• Nie Znaleziono Wyników

Fatigue assessment in FPSO mooring design using moment based Hermite Approximation

N/A
N/A
Protected

Academic year: 2021

Share "Fatigue assessment in FPSO mooring design using moment based Hermite Approximation"

Copied!
9
0
0

Pełen tekst

(1)

Abstract

Hermite models are verified to describe analytically the statistics of nonlinear responses based on the re-sponse mean, standard deviation, skewness and kur-tosis. Cothparion are made bétween the numerical fittings and data obtainéd fröm a nonlinear moor-ing simulation programme obtained for rioored 250 FCdwt FPSO. Good agreerents ä.re obtained own to the lowest levels of probability in thé simulations. This inestigation is an intermediate step in the de velopment of a complete method to calculate the statistics of nonlinear mooring loads loads in arbi-trarv wave spectra. Since the Hetmite models turn out to give ver accurate results, the main attention should be addressed to thé calculation of the statis-tical response moments in nonlinear conditions.

i

Introduction

To calculate fatigue damage or extreme loads in the mooring system of an FPSO, dynamic responses are usually defined as linear or slightly non-linear pro-cesses. lt is known however,tha.t the mooring loads

can show a signißcant nonlinear behaviour, both

from the low frequency motions of the vessel á.s well is from ihe high fréquecy chain dynamics. This

aftcts hoi h the distribution of extremes used in ul tniiate si iength problems and of the rainflow

distri-hittion which is required to perform fatigue

assess-vcor

rchtf

Mekeweg 2 2E23 CD Oft 0th - o1 i'81$

L

Fatigue assessment in FPSO mooring design

using moment based Hermite Approximation

R.H.M.Huij smans (Maritime Research Institute Netherlands) L.J.M. Adegeest (Det Norske Ventas, Høvik)

February 1998

ments. First validation of our numerical approach for realistic type of vessels is based on results from a study performed for the Royal Netherlands Navy for

an Mfrigate type of vessel. The attention in that project was focused on the global hull girder loads in the nonlinear regime The experiments comprised amongst others, fre running tests in head waves and following waves. The free-running tests were per-formed in regular and irregular waves, In this study significant nonlinear hull girder loads were observed under severe relative motion conditions either in the

aft or at the bow. Sag/hog ratio's higher than 2 were observed both in head seas and in following seas.(partly reported in Adegeest (1995)

Thé present article inVestigates the applicability of momentbased Hetnite models to describe analvt-icallv the marginal and extreme probability distri-butions and crossing rates of. a nonlinear response. These models use the first N statistical moments of the response (mean, standard deviation, skewness, kurtosis, etc.). The analytical results are compared with numerical simulation results and are found to cömpare well.

This investigation is an intermediate step in the de-velopment of a complete method to calculate the statistics of nonlinear hill girder loads in arbitrar' wave spectra. Since thè Hermite models turn out to give very accurate results, the development of higher ordér Vôlterra approach to calculate directly the statistical moments in arbitrary wave spectra in the frequency dotnaip is supported after which a

(2)

very efficient procedure to perform nonllnea tatis-tical analysis can be developed. However the knOwl-edge of the second order transfer function (QTF) of the drift forces or low frequency motions is essential

for the estimation of the distribution of the turret loads. From model tests or numerical simulations the underlying statistics is by no mans easy to be retrieved. The numerical simulations that are used in this article are based on the work of Wichers and Huijsmans (1992) and Dercksen et al (1994). Their numerical simulatiOns compare favourably with the results of model tests.

Recently Standberg (1997) discussed a procedure to obtain the underlying statistical properties using

crossbispctral techniques and a "preguessed"

linear response deconvolution for the estimation of the QTF of the wave drift forces. However for moor-ing systeths where highly nonlinear restormoor-ing forces and dynamic chain effects play a prominent role, such procedure may prove to be very rude and inac-curate. The suggested procethire of standsberg for applying simplified deconvolution techniques may i hen be used in an iterative manner.

2

Short-term statistical analysis

of nonlinear responses

A significant nonlinear behaviour of the loads on i tirrets in FPSO's has been observed in. the model tc'sts. Irregular wave tests showedconsiderable skew

atid kurtosis in the probability density functions of

tlic' t urret loads.

lliis severely nonlinear behaviour has an important itii pact on the strategy that has to be followed in a

statistical analysis of turret lOads, both of the pre-(hei ion of extreme loads and of fatigue load distri-hut ions.

\ r('view is given of different methods to perform si ¿il istica.l analyses of nonlinear responses, in

partic-ilat those which are applicable to analyze

nonlin-r t tirret loads. Subjects of interest are firstly the.

2

marginal response probability density function and secondly the extreme value and range distributions using crossing rate statistics.

2.1

Closed

form

nonlinear

statistical

models

Analytical derivations of nonlinear statistical

quan-tities such as probability densities and crossing statistics are based on the. knowledge of the linear and nonlinear frequency response functions.

Un-der certain conditions, it is possible to Un-derive closed formulations describing the statistical properties of nonlinear processes that can be described using such a set of frequency response functions.

Kac and Siegert (1947) presented the inathemati-cal fundamentals to inathemati-calculate the probability den-sity function of the secord order problem, followed by Bedrosian and Rice (1971) nd Neal (1974). Use was made of a second order Volterra modelling of

the system's response.

For practical applications, the main attention should

be focused on the distribution of the peak

val-ues and of the distribution of response ranges.

Based on the same principles

as applied in the

second order theories, referred to above, Longuet-Higgins (1963), Longuet-Longuet-Higgins (1964) and Vinje and Skjordal (1975) both showed that for slightly quadratic processes, the distribution of the extreme values can be found using the joint probability den

sity function of the variable itself and of its time derivative. The resulting expressions are presented in the form of perturbatión series expansions, called Edgeworth or Charlier series. Jensen and Feder-sen (1981) applied this method to predict extreme hull girder loads in ships. The fatigue damage in-flicted on a construction by non-Gaussian wave-induced sttesses Was analyzed by Jensen (1991). The results of these practical applications were

ob-tained on the assumption of, again, a weakly second order behaviour arid a narrow band response

(3)

The complication with the measured turret load responses is the significant contribution to the to-tal response of not only second order but also of

third order components. This phenomenon has bêen observed from model tests exemplifying the effects of chain dynamics on the low frequency motions (Huijsmans and Wichers (1992)) The higher than second order behaviour of the responses is in

con-flict vith the basic assumption made in the

an-alytical approaches above i.e. the occurrence of small second order non linearities. In another study (Adegeest (1995)) showed that neglecting the third order contribution, it was not possb1e to make ac-curate predictions of the nonlinear hull girder loads for flared hull shapes

DaIzell (1984) reported theoretical derivatiOns in-cluding comparisons with simulations foi the proba bility dénity fünctions of third order systems. The theory was based on an extension of Vinje and Skjordalîs method (1975). In a similar way use vas made of a third. order functIonal polynomial modelling or Volterra modelling in which the linear and nonlinear frequency response functions were as-suined to be available. Including third order nonlin-earities required the derivation of the joint densty of t he response and two derivatives instead of one. Yo legant way was found to approximate the max-hua and minima of the responses. In conditions with strong nonlinearities, sometimes negative probabil-irv densIties vere found which are obviously

unreal-istic results. For relatively weak nonlinerities, how-ever. good predictions could be made of the lower

uuid upper tails of the distributions.

2.2

Approximate moment-based

statisti-cal models

:\s ong as closed formulations for the statistical characteristics of third order responses do not give "ahst Ic and reliable answers in all the conditions. i Ir pId)babilistic quantities can be computed using

ai approximate statistical method. These

approxi-ita t r odds are based on the knowledge that that

the response. statistics are reasonably well described

by a limited number of spectral or statistical

re-sponse moments.

Wirsching and Light (1980) published a method to predict the fatigue damage under wi4e band

stresses. Based on spectral moments in combina-tion with the applicacombina-tion of the rain flow counting method, the fatigue damage was predicted.

Non-linear responses generally result in wider response spectra because of sub- and super-harmonic contri-butions. This by nonlinearities increased spectral broadness is accounted for by the larger higher or-der spectral moments, which are very sensitive es-pecially to super-harmonic components. \Virsching and Light presuted expressiors which only depeid on the spectral moments m0, m2 and m4.

Using Hermite moment formulations as initially pre-sented by Winterstein Ç1988), it is possible to cal-culate approximate probability density functions, crossing rates and extreme values based solely on knowledge of the statistical response moments. The basic idea of a Hermit e moment node1 is that a non-linear response y(t) is matched to a Gaussian pro-cess U(t) by an appropriate function g according to y(t) = g[U(t)J. The transforration function g is taken as a N-terms I-Termite series. The shàpe of the function is controlled by the statistical moments of the response. Winterstéin showed that uing.the first four moments. good results could be obtained for the probability density functions, spectral densi-ties, extremes and ranges for various nonlinear

pro-cesses.

2.3

Proposed method

Since the proper working of the proposed Volterra modelling was validated for hull girder loads in head waves (Adegeest 1994b). it's worthwhile to investi-gate the pp1hcability of the afòremetttiored approx-irnate ethóds to analyze the statistical propetties of the nonlinear hull girder load responses.

(4)

re-quired statistical moments in arbitrary conditions, i.e. the mean, variance, skewness and kurtosis in dif-ferent wave spectra. The Volterra modelling makes it possible to calculate these quantities directly in the frequency domain.

in summary, the following procedure is proposed to calculate the short-term statistics of nonlinear hull girder loads in arbitrary wave spectra:

Calculation of the set of first, second and third order frequency response functions using a non-linear time-domain program or from model test identification techniques Standsberg (1997). Definition of a realistic wave scatter diagram,

pec ra and operational profile.

Calculation of statistical and/or spectral mo-inents according to the ã.pproxinate higher

or-(1er Volterra modelling.

-I. Calculation of the marginal distributions,

cross-ing rates and other statistics uscross-ing

moment-based Hermite models.

Prior to fully developing the outlined method, par-ticulariv the development of the direct method to compute the statistical moments in the frequency domain, t he applicability of the Hermite models is

verified.

3

Momeit-based Hermite Models

lit lit eral lire, \Vinterstein's approach based on Her-unte polynomials is widely used and seems to have t he most consistePt theoretical background. In this sN:l ion a brief review is given of the theory on Her-.mmijte imiodels as described by \Vinterstein in. a

num-ber of papers (1987, 1988). The starting point of the derivatiofl is the assumption that tie marginal

dis-I 111)111 iùii of virtually any nonlinear response can.be miat :hed to a Gaussian process by applying an ap-l)ToPriite monotone function g. Polynomial approx-nieit tamis Io!J are defined based on a limited number

4

of statistical moments of the response. A distinction had to be made between softening responses, which are characterized by wider tails than the Gaussian distribution i.e. kurtosis a4 larger than 3, and hard-ening responses, which have narrower tails i.e. kur tosis a smaller than. 3.

3.1

Matching Hermite models with

sta-tistical moments

Given N response moments of a slightly nonlinear response y(t), the transformation to a standardized Gaussian repbnse U(t) is taken as an N-term Her-mite series (Winterstein (1987)):

= p0(t) = g(U(t))

= + cnHeni(U(t))]

(1)

where p and o are respectively the mean and Stan-dard deviation of y(t). After substitution of the first thtee Hermite polynomials by their equivalents in terms of powers of U, respectively Heo(U) = i

He1(U) = U, He2(U) = U2 - i and He3(U)

U3 - 3U, one gets

o=K[U+c3(U2-1)+c4(U3-3U)+...j

(2)

The time variable t as been omitted for brevity. The shape of the staidardized distribution i

con-trolled by the coefficients c,1. the scaling factor ensures that has unit variance. It was derived by \Vintersteîn that for 1V = 4. the approximate analyt-ical solutiön for these coefficients can be expressed in terms of the central response moments, defined as

E[(t)1 where for n = .3 the skewness a3 and n 4 the kurtosis a4 are found. The results for c3 and c4 are found by matching the following equations

(5)

24 a3 6 O4 - 3 c3 = 6c3c4 c4 2c32 + 9c42

The scaling parameter i is given by

i

K

The first order fitting is foi.ind after neglecting all second order contributions in equation 3 to 5, re-sulting in i añd a4 - -3 24 (6) a3 (7) 6

After ignoring the term 2c in equation 4, an ap-proximate so1utio for c3 and. c4 was found, the so-called second order Wiïtersteiri model:

4 + 2./i + 1.5(û4 - 3)

Matching the Hermite series coefficients with the exact, numerically computed result, slightly

dii-frent

coefficients were found, by Mansòur arid

Jensen (1.994):

Torhaug (1996) reports a third set of coefficients which were derived by \Vinterstein t al in an un published report (1994): [1 + 1.25(a4 .3)]l/3 - i. C4 = -- 10

L-

c14-3 1.43a-., i-0 0.8 (1 2) a3

fi -

0.0151a31 + 0.3a32

-

6 L

1+0.2(a4)

(13)

This set of coefficients results -in distributions which ate very similar to Mansour's and Jensen's fit.

3.2

Crossing rates arid probability

densi-ties

The mean rate-at which the stationary resporse

yo(t) crosses level Y from below is defined as vy(y).

If the Hermite series is monotone, the crossing

statistics are defined -by

vy(y)=vyo(o)±voexP(_)

(14)

where u0 is defined as max[vy(g); oc < y < which is a measure of the rate of response cycles or the 'average' response frequency. The normalized rate Liy(y)/vo approximates both the failure rate per response cycle (for extremes) and the probability thä.t a response peak exceeds level y in narrow-band

responses (for fatigue).

The tuargirial distribution of the nonlinear response i given by

py(y)o' = pv0(o) =

dg0

du

(1.5)

where (u) is the standard normal probability deny sitv function

(u)çcxP()

(16)

v/1 +15(a - 1

(4 = 30 (11) 5.8 + 2v"l ± 1.5(a4 - 3)

l'liese coefficieiits vere shown tó give better results

iii i lic tails of the ditribution and vill be used from

no\ on. /i + 1.5(c14 - 3) - i C4 18 a3 C3 = C4 C3

(6)

Since He,(u) nHe...1(u), it follows that du/do = I/(d?ja/du) equals

N -1

[K(1

+ >J(n -

1)cnllen.2(u))]

(7)

This equation should be larger than, O to satisfy the condition of monotony, or, if.N = 4:

The results for the crossing statistics and margin.l distribution depend explicitly on u. This requires that for a specified response level o the

standard-ized response g(u) has tò be inverted to tnd u(0) = g' (go).

Comparison

with

numerical

simulations

The ca1cu1ted irregular wave results of the moored FP.SO are used to verify the last step in the proposed method, i.e. the calculatiòn of probability densities and crossing rates using approximate moment-based Herinite modéls.

The degree of fitting of the different Hérmite models is verified by substituting the measured statistical niotnents into the expressions for the coefficients c3 and c4. Three different Hermite modelings defined in the previous chapter are considered:

The second order fitting defined by the coeffi-deitts given in equation S and 9.

The set of coefficients as derived by Mansour and Jensen, given in equatiOn 10 and 11. The First order coefficients derived by Winter-stein et al, given in equation 6 and 7.

Results are presented for the turret loads in head

seas. lite Qbtained statistical moments per condi-I jutis a r' list ed itt table i

6 Heading: Wave height: Peak period: 14kN] o [kÑ] Skewness Kurtosis

4.1

MarginaI probability densities

The resulting marginal distributions are plotted on linear scale see. figures 1, 2. It is clearly shown that the nonsyrnmetryin the probability densities is ac-curately predicted especially using Mansour's and Jensen's fit or Winterstein's second order fit. Down to the lowest measured probability levels the densi-ties are accurately approximated especially for the survival Wave con4ition as shown in figure 1.

Slightly less accurate results are obtainéd using the original first order fit, in particular in the tails of

the distribution.

5

Rainflow damage simulations

For sufficiently narrow bafided stress responses 'it is common to relate the growth in fatigue damageD(t) to ranges of the stress responsé y(t). Particularily Miner's rule assumes that a stress range R causes an incremental fatigue damage 5D = cR". The con-stants c and n depend on material and construction details.

For narrow-band Gaussian processes, the expected damage is equal to

E[!.D1Lirtear = E[CR''Linear

= (19)

Head seas Head seas JI. 16.0 m H5 = 5.6 m T0=13.Os

T0=8.Os

-2350'

-537 3040 680 1.027 0.496 3.929 3.392

Table 1: Measured statistical moments of the turret < 3c(1 - 3c4) (18) load on .a moored FPSO

du

(7)

using the. Hermite model formulation, a correction factor -y can be. derived which accounts för band-width and nonlinear effects such that the expected. damage for the nonlinear systems becomes

'1 D'1

-

.i..1CLt. jNonlineer

= 7E[CR]

Linear (20) with (Winterstein (1987)): -y

= (V1_E2k)fl[1+n(n_1){

c4(1 - E2)± 2E2c +

((1 _2e2)22±5)±9E4)

]] (21) in which z is the response band width. The

correc-i correc-ion facto fór a narrow band process satisfying the first order \Vinterstein model follows after substitu-tion of z 0, 1 and for c4 and c3 the first order

results ¿.s given in equations 6 and 7, hence

-y = i + n(n 1)c4 (22) Normally, n is between 2 and 5. Since we only want to illustrate the effects of nnlinèar contrIbutions in

fatigue calculations, we consider n to be constant and ecival 3 and c equal 1. Using this values, the correction factor for gamma becomes

-y=1+6c4=1+

- 3

(23)

The estimation of the factor obtained from the rainflow calculations are displayed in the next table

2.

6

Concluding remarks and

rec-ommendations

..\ wet hod was presented to perform statistical anal-ys('s of nonlinear hull girder loads. The method is

E[D]

Nonlinear

Table 2: Measured and approximated correction fac-tors fatigue life

based on the calculation of the first four statisti-cal moments of the nonlinear response in combina-tion with an approximate moment based statistical model to calculate the probability densities, crossing rates.

The mean, standard deviation, skewness and kurto-sis obtained from simulations are used as input to three different approximate moment-based Hermite

models. The resulting probability densities have been compared with the sirnuated equialents \'er good cörtiparisotis were found for the urret loads on the FPSO. Also the effect of nofl-linearity on the fatigue life was considered A.s vas shon for an ar tificial SN curve, the effect of nonlinearit on the fatiguge life was demonstrated to be quite signifi= cant.

The approximate higher order Volterra. series expan-sion is a good candidate for the. further development

of the iderttification technique as put forward by Standsberg (1997)

References

Adegeest, L. J. M. (1994a, February). Experimen-tal irivesigation of the influence of bow flare and for\ard speed on the noñlinear vertical rnotlóns, bending moments and shear forces in extreme regular waves. Technical Report 993 MEMT 32, Delft University of

Technol-Heading: - Head séas Héad seas

Wave height: 16.0 m 5.6 m Peak period: 13.0 s 8;0 S Skivness 1.027 - O.496 Ku rtosis 3.929 3.392 approx. eq.(21) 1.232 1.082 y rainfiow 1.32 1.003

(8)

ogy, Ship Hydromechanics Laboratory

Adegeest, L. J. M. (1994b). Third-order volterra modeling of ship responses based on regular wave results. Technical Report 988 MÊMT 30, Delft University of Technology, Ship Hy-dromechanics Laboratory.

Adegeest, L. J. M. (1995). Nonlinear Hull Girder Loads in Shipà. Ph. D. thesis, Deift University of Technology, The Netherlands.

Adegeest, L. J. M. (1996). Third-order volterra modelling of ship responses based on regular wave results. In Proc. of 21st Symp. on Naval Hydrodynamics. Nat. Academy Press, Wash-ingtçn DC.

Bedrosian, E. and S O. Rice (1971). The output properties of volterra systems (nonlinear sys-tens with memory) driven by harmonic and gaussian inputs. In Proc. of the IEEE,

Vol-ume 59.

Beridat. J. S. (1990). Nonlinear System Analysis L Identification from Random Data John Wi-lev and Sons.

Dalzell. J. F. (984). Approximations to the prQb-ability density of maxima and minima of the response of a noiUnear system. Technical Re-port EW-22-84, US Naval Academy; Annapo-lis. Maryland.

Dercksen, A, and Huijsmans. R. H. M. and Wich-ers . J. E. W. (1991). A Improved Method for calculating the contribution of hydrodynamic chain damping on low frequency vessel mo-t ions. In Proc of mo-the 23rd Offshore Technology

Conference 1991 paper no 6967.

Jluijsinans, R. H. M. and \Vkhers ,

J. E. W.

(1992). Computation model oti a chain turret nioorecl tanket in irregular seas. In Proc ofthe ?2,ì.d 1h Offshore Technology Conference 1991

/XI/)CV no6594.

.kiisen. .J. J. (1991). Fatigue analysis of ship hulls

tinder non-gaussian wave loads. In

.'Iurinc-8

Structures, Design, Construction and Safety, pp. 279-294. Elsevier Applied Science.

Jensen, J. J. and P. T. Pedersen (1981). Bending moriierits and shear förces in ships sailing in irregular waves. Journal of Ship Research 25, 243-251.

Kac, M. and A. J. F Siegert (1947). On the the-ory of noise in radio receivers with square law detector. Journal of Applied Physics 18,

383-397.

Longuet-Higgins, M. S. (1963). The effect of

non-linearities on statistical distributions in the

theory of sea waves. Journal of Fluid Mechan-ics 17, 459-480.

Longuet-Higgins, M. S. (1964). Modified Gaus-sian distribution for slightly nonlinear vari-ables. Radio Sci. J Res. 19, 1049-1062. Mansour, A. E. and J. Juncher Jense.n (1995).

Slightly non-linear extreme loads and load combinations. J. Ship Research Vol 39 nO. 2 pp 139-149

Neal, E. (1974). Second-order hydrodynamic

forces due to stochastic excitation. In Proc. of 10th Symp. on Naval Hydrodynamics. Standsberg , C. (1997). Linear and nonlinear

sys-tem identification iti model testing. In Proc. of OTRC workshop on Nonlinear Design .4s-pects of Model tests.

Torhaug, R. (1996). Extreme response of nonlin-ear Ocean structures: Identification of minimal stochastic wave input for time domain simu-lation. Ph. D. thesis. Stanford University. \Tinje T. and S. O. Skjordal (1975). On the

calcu-lation of the statistical distribution of maxima and miiima of slightly noii-linear, quadratic, stationary stochastic variables. International Shipbuilding Progress 22, 265-274.

Winterstein, S. R. (1957). .\Iomentbased herinite models of raíidoin vibration. Technical Report R-219, Technical University ofDenmark. Lyti-g by.

(9)

\Vinterstein, S. R. (1988). Nonlinear vibration models for extremes and fatigue. Journal of

Eng. Mech. , ASCE 114(10), 1772-1790.

Winterstein, S. R., C. H. Lange, and S.

Ku-mar (1994). FITTING: A subroutine to fit four möment probability distributions to data. Technical Report RMS-14, Reliability Marine

Structures Prograri, Civil Eng. Dept., Stan-ford UniveritV.

\Virsching. P. H. and M. C. Light (1980). Fatigue

under wide band random stresses. Journàl of the Structural Division, ASCE 106(ST7),

1.59:3-1607.

-3

1(.

tctt

ri-haI,Iit iit. otwrr Iöids t ¿, ni

itisrc min fr.,rn SiniuIatiOnS

Henrut .\ppro. 2nJ crdrWiecs'in

t Appras.. i rst otder Winerstein

M&czist,ur .ter.scit

Figure I: Probability density funçtions of the turret load for an moored FPSO in head waves of 16.0 rn

sign. waVe height

Tr.-hthditv rstr rttirr icdc S n ni

j' ',d

-n5,.

-

t tiz,,.,rzj'i Sir.ut,tic,n

t tr?n; r- .-\pprQs.. 2 nJ .rdr W irt- n

- - - t ..,i r t ri rd er \ in,.r.t I' - - - ?¼1çri 5t,LII .Ii:ri-cn

'igure 2: Probabilit density functions of the turret load for ari moored FPSO in head waves of .6 in sign. wave height

Cytaty

Powiązane dokumenty

In [3], the approximate solutions of the standard 3 × 3 Euler equations are proved to satisfy stability properties as soon as a relaxation scheme is used.. In the present work,

Abstract. Some new inequalities of Hermite–Hadamard type for GA-convex functions defined on positive intervals are given. Refinements and weighted version of known inequalities

A large collector drop of radius R and terminal velocity V(R) falls through a volume containing many smaller drops of radius r and terminal velocity V(r).. In some time interval

Calibration of such a model against the impulse response function obtained from time series analysis at observation wells can be achieved by computing head values at the

With the help of Theorem 7 and the Lemma, we can easily prove Theorem 9.. , be some sequences of complex numbers. .) be an asymptotic sequence. .)... By using the previous Lemma, we

Numerical results on uniform and adaptive grids are shown and compared with the biquadratic Lagrange interpolation introduced in (Campos Pinto and Mehrenberger, 2004) in the case of

Solid Edge® software for Wiring Design enables the creation of fully functional and manufactur- able designs in a seamless mechanical computer-aided design (MCAD) and

In this paper I argue that the specific musical authenticity of improvisation in different kinds of music (especially, but not only, in Jazz and Free improvisation