Table of contents

In this paper, we propose a neural-network-based approach to acquiring angular acceleration from a noisy velocity signal. Our scheme consists of two cascaded neural networks: neural network I (NN I) and neural network II (NN II). NN I attenuates harmful measurement noise from the velocity input. NN II further reduces the residual noise level, and gives the one-step-ahead prediction of the final acceleration signal. As an illustrative example, we discuss the application of our method in the elevator velocity and acceleration acquisition problem. Two different kinds of neural network model are employed here, the back-propagation neural network (BP) and the adaptive-network-based fuzzy inference system (ANFIS), to act as NN I and NN II. We also compare the performances of these two neural networks using numerical simulations.

A new multiple cell photo-sensor instrument has been constructed to study the kinetic process of crystallization from solution. Tetrahydrofuran (THF) hydrate formation from water/THF solutions was chosen as an example application. The instrument simultaneously monitors the entire hydrate formation process in a group of 12 water/THF samples, by measuring four physical parameters: transmitted light intensity (Pi), the change in the polarizing state of emerging light (B), the half depth diameter of the vortex (D_v) and magnetic bar spinning rate (V_str). The appearance and location of the new THF hydrate phase were identified by analysing the following parameters plotted as a function of time: transmission, vortex and change in the polarizing state of the emerging light. The macroscopic completion of hydrate crystal growth was determined as the time taken for the magnetic bar inside the sample tube to cease movement. These studies show that hydrate crystallization from 27:1 mole ratio solutions of water/THF commence at the vapour-liquid interface and near the internal wall of the glass sample tube. This was in contrast to stoichiometric mole ratio solutions of water/THF in which THF hydrate formation commenced in the bulk liquid phase.

Clad glass fibres represent a promising medium for combined applications in medicine. Guidelines are presented for the selection of promising novel fibre material for device designs of combined acoustical and optical applications. This paper presents theories of weak guiding fibres and analogies, as well as differences, between acoustical and optical fibres and simple formulas and functions for the fibre parameters as help for practical research and development. Our results (graphically represented as figures 2-4) show from the selected fibres (tables A1 and 1) that the usual glass fibres are not as good for acoustical as for optical applications. Depending on the major area of application these results could assist designers to make better choices for their practical work.

The spectral density function of turbulent velocity fluctuations can be estimated from randomly sampled LDA data by first computing a discretized autocorrelation function (using the slotting technique) followed by a Fourier transform of this function. The spectral estimates obtained in this way have a large statistical scatter in the high-frequency range. This work focuses on a spectral estimator with a much reduced statistical scatter (approximately 1 decade), enabling the retrieval of the spectral density function up to higher frequencies. This spectral estimator is based on a modification of the standard slotting technique, known as `local scaling' or `local normalization', in conjunction with a window of variable width. A series of benchmark tests for spectral estimators indicated that this estimator yields good overall results. This paper explores the characteristics of the new spectral estimator with regard to the effects of velocity bias and the presence of uncorrelated noise in the velocity data.

A laser/EMAT system is described for non-contact monitoring of metal thickness. In a B-scan imaging configuration, quantitative time-of-flight analysis of laser-generated shear waves and longitudinal-shear mode-converted waves can be used as an effective method for thickness monitoring. Q-switched Nd:YAG laser pulses with energies of ~18 mJ were delivered to the material surface via an optical fibre and focused to a line source by a cylindrical lens. Transient acoustic waves propagated through the plate to be reflected from the far surface. Waves returning to the surface, including L-S mode-converted waves, were detected using an EMAT sensitive to in-plane motion. B-scans were generated as the sensor head was moved along the material's surface, forming a 2D intensity profile that made changes in plate thickness easy to visualize. The L-S and S-L mode-converted waves provided a method of simultaneously monitoring two different points on the far surface, enabling any changes in the material thickness to be clearly identified.

A monolithic x-ray interferometer made from silicon and a scanning tunnelling microscope have been combined and used to calibrate grating structures with periodicities of 100 nm or less. The x-ray interferometer is used as a translation stage which moves in discrete steps of 0.192 nm, the lattice spacing of the silicon (220) planes. Hence, movements are traceable to the definition of the metre and the nonlinearity associated with the optical interferometers used to measure displacement in more conventional metrological scanning probe microscopes (MSPMs) removed.

We report on the intrinsic or material quality factor (Q-factor) of a number of different materials, possible suited for the construction of low loss flexures. Clamped foil flexure membranes made from titanium, a glassy alloy, Cusil and Havar (high strength non-magnetic alloy), as well as monolithic flexure membranes made from titanium, niobium and Marval 18 (maraging) steel, have been tested. The results show that the highest intrinsic Q-factor occurs for monolithic flexure membranes constructed from niobium, with maraging steel also yielding a high intrinsic Q-factor. Other materials tested were shown to be significantly inferior. For a pendulum of mass 10 kg, with a stress safety factor of δ = 0.3, we predict a maximum Q-factor of 3.1×10⁷ for a niobium flexure, and 2.9×10⁶ for a maraging steel flexure.

The concept of a model-based analysis of key comparisons is proposed and illustrated by applying it to data from a regional key comparison of accelerometer calibrations on a scale of frequencies. A physical model of the frequency dependence of the accelerometers' sensitivities is used to calculate reference values. The parameters of the physical model are determined by weighted least squares, and the resulting model is shown to conform with the data.

Uncertainties associated with the reference values calculated by the physical model are smaller than those associated with reference values obtained by standard analysis. This can lead to a more favourable assessment of the degree of equivalence of single laboratory measurement values as expressed by calculated E_n-numbers. The degree of equivalence of single laboratory measurement values is quantitatively calculated by both model-based analysis and standard analysis, and the results obtained and their differences are discussed.

An advanced method is presented, which allows the measurement of aperture areas with diameters smaller than 50 µm. The method follows the comparator technique and compares the photocurrent of a detector covered with a reference aperture of known area with the photocurrent of the same detector equipped with the aperture under investigation. Using typical reference apertures with diameters of some millimetres, the very small measured aperture area results in extreme large ratios of measured photocurrents. This requires sophisticated detector and measurement electronics to obtain high accuracy. Significant improvement with respect to common set-ups has been achieved by applying a silicon trap detector instead of a single photodiode. The advanced set-up allows areas with diameter below 50 µm to be measured with uncertainties in the order of 10^-3. Application of the method to four ultra-small apertures with nominal diameters in the range from 300 µm down to 50 µm is presented.

Optical triangulation displacement sensors are widely used for their non-contact measurement characteristic, sub-micrometre order resolution, simple structure and long operation range. However, errors originating from surface inclination, speckle effect, light source fluctuation and detector noise restrict the wider use. In order to minimize these errors, a structure for an optical triangulation displacement sensor, which is composed of an incoherent source and a linear CCD (charge coupled device), has been proposed. However, using a linear CCD causes several problems in signal processing. Because CCDs consist of spatially discrete pixels, it is impossible to detect the peak position movement within a pixel. This is an inherently severe problem of CCDs. Therefore, a new approach is needed. In this paper, we propose an adequate signal processing algorithm to overcome the limited resolution problem of a CCD for the proposed sensor structure. The signal processing algorithm is composed of a pre-processing algorithm and a modified cross correlation algorithm. With the help of the proposed algorithm, the limited resolution problem of the CCD can be solved.

A new configuration with only one cylindrical attracting mass for the determination of Newtonian gravitational constant G is proposed. The uncertainty in G introduced by determining the separation between the attracting masses in usual configurations can be avoided naturally, due to only one cylinder being used. Furthermore, an optimum ratio of the separation between the test masses to the length of the attracting mass can greatly reduce the uncertainty in determining G due to the uncertainty of measuring the separation between the test masses.

In this paper, an experimental procedure for the estimation of the volume velocity fluctuation at the open end of the exhaust tailpipe of an internal combustion engine, from in-duct pressure measurements in the tailpipe, is described and discussed. In order to check the accuracy in the estimation of the fluctuating volume velocity, and to obtain an estimate of the sensitivity of the procedure to uncertainties in relevant data, a full non-linear one-dimensional flow calculation was used. The experimental set-up used and the complementary measurements performed are described, and comments on the quality of the results obtained and an evaluation of the difficulties found are given. The procedure is validated in an indirect way, checking the sensitivity of the results to a change in the geometry of the exhaust line and their correlation with tailpipe noise measurements. These comparisons show good agreement in the trends observed. However, the quality of the results is highly dependent on the emission conditions and the in-duct pressure measurement position.

A pulsed quartz microprobe has been developed for time-resolved, extractive sampling in a time-varying (flickering) flame. The probe has been used for tunable diode laser absorption spectroscopic detection of carbon dioxide and nitric oxide in a 10 Hz, acoustically forced, air-diluted natural gas, non-premixed flame. The design of the microprobe satisfies the design requirements of rapid temporal response, low dead volume and using chemically inert materials. For selected heights and radial locations, nitric oxide and carbon dioxide mole fractions were found to track one another through the flame cycle. Temporally resolved nitric oxide concentration profiles were also compared with Abel-transformed CH luminosity profiles. Correlation of profile maxima suggests that the probe disturbances to the flow are minimal.

In this paper we introduce a new mathematical model, which can be used to investigate the response of a long-period fibre grating to changes in ambient refractive indices of values both greater and less than that of the cladding. A numerical method to calculate the HE core and cladding modes is presented. The results show that even when the ambient refractive index is higher than that of the cladding, the coupling (resonance) wavelengths experience a measurable shift (a few nm). The result obtained suggests that LPGs coated with a material of higher refractive index than the cladding may be used as index sensors.

The spectral dependences of the retardance and of the birefringence for three polymeric sheets of polyethylene, polyamide and polyethylene-polyamide blend between 420 and 680 nm are studied. Retardance measurements are carried out using a recently developed model which allows for calibrating two phase plates simultaneously from a single adjustment of the optical system. A new parameter that expresses the retardance introduced when light crosses 1 µm of the material is found to be useful in characterizing birefringent materials and in constructing phase plates for use at single wavelengths and achromatic phase plates.

The aim of this paper is to present a high voltage pulse generator which is able to generate 3 to 30 kV pulses through a 50 Ω load impedance, lasting less than 1 ns and having a rise-time of a hundred picoseconds. The pulse repetition frequency is superior to 1 kHz. The technology of this coaxial generator based upon the technology of a high pressure gas switch is described.

Dedicated probes and attenuators were constructed in order to measure pulses delivered by the generator without perturbation. The procedure and the calibration tests are also presented.

Finally, a simulation using Spice software allows the determination of the parameters which limit the performance of the system in order to model and optimize the transient response of the generator. Simulation results are compared to experimental ones and validate the Spice model.

A narrow absorption feature in an atomic or molecular gas (such as iodine or methane) is used as the frequency reference in many stabilized lasers. As part of the stabilization scheme an optical frequency dither is applied to the laser. In optical heterodyne experiments, this dither is transferred to the RF beat signal, reducing the spectral power density and hence the signal to noise ratio over that in the absence of dither. We removed the dither by mixing the raw beat signal with a dithered local oscillator signal. When the dither waveform is matched to that of the reference laser the output signal from the mixer is rendered dither free. Application of this method to a Winters iodine-stabilized helium-neon laser reduced the bandwidth of the beat signal from 6 MHz to 390 kHz, thereby lowering the detection threshold from 5 pW of laser power to 3 pW. In addition, a simple signal detection model is developed which predicts similar threshold reductions.

This paper presents a simple system for the dynamic compensation of nonlinearity in a homodyne laser interferometer that can be used for high-precision length measurement. The computer collects two phase-quadrature signals, and calculates the DC offsets, the AC amplitudes and the difference from the phase-quadrature by the elliptical fitting with a least-squares method. The control signals for adjusting these ellipse parameters are fed into the automatic control circuit through the D/A converters so that the offsets are zero, the amplitudes are same and the phase difference is 90°. As a result, the nonlinearity is eliminated electronically. The system can be used in applications requiring the real-time compensation of nonlinearity. Experimental results demonstrate that the nonlinearity of the homodyne interferometer can be reduced to sub-nanometre over the measuring range of 100 mm.

A simple and reliable method of measurement of the very short delay (tens of nanoseconds) between the pre-ionizing and main discharge currents in a TEA CO₂ laser operated in the spiker sustainer configuration has been demonstrated. In this method light emitted from the pre-ionizing spark and the main discharge are first collected separately using optical fibres and then the delay between the onsets of their emission is measured with the help of simple opto-electronic circuits. This method is noise free, since it does not involve measurement of large pulse currents, and can be useful for the measurement of similar delays in the operation of other TE gas lasers.

High precision positioning mechanisms with accurate displacement resolutions find employment in many industrial applications. In devices such as incremental encoders, the high resolution comes at the expense of increased sensor bandwidth. This design note describes a low-cost method for implementing effective bandwidth reduction, without reducing the achievable displacement resolution.

Having approached this book with eager expectation, I have to admit I was soon disappointed. Firstly, it seems to me that the title is rather misleading: how many would think of `Analog Electrical Devices & Measurements' - Chapter 6; `Sampling, Digital Devices & Data Acquisition' - Chapter 7, or `Temperature Measurements' - Chapter 8, as Mechanical measurements? This aside, however, I find the coverage both patchy and uneven; there are some excellent and useful sections, obviously those where the authors have personal experience, such as for instance Chapter 11, `Strain Measurements', and also some useful work on Probability and Statistics and Uncertainty Analysis, but the appreciation of a problem is often too superficial, with many sources of error being overlooked or ignored.

However, the most serious problem with the book is the number of errors, some of such a fundamental nature that one is inevitably driven to consider just how much of the rest one can consider reliable.

Some instances will show what one must expect. These examples were not found by a careful trawl in depth, so must be considered as representative only.

The definition of the ohm as the resistance of a column of mercury was abandoned many years ago; a standard cell must never be used to deliver 100 mA, let alone for a few minutes, and in any case, both volt and ohm have been defined for more than ten years by quantum effects.

Problem 1.37 specifies an LVDT being driven by a d.c. voltage, when of course only a.c. can be used. (There are `d.c.' LVDTs but these contain electronics to generate the necessary a.c.) The accompanying diagram also shows the connections incorrectly and the text contains the sentence `outputs a voltage which is linear to the input'. Which input? The electrical one or the displacement? Such looseness abounds, even ignoring the use of `output' as a verb.

Page 36: `Analog describes a signal that is continuous in time'. Alas, not so, analog describes a signal where the intelligence is represented by a continuously variable level.

Page 43: Fig. 2.8 seems to show (quite clearly!) that subtracting a d.c. offset actually causes a waveform to change shape.

Page 197: `An a.c. meter indicates a true rms value for a simple periodic signal only'. Absolutely wrong: an a.c. meter (the rectified moving coil type is meant) indicates average only, whatever it may be scaled.

The authors' application of mathematical analysis to the consideration of errors is both praiseworthy and necessary but reveals their obvious preference for an analysis of errors rather than for the application of careful thought to remove sources of error first, before applying analysis to those that inevitably remain. An example of this lack of thought is revealed in Fig. 6.4, a simple multirange ammeter, where the moving coil would vaporize if the switch were ever operated under load or there were ever any jump in contact resistance. Amazingly there is no discussion of the universal shunt, nor of the necessity of a swamping series resistor.

There is very little discussion of what most people would actually think of when considering `Mechanical Measurement'. Importantly, there is no discussion on the fundamental and vital differences between a micrometer and a caliper - that the former indicates its own true zero, unless it is very seriously damaged, while the latter is subject to many errors caused by wear or by quite minor damage, a further indication of inattention to some vital considerations. The authors seem also to be unaware that micrometers can be used in at least as large a size as a caliper.

There is no mention at all of modern precision coordinate measuring machines - only one of several omissions.

The arrangement of the knife edges of a precision balance in Fig. 12.13, purporting to show a precision balance, is drastically wrong.

A pencil type pressure gauge has no spring, which rather makes nonsense of the rest of the analysis.

It is surprising and regrettable that so many errors (and there are many more) should have survived into a third edition. A major revision is called for before the book could be wholeheartedly recommended.

This book is certainly a useful addition to one's library for its chapters on probability of statistics and analysis on uncertainty, but it would be most unwise to use it for general guidance on the theory of measurements for one not already thoroughly familiar with the design of actual instruments.

J C Taylor

Measurement activities in advanced industrial countries are estimated to account for somewhere between 3% and 6% of a country's Gross National Product. A well-founded National Measurement System is crucial to economic well-being and the quality of life for all citizens. It is like the oxygen we breathe - unseen and taken for granted - but it is absolutely essential for life as we experience and know it. That all students of science and technology should understand the fundamental nature of measurement and something about its implementation using instrumentation should be self-evident to even the dullest Head of Department, Dean of Faculty or Vice Chancellor of our colleges and universities.

The subject's boredom factor long since disappeared. The science, the technology and the application of measurement systems are all exciting flagship activities in any endeavour in the manufacturing and process engineering industry, transport, medicine and bioscience, natural and built environment, and so on. Information technology, computing and control automation and quality engineering are like stranded whales without underpinning by and integration with measurement systems. Technological innovation and commercial exploitation are non-starters without advanced measurement. You can forget globalization and advanced technology if you fail in supporting a sound measurement infrastructure.

The key factor for production of measurement strength is a trained and skilled labour force which understands the need for measurement fidelity and can interact intelligently with measurement experts. Therefore the key tests to be passed for any general textbook in the area of measurement and instrumentation relate to depth on principles and breadth on implementation.

Alan Morris has updated his 1988 and 1993 editions of this measurement book to emphasize advances in the use of intelligent and distributed instrumentation, measurement reliability and safety systems, and formal standards governing instrument calibration procedures and measurement system performance.

The text is divided into two parts: principles of measurement in Part 1 (240 pages), and instruments and sensors to measure various physical quantities (200 pages). Mathematics is kept to a minimum, and worked examples and self-assessment questions (with answers) are provided. There is reasonably good coverage of the necessary topics for undergraduate introductory courses to the subject, but it has not been possible to provide in-depth coverage in all topics suitable for higher-level courses and for adequate updating by practising engineers.

Unfortunately there are two serious limitations to the book. Firstly, the two chapters on measurement errors and calibration have not kept pace with developments since 1993 with the introduction of the ISO Guide to Uncertainty of Measurement, and changes in the UK National Measurement System Infrastructure (UKAS, ISO 17025, etc). Everybody should now be able to understand intelligently (and, for some, develop) the standard table format for evaluation of uncertainties of measurement. Measurement should now be seen as a measurement range within which the true measurement value might be expected to lie with a given confidence level. Measurement is no longer just a number with an error attached. Measurement is all about how to reasonably deal with uncertainty and non-repeatability of observation in the physical world.

My second concern about the book is a lack of a good set of up-to-date references and bibliography for all chapters. For example, during the past ten years an abundance of good books on all types of sensor have appeared. The SENSOR and TEST exhibition in Nuremberg in May 2001 had some 860 exhibitors of commercial products. The book by Alan Morris contains only one inadequate chapter (23 pages) on the subject of sensors. An introductory text must lever people into the subject field and allow follow-ups.

I am afraid that during the past decade much of industry has failed badly to keep abreast of good measurement practice and advanced measurement technology and this is particularly true in Britain. Pedagogues must help to force the pace of change in a knowledge-based economy. A strong link between research and teaching is essential for this to happen.

Barry E Jones

This book is described by the authors in the Preface as an introductory text for graduates in physics and engineering as well as serving the needs of technical staff. The book shares its title with an earlier (1981) publication by the first two authors. It would be wrong, though, to think of the current book as simply a new edition, since it has obviously been completely rewritten.

The scope of the book is wide, reflecting the diversity of activity in the medical physics and bioengineering profession in the United Kingdom. The topics covered include Biomechanics, Biomaterials, Physiological Measurement, Imaging using X-rays, Radioisotopes, Magnetic Resonance and Ultrasound, the effects of ionizing and non-ionizing radiation, and associated subjects including mathematics and statistics and safety-critical systems design. Such a wide scope means that the depth of coverage for each subject is limited and somewhat patchy; however, the authors largely fulfil their aim of producing an introductory text.

Each major chapter of the book is largely self-contained, starting with a useful `introduction and objectives' section, and concluding with a set of problems consisting of a number of short questions intended to help the reader check that he or she has understood the material, and then some longer questions. Outline answers are provided for some of the longer questions. Each chapter also includes a bibliography. The structure means that the book does not work particularly well for reference. However, it is quite feasible to use one or two chapters as an introduction to a specific area.

The book is obviously intended as a teaching tool, and this is reflected in the literary style, which is accessible and easy to follow. The balance between descriptive text and mathematics varies between sections, but in view of the diverse subject matter, this is probably inevitable. The overall impression is that a good balance has been struck. In places, the authors have included some very basic physics, but, given the breadth of both the subject matter and the intended audience, this is reasonable.

There are justifiable criticisms of this book. The coverage of some clinically important areas is rather superficial - for instance, pulse oximetry, one of the most widespread physiological monitoring techniques, is covered in less than one page - whilst other less `mainstream' topics are given much greater space. Also, the overall choice of material reflects the book's British origin, and will be less relevant in countries where the links between medical physics - concerned primarily with ionizing radiation - and biomedical engineering are weaker.

That said, this is undoubtedly a valuable book, which will surely find a place in the libraries of departments of Medical Physics and Bioengineering and on the recommended reading lists for both BSc and MSc courses in this area. Practising scientists and engineers will also find the accessible introductions that the book provides a good route into a new area.

Overall, this is a useful and worthwhile book which can be recommended to practising scientists and engineers, to teachers of medical physics and bioengineering and to students.

Geoff Cusick

This book is orientated towards readers who already have an understanding of state machine logic design, assembly-language programming and basic operating systems. Each chapter contains a set of exercises and a number of pertinent and up-to-date references. All chapters except the first incorporate several worked examples.

Of the eight chapters, the first two are of an introductory nature. The first chapter gives a historical overview of computer architectures and encompasses the von Neumann instruction set architecture. In the first section I discovered that the Enigma- breaking valve-powered Colossus computer was built at Benchly and not Bletchley Park (see http://www.bletchleypark.org.uk) in the early 1940s. This unfortunate inaccuracy was not noted on the Cambridge University Press web pages and is still present in the 24-page sample version of the book on their site.

Performance models and evaluation methods are described in Chapter 2, which incorporates, amongst others, Moore's, Grosch's and Amdahl's Laws. These are used further on in the book. User instruction set design is covered comprehensively in Chapter 3, taking up 78 of the book's 318 pages. This chapter describes CISC and RISC architectural styles using the Intel x86 family as the token CISC architecture and the MIPS R2000 as the token RISC architecture.

The motivation for me reviewing this book was my interest in computer memory systems. Professor Cragon covers this topic well in Chapter 4 and the reader can learn about a plethora of memory systems including: segmented & paged virtual memory; interleaved real memory and several different types of caches.

The reader's `journey' into computer architecture implemen- tation becomes even more detailed in Chapter 5, which examines processor control design. There is a good description of microprogrammed control and microinstruction set architectures that served to remind me what progress has been made in this field since I encountered it 20 years ago.

The performance models described at the start of the book are applied to example RISC/CISC architectures in Chapter 6, which details pipelined processors and the effects, amongst others, of data dependences and interrupts on the pipeline's performance. Input/Output systems are covered in Chapter 7, including a section on queuing theory, which is used to explain performance issues associated with an example of a hard disk drive's response time.

Finally, in Chapter 8 the role of the operating system in a computer architecture is acknowledged. To achieve this, quite a lot of emphasis is given to the IBM S360 (and successors) family. However, this is balanced out with an analysis of the Intel Pentium Pro CPU, with which younger readers will be more familiar.

In conclusion, this is a good book and would make a useful addition to more traditional computer architecture textbooks. Harvey Cragon's writing style is accessible and is bolstered by useful context-enhancing `nuggets' of background information scattered throughout the book.

Fraser Dickin

This book covers the analysis methods of planar transmission line structures from basics to advanced levels. All important fundamental concepts and principles are covered as far as is possible within a text of reasonable size. The content of the book may be split into three parts.

The introduction gives a perspective of realized applications that require a thorough mathematical treatment.

In the second part the essential fundamentals of electromagnetic theory are explained. A reader with some basic knowledge of vector mathematics can easily follow the treatment of Maxwell's equations in the text. The author not only explains the rules, but also gives valuable hints on how to solve special problems, thus reducing the otherwise inevitable search of the literature.

In the third part various methods used to solve Maxwell's equations are described and applied to many to the usual realized line systems. A mathematical treatment as well as practical formulas and results of commonly used structures are provided. The examples guide the reader from the start to the solution of the problem (e.g. Schwarz-Christoffel Mapping).

This book, with its thorough mathematical treatment, addresses students of electromagnetic theory but it also addresses the engineer who is in need of knowledge and practical, easy-to-apply formulas for the various line systems.

It would have been better if a discussion of important and emerging numerical methods such as FDFD and FDTD had been included.

F J Schmückle

The appearance of this book is quite timely as it provides a much needed state-of-the-art exposition on fault detection and diagnosis, a topic of much interest to industrialists. The material included is well organized with logical and clearly identified parts; the list of references is quite comprehensive and will be of interest to readers who wish to explore a particular subject in depth.

The presentation of the subject material is clear and concise, and the contents are appropriate to postgraduate engineering students, researchers and industrialists alike. The end-of-chapter homework problems are a welcome feature as they provide opportunities for learners to reinforce what they learn by applying theory to problems, many of which are taken from realistic situations.

However, it is felt that the book would be more useful, especially to practitioners of fault detection and diagnosis, if a short chapter on background statistical techniques were provided.

Joe Au

Large Eddy Simulation (LES) is an approach to compute turbulent flows based on resolving the unsteady large-scale motion of the fluid while the impact of the small-scale turbulence on the large scales is accounted for by a sub-grid scale model. This model distinguishes LES from any other method and reduces the computational demands compared with a Direct Numerical Simulation. On the other hand, the cost typically is still at least an order of magnitude larger than for steady Reynolds-averaged computations. The LES approach is attractive when statistical turbulence models fail, when insight into the vortical dynamics or unsteady forces on a body is desired, or when additional features are involved such as large-scale mixing, particle transport, sound generation etc. In recent years the rapid increase of computer power has made LES accessible to a broader scientific community, and this is reflected in an abundance of papers on the method and its applications. Still, however, some fundamental aspects of LES are not conclusively settled, a fact residing in the intricate coupling between mathematical, physical, numerical and algorithmic issues. In this situation it is of great importance to gain an overview of the available approaches and techniques.

Pierre Sagaut, in the style of a French encyclopedist, gives a very complete and exhaustive treatment of the different kinds of sub-grid scale models which have been developed so far. After discussing the separation into resolved and unresolved scales and its application to the Navier-Stokes equations, more than 140 pages are directly devoted to the description of sub-grid scale models. They are classified according to different criteria, which helps the reader to find his or her way through the arsenal of reasonings. The theoretical framework for which these models have mostly been developed is isotropic turbulence. The required notions from classical turbulence theory are summarized together with notions from EDQNM theory in two concise and helpful appendices. Further sections deal with numerical and implementational issues, boundary conditions and validation practice. A final section assembles a few key applications, cumulating in a condensed list of some general experiences gained so far.

The book very wisely concentrates on issues particular to LES, which to a large extent is sub-grid scale modelling. Classical issues of CFD, such as numerical discretization schemes, solution procedures etc, or post-processing are not addressed. Limiting himself to incompressible, non-reactive flows, the author succeeds in describing the fundamental issues in great detail, thus laying the foundations for the understanding of more complex situations. The presentation is essentially theoretical and the reader should have some prior knowledge of turbulence theory and Fourier transforms. The text itself is well written and generally very clear. A pedagogical effort is made in several places, e.g. when an overview over a group of models is given before these are described in detail. A few typing errors and technical details should be amended in a second edition, though, such as the statement that a filter which is not a projector is invertible (p 12), but this is not detrimental to the quality of the text.

Overall the book is a very relevant contribution to the field of LES and I read it with pleasure and benefit. It constitutes a worthy reference book for scientists and engineers interested in or practising LES and may serve as a textbook for a postgraduate course on the subject.

Jochen Fröhlich

The book provides an introduction to multivariate data analysis using linear modelling and its applications to quality assessment with special emphasis on chemometrics and sensory science. The aim of the book is to help students and researchers (the problem and data `owners') to analyse their large empirical data sets with minimum prior knowledge of algebra and mathematical statistics.

The text (445 pages) is divided into four parts: overview (76 pp), methodology (156 pp), applications (122 pp) and appendices (72 pp). In the first part, the authors motivate a reader to use multivariate methods, explain concepts of quality assessment and provide `a layman's guide to multivariate data analysis'. The importance of background knowledge about the problem and good quality of input data is emphasized. A research project with the aim of obtaining new facts from an empirical data set is divided into six steps from the original question to the unfolded answer. These steps are then formally followed in all the presented examples. Finally, the principles of soft bilinear modelling (the basic data-analytic tool used in the book) are described at a glance. In the second part, the method of partial least squares regression is derived from the methods of linear least squares regression and principal component analysis, including its individual variants and applications (multivariate calibration, prediction, discrimination and classification). Individual chapters are dedicated to validation of results and experimental planning. Part three describes five specific experiments: analysis of NIR spectra, analysis of questionnaire data on the quality of the working environment, prediction of toxicity from chemical structure, quality monitoring of a sugar production process, and exploratory search for optimal conditions preventing loss of quality in stored food. Appendices in the final part provide additional information to every individual chapter. The text is supported by 97 figures, 19 tables and 114 references.

Most derivations and explanations in the book are based on examples which are described in almost all possible details except calculations: the main emphasis is on experimental design, organization of input data tables and, above all, on interpretation and validation of results. Computing itself is supposed to be performed by some available data analytic software such as, e.g., The Unscrambler (http://www.camo.no) or PLS_Toolbox in Matlab http://www.mathworks.com). However, the book provides a rather general guide independent of any specific software.

The beginner will probably benefit most from the great experience of the authors: the description of advantages and risks in multivariate data analysis is well and proportionally balanced and documented. Minimum abstraction and mathematical formalism may also bring the book closer to a wider readership. At the same time, the authors' approach is fair and responsible: they encourage readers to work independently but they clearly mark the limits beyond which the reader should seek help from a professional statistician. The absence of abstraction also does not necessarily prevent some generalization: while the reader will probably solve new problems by analogy with those described in the book, the examples presented are accompanied by lists of related problems in order to demonstrate the wide spectrum of possible applications.

The strengths of the book determine also some of its limitations: although it is intended for `laymen' and contains minimal mathematics, it is not too easy to read. Many references to other parts of the book, descriptions of the authors' intentions and directions on how to read the book (often useful) sometimes distract rather than focus the reader's attention on the main problem. A diligent reader, however, will find the book useful: beginners and data-analytic software users as a good start and introduction, more advanced users and teachers as a possible source of inspiration and a good teaching aid.

Martin Samal