, machinery and buildings. Guidance on measurement and evaluation procedures and frequency weighting curves are provided for a range of situations. The foot to head vertical and side to side horizontal frequency weighting curves, for an evaluation of perception in this international standard are in fact slightly different to those referred to in BS6472(1992), which are compared in Figure 3.6. There will inevitably arise departures between the desired and applied frequency weighting particularly at low frequencies, due to the practical constraints of applying weightings by digital means (Lewis, 1985).

This standard advocates three methods of evaluation, which are ; (i) A basic evaluation method using weighted root mean square acceleration for the event, which can be used in environments with a low crest factor(<9), (ii) Additional evaluation in terms of the running r.m.s using a 1 second integration time constant, which takes into account occasional shocks and transient vibration, (iii) The fourth power vibration dose method (VDV) which is more sensitive to peaks than the basic evaluation method.

A more specific international standard dealing with an evaluation of vibration in buildings was issued as ISO 2631-part 2 in 1989, although it is now currently under revision (see ISO/WD/2631-part2, 1998). No guidance was given on acceptable values, which was left to National bodies, and in the UK this resulted in BS6472 (1992), which was largely based upon this earlier international standard.

There are other ways in which vibration from trains may be evaluated. For example, the German Standard DIN 4150 part 2 (1992) requires the calculation of a different parameter known as KB values. In the USA (Remington et al, 1987; US Dept of Transportation, 1995) report a range of single number values (unweighted velocity) appropriate for various categories of use, making allowance for frequency of the events.

3.4 Evaluation of Human response to Noise in Buildings

Noise that arises in buildings from railways may be either predominantly via an airborne path, or reradiated noise due to the ground vibration. In the latter situation, ground vibration will excite the structure, which will cause bending waves in walls and floors, which can radiate noise into the building interior, much like a loudspeaker. This noise is a low frequency rumble, which is referred to as reradiated or groundborne noise. It occurs in the frequency range of 30Hz to 250Hz. Groundborne noise is particularly associated with underground railways. For railways at grade, the groundborne noise is more often masked by the airborne path which tends to dominate, unless the building fabric affords good attenuation, or the rooms are located on a façade naturally screened from the source.

Unlike airborne noise, there is no agreed method to evaluate groundborne noise. There is general acceptance that the A-weighted sound level is a good descriptor. The A-weighting (BS5969, 1981) is a method for making a sound measuring device respond in approximately the same manner as the human ear, which is comparatively less sensitive to low pitched and high pitched sounds. There is however some concern that the A-weighting does not properly account for very low frequencies (<100Hz), especially when the spectrum is unbalanced by dominant low frequencies (Broner and Leventhall, 1983). The characteristic unbalanced spectrum of a rumbling sound from underground trains therefore raises reasonable questions about the applicability of an A-weighting for this type of noise (Kraemer, 1984; Walker_et_al, 1997). Despite these concerns, no alternative weighting has been adopted. Noise measurements are most commonly expressed as maximum noise level (LAmax) for individual train pass-bys.

The sound measurement process involves r.m.s. averaging which takes place over a defined interval. Two averaging times have been internationally agreed (BS5969,1981), as 0.125 seconds known as fast (F) response, and over 1 second, known as slow (S) response. The use of slow or fast time response on the measuring instrument is unfortunately not always stated when measuring and evaluating groundborne noise, but when it is, it is more commonly required to be slow response. The emphasis to slow response is primarily historic, based upon the observer being more able to read the slow excursions of an analogue meter. It is relevant to note that standards which deal with airborne noise are nowadays requiring fast time response, such as; 'BS5228, part 4:1992 Noise control on construction and open sites' and 'BS4142:1995 Method for rating industrial noise affecting mixed residential and industrial areas'. Despite this trend, the railway industry favours slow time response, given that it leads to a lower level of LAmax by about 3dBA compared to fast, and is the basis on which most of the measurements in industry have been made and evaluated (see Table 7.12 of Chapter 7). Whilst LAmax is a simple descriptor, it takes no account of the duration and number of events. Some criteria (US DOT, 1995) make a simple adjustment when the number of trains exceed a certain number, for which lower limits apply.

Alternatives to LAmax are LAeq calculated over a specified period. The LAeq index is simply the equivalent continuous sound level that has the same energy as the time varying sound pressure level calculated over a specified time period (BS6698, 1986).

Most of the guide values for groundborne noise are based upon absolute levels, although it is feasible to provide guide values in terms of limiting excursion above ambient levels, in a manner which addresses audibility and nuisance. This is the philosophy adopted in BS4142 (1995), but which relates to assessing noise from industrial premises on residential areas. This standard compares the noise level from the source (specific noise level), measured using the LAeq index, and compares it with the background noise in its absence. If the noise source has a certain character that makes it more obtrusive, such as tonal or impulsive, a penalty of 5dB is added to the 'specific noise level' to obtain what is called the 'rating level'. It is the difference between the 'rating level' and the 'background level' in absence of the specific noise source that is significant. The 'background level' is obtained from a different noise index, the L90 level. This is the level that is exceeded for 90% of the time, and approaches the troughs, or minima in the time varying noise level.

If the difference between the 'rating level' and the 'background level' is 10dB or more, then complaints are likely. If the difference is 5dB, it is of marginal significance. Smaller differences give rise to a lower likelihood of compliant. Difference of (-10dB) is a positive indication that complaints are unlikely. As stated, this method of evaluation was developed to assess the effect of fixed industrial sites on residential areas, and was never intended to address transportation noise.

This approach whilst seeming to be a reasonable way to evaluate and therefore minimise nuisance, is however quite understandably objectionable to the railway industry. This is because it can impose very high and variable environmental constraints, where neighbours affected by groundborne noise could reduce background levels by attenuating other noise sources to very low levels, by for example a higher degree of airborne sound insulation. Absolute levels are therefore appropriate. These absolute levels can then be set at levels that take into account factors such as speech intelligibility, which is critical in offices, or sleep disturbance that is critical in residential situations.

The interference of noise with speech communication occurs when an interfering noise renders the desired signal of speech inaudible. The higher the levels of this interfering (masking) noise and the more energy it contains at speech frequencies, the greater will be the percentage of speech sounds that are inaudible to the listener. The redundancy of speech is such that even if a particular sound is masked or omitted, the remaining sounds are often sufficient to convey the message, although this can impose an additional strain on the listener. It is significant to note that groundborne noise from railways arise at low frequencies (<200Hz), whereas speech intelligibility is more affected by frequencies between 250Hz and 4000Hz. Table 3.4 describes how different indoor noise levels can affect communication.

Table 3.4 Effect of indoor sound level on communication (Taylor, 1994)

Indoor Sound Level dBA
Male Voice Level necessary for communication at a distance of 3.6m*
Normal Voice
Raised Voice
Very Loud Voice
Communication impossible

*For a female voice all levels should be reduced by 5dBA, and a further 6dBA for every doubling of distance.

This would indicate that a limit value of about 50dBA would allow communication in a normal office environment. The US Environmental Protection Agency (1974) are quoted by WHO (1980) that 100% sentence intelligibility can be achieved for a normal voice in a typical living room at a speaker-listener distance of 1m at a steady background noise level of less than 45dBA.

In considering disturbance to sleep it is relevant to note that noise is not the only cause of disturbance, where it is reported by Rice and Morgan (1982) that about 20% of the population suffer from sleep disturbance from causes other than noise. The most vulnerable times are said to be when people are trying to get to sleep (2200-2400) and the latter part of the night when lighter sleep stages predominate. They provide a review of studies from which they summarise values at which 25% of the population are likely to suffer sleep disturbance, shown in Table 3.5.

Table 3.5 Outdoor levels where 25% population likely to suffer sleep disturbance, from all causes including the source under consideration (Rice and Morgan, 1982)

Noise Source
Peak dB(A)
Steady Noises

On this basis they suggest a criteria for external noise levels, at 75dBA Lmax or 55dBA in terms of Leq(2200-0700), where about 25% of the population are likely to be awakened, which is only 5% above the level of awakenings that arise from other causes.

These external noise levels can be converted to an internal level, by noting the approximate attenuation of 15 to 20dBA which is assumed for a single glazed closed window. The attenuation is only about 10dBA if the window is partially open. This equates to an internal noise level of 55-60dBA Lmax and 35 - 40dBA Leq(2200-0700), for closed windows.

A more stringent set of guidelines can be taken from the World Health Organisation (WHO, 1980), which refers to studies where the probability of subjects being awakened by noise is, 5% at a peak sound level of 40dBA, with the probability increasing to 30% at 70dBA. Defining sleep disturbance as changes in sleep stage (EEG 'electroencephalogram' changes), increases these probabilities to 10 % at 40dBA and to 60% at 70dBA. They found that a difference in sensitivity to noise is related to factors such as age and sex of the subject. Adaptation was observed but only when noise stimuli are of a low intensity. A fluctuating noise level was reported to be more disturbing than an equivalent steady continuous noise. They found that sleep disturbance becomes increasingly apparent at noise levels in excess of 35dBA Leq, and therefore recommend levels significantly less than this to preserve the restorative process of sleep.

Differences between the guide values in these two sources arise, and may in part be explained by the question that each study considers. The studies by Rice and Morgan set out to address the noise level which does cause a significant problem with sleep disturbance, whereas the WHO studies addressed the question of what noise level can one be sure there is no sleep disturbance (Taylor, 1994). There is currently no formal guidance on acceptable levels to avoid sleep disturbance from groundborne noise due to railways. It is however certain that individual complaints of sleep disturbance may occur, however low the noise is kept.

It is instructive to examine how the guide values for the most recent and prominent underground rail schemes in the UK came into being (Londons' Jubilee Line Extension due for completion in the year 2000, and the proposed Crossrail scheme currently under consideration). These schemes were based upon a design objective in terms of Lmax 40dBA (slow response) for a single train, and this is taken directly from the experience of London Underground, where they found complaints predominantly occurred when levels reached and exceeded this value. They also conducted a social survey where 277 interviews were carried out, with vibration measurements at 58 houses of which full noise and vibration measurements were undertaken at 25 addresses. The significant result from this survey, albeit of a small sample, showed that 11.5% were very annoyed at an Lmax,S level of 40dBA with an increase in annoyance above this level. Whilst the detailed review of this survey by Taylor (1994) indicates a poor statistical correlation of annoyance with noise level alone, and sights improved correlation when effects of perceptible vibration and groundborne noise are considered together, it is important to note that a significant proportion were very annoyed which leads to at least a substantial proportion which must be quite annoyed as well. Yet this was considered entirely acceptable, because the degree of dissatisfaction is similar to the level that arises from noise standards being applied to other transportation sources, summarised in Table 3.6.

Table 3.6 Degrees of dissatisfaction when noise standards are applied to other transportation sources (quoted from Taylor, 1994)

Transportation source
% of people very annoyed
Aircraft noise (Heathrow)
Aircraft noise ( London City)
Road noise
Airborne Rail Noise

It is also important to note that the design aim of 40dBA Lmax applied from 0600 to 2400 which were the operational times envisaged, during which sleep disturbance would be an issue in the late evening and early hours. It was argued that suggested noise level criteria for sleep disturbance for various forms of transportation sources was at a maximum level of 50 - 55dBA Lmax, which was significantly bettered in the design aim of 40dBA. A higher standard was however to be aimed for in reference libraries, churches, theatres, schools, hospitals where reasonable and practicable. It is relevant to note that this design aim was based upon predictions using worst credible assumptions, and that errors in the accuracy of prediction were more likely to result in lower rather than higher levels (Taylor, 1994).

There is currently no guidance on measurement procedures to evaluate groundborne noise, where important factors such as relevant measurement parameter, time weighting, and measurement location in a room, or location within a building need to be agreed.

It has already been noted that slow and fast time weightings can lead to a difference of 3dBA. Measurements close to a wall or at the centre of a room can also lead to significant differences, due to standing waves, where the measurement position could coincide with a node or an anti-node.

Noise levels vary room by room and storey by storey (e.g. 2 to 3 dBA per storey height, see Kurzweil and Ungar, 1982; Sharif, 1999). Such a difference in SPL is perceptible, in line with the generally assumed subjective impression (loudness), shown in Table_3.7.

Table 3.7 Benchmarks for subjective impression (loudness) for changes in SPL

Change in SPL

(dB or dBA)

± 3
± 5
± 10
Apparent Loudness Change




Twice(or half)

as loud

* A lower SPL change may be perceived under laboratory type conditions

For an unbalanced spectrum with very low frequencies, smaller changes in SPL at low frequencies can cause larger changes in apparent loudness than is shown (Kraemer, 1984)

Whilst there is no international agreement on how to evaluate groundborne noise, there is some consensus (Walker et al, 1997) that for levels at and above 50dBA Lmax during daytime, there is likely to be significant adverse reaction. For residential situations the Lmax noise levels for which there is likely to be very little adverse comment can be taken as about 30dBA for day time, and about 25dBA during night time. These levels do not make the train pass-by inaudible, but they are considered to be reasonable. These values for low adverse comment are clearly significantly more stringent than those applied in the recent UK rail schemes, which have been referred to. However, it may be argued that since it is not practical to eliminate noise from transportation sources, one must weigh up the benefit of a transportation system to the greater part of the community against the noise impact that arises to the few.

3.5 Effects of Vibration on Equipment

The sensitivity of equipment or processes to vibration will be as varied as the variety of equipment and processes that exist. Equipment or processes are more often sensitive to differential motion arising between components. For example, IVF treatment involves injecting an egg held on one arm, whilst the needle is held on another arm which is manoeuvred to pierce the egg. Differential motion between these two components would clearly damage the egg. Vibration of a laser beam source or even vibration of an intervening prism or mirror could cause the beam to deviate on a target, such as might arise between an etching laser and a circuit board. Equipment and processes are constantly evolving and therefore their susceptibility to vibration will vary.

Early accounts concerning the problem of vibration in laboratories were given by Steffens (1970) and Ferahian and Ward (1970). Much of the equipment referred to by them may now be superseded or obsolete. However, they gave advice on practical measures to take in deploying sensitive equipment, which are still pertinent today. These are summarised as follows; (i) site the sensitive area as far away from sources of vibration either within the building or as far as possible from external sources such as road and rail traffic. (ii)The most favourable location in a building is likely to be on a solid ground floor or basement. (iii) Mechanical equipment in a building should be well balanced, and properly isolated. (iv) Doors closures may need to be used to reduce the effect of door slamming. (v) The sensitive equipment may need isolation itself. (vi) The laboratory bench will need to be robust and be clear of direct contact with walls (vii) The entire laboratory, or laboratory floor may need isolation. (viii) If located on a suspended floor, the floors should be stiff to minimise the effects of local activities.

The latter issue is now a greater problem with the lightweight trend in construction. For example, footfall induced vibration may require detailed consideration (Wyatt, 1989; Osborne and Ellis, 1990; Williams and Waldron, 1994).

Rather than qualifying an area for a particular piece of equipment it is possible to design or qualify an area for a class of instruments. Various generic vibration criteria now exist. Two widely used and almost identical criteria for laboratories and microelectronic facilities are due to Ungar et al (1990) and Gordon (1996). Amick (1997) refers to the criteria due to Gordon in full, which is reproduced here in Table 3.8 and Figure 3.7. These are called vibration criterion (VC) curves, for classes 'A' to 'E', which relate to detail size, either to particle (cell) size in pharmaceutical and medical research, or line width in case of microelectronics fabrication. The logic here is that what might be an acceptable error for a large line width may be unacceptable at much smaller geometry. These criterion curves apply over the frequency range 8Hz to 100Hz, with a constant acceleration limit below 8Hz and are based upon r.m.s. velocity measured in one third octave bands. This is by no means the only type of spectra that can be used, where constant bandwidth and spectral density may also be referred to, but the results are not interchangeable.

Table 3.8:Generic Vibration Criteria Advanced Technology Facilities (Gordon, 1996)

Criterion Curve
rms Amplitude* mm/s
Detail size microns
Description of Use
Adequate in most instances for optical microscopes to 400X, microbalances, optical balances, proximity and projection aligners, etc.
Appropriate for optical microscopes to 1000X, inpection and lithography equipment (including steppers) to 3mm line widths.
A good standard for most lithography and inspection equipment (including electron microscopes) to 1 mm detail size.
Suitable in most instances for the most demanding equipment, including electron microsopes (TEMs and SEMs) and E-Beam systems, operating to the limits of their capability.
A difficult criterion to achieve in most instances. Assumed to be adequate for the most demanding of sensitive systems, including long-path, laser based small target systems, and other systems requiring extraordinary dynamic stability.

* As measured in one-third octave bands of frequency over the range 8 to 100 Hz.

The detail size refers to the line width in the case of microelectronics fabrication, the particle (or cell) size in the case of medical and pharmaceutical research, etc.

The criteria due to Ungar et al (1990) are almost identical to Table 3.8, except for slight differences in the description of the generic classes of equipment and detail size. They also provide additional criteria for; ordinary workshops (0.8mm/sec), offices (0.4mm/sec) and operating theatres (0.1mm/sec) based upon third octave r.m.s. velocity.

It is clear that the growing trend of miniaturisation in electronics, the inroads in gene manipulation and growing use of lasers (e.g. in eye surgery or interferometers) will make some categories of equipment and processes increasingly sensitive to vibration. The sensitivity of some of these (VC-E of Table 3.8 at 0.003mm/sec) may be adversely affected by ground vibration that can arise from the swaying of trees in the wind, waves breaking on the shore, or even acoustic coupling of airborne noise, let alone the magnitude of vibration that can arise from trains.

Computers have also attracted a lot of concern when vibration arises from demolition and construction works (Boyle, 1990). This was particularly the case in the banking and finance industry, where possible corruption or loss of data evoked strong commercial concerns, as such effects may go unnoticed.

Vibration limits for computer manufacturers normally take the form of displacement limits at low frequencies, typically 5Hz to 10Hz (from 0.04mm to 0.25mm 0-pk), velocity limits in a mid frequency region 10Hz to 30Hz (2mm/sec to 10mm/sec 0-pk), or a straight transition to acceleration limits at higher frequencies (0.1g to 0.5g). Such limits are generally applicable to tonal vibration, as this is the basis of shake test that provides tolerance figures. Some manufacturers also subject products to random vibration and quote limits in terms of spectral density. A higher limit for shock, varies from 1g to 10g with a half sine pulse duration of 3msec to 11msec. The limits are applied at the support surface, whilst the equipment is operational. Higher limits are appropriate when the equipment is not operational and higher still during transit.

Concerns about malfunction are in some cases supported by conservative limits put forward by the manufacturers themselves. For example, ICL (1990) suggest limits in service 1/5th the level at which their products pass a shake test whilst in an operational state.

It should be recognised that equipment failure can arise in service due to other factors such as; electromagnetic interference, electrostatic discharge, power line irregularities, etc. Furthermore, measurements of vibration on floors due to footfalls, door slamming, dropping boxes of printing paper, or the levels of vibration on a table top in normal service, are found in some cases to be much higher than the manufacturers' limits would suggest is safe. IBM (1988) have reported experience of computers exposed to seismic excitation which showed that they appeared to tolerate high levels (0.5g), provided the false floor did not collapse. In some cases damage did occur, but was limited to loosened card and connectors, requiring simple reconnection. Some of the field knowledge is reflected in recent criteria from IBM (1990) which indicate that their computers can in fact sustain high levels of vibration, which they readily endure in a normal working environment. Some manufacturers, such as ICL (1990) on the other hand continue to suggest conservative limits shown in Figure 3.8, yet these are likely to be breached in a day to day working environment. The variety of computers available from a manufacturer, from floor mounted, desk top and portable units means that limits will be product dependent, as seen by the different shape of the curves in Figure 3.8.

A British Standard which provides recommendation for the 'Accommodation and Operating Environment of Computer Equipment' (BS7083, 1989) did not address vibration at all, despite computer manufacturers stating that vibration is one of the relevant environmental factors for reliable operation.

Vibration in buildings close to railways would ordinarily not be at a level to cause concern about computer equipment. However there are some stringent vibration limits put forward by certain computer manufacturers which may suggest that there would be a problem and a possible breach of warranty. At such levels however, it is likely that some adverse comment would also arise from occupants, and there may also be unacceptable movement of monitor screens.

A building will also house equipment associated with mechanical services, which will not be sensitive to vibration at the levels expected from railways, but may be affected adversely under earthquake excitation, a source not considered in this thesis.

The greatest difficulty in evaluating and setting criteria for any type of equipment is that vibration in buildings is a mixture of tonal, random and transient events. Whilst spectral analysis is a common form of evaluation which may use third octave, constant bandwidth or spectral density, the difficulty is how does one treat tonal, random and transient events on an equal footing. Another difficulty is how to treat simultaneous signals in different bands (Amick, 1997).

The problem stems from the fact that equipment suppliers test their equipment using either random excitation or swept sine (BS EN 60068-2-64, 1995), and there may be a separate bump/topple test. Whilst this tells us what the equipment should endure under these specific excitations the problem is how to measure and compare field measurements on a like for like basis. The industry is responding to these challenges by demanding a more unified way of measuring and reporting data, such as in a document from the USA 'Recommended Practice for Measuring and Reporting Vibration in Microelectronic Facilities' (IES-RP-CC024.1, 1995). There is also work in hand to revise International Standard ISO 8569 for 'Measurement and Evaluation of Shock and Vibration Effects on Sensitive Equipment in Buildings', under ISO Technical Committee 108/SC2/WG16.

Ahlin (1996) has proposed a new method to characterise vibration for classification of vibration environments, which is said to work for periodic, random and transient vibration. Whilst the method involves response spectrum techniques and has some merits, its disadvantage is that it is not widely used in Industry as yet and therefore may not be readily adopted by manufacturers.


This Chapter has described factors to consider in evaluating the effects of vibration on the building structure, the occupants and equipment.

Railway-induced vibration is very unlikely to cause damage to the structure, although lack of experimental research on traditional building materials and the variety of guide values in standards will leave the rail industry open to the question of accelerated or triggered damage from long term exposure.

Contrary to the advice given in BS6472 (1992), it was found that VDV guide values do not lead to an environment with low probability of adverse comment, and a case was made to BSI to initiate a revision of this standard. The current VDV guide values in the standard should therefore not be used to influence a decision to adopt base isolation at a site, and compliance with VDV guidance using measurements in a completed building should not be used to infer adequate performance of base isolation.

Certain equipment and processes were shown to be very sensitive to vibration. Vibration levels expected at lineside do not adversely affect computers, but conservative limits put forward by some manufacturers may be breached at certain sites.

Base Isolation is unlikely to be adopted over concerns of vibration damage from railways. It is more likely to be used to reduce the effects of vibration (and or groundborne noise) on occupants and make a site suitable for a class of sensitive equipment or processes.