Abstract

A technique based on the Wiener path integral (WPI) is developed for determining the stochastic response of diverse nonlinear systems with fractional derivative elements. Specifically, a reduced-order WPI formulation is proposed, which can be construed as an approximation-free dimension reduction approach that renders the associated computational cost independent of the total number of stochastic dimensions of the problem. In fact, the herein developed technique can determine, directly, any lower-dimensional joint response probability density function corresponding to a subset only of the response vector components. This is done by utilizing an appropriate combination of fixed and free boundary conditions in the related variational, functional minimization, problem. Notably, the reduced-order WPI formulation is particularly advantageous for problems where the interest lies in few only specific degrees-of-freedom whose stochastic response is critical for the design and optimization of the overall system. An indicative numerical example is considered pertaining to a stochastically excited tuned mass-damper-inerter nonlinear system with a fractional derivative element. Comparisons with relevant Monte Carlo simulation data demonstrate the accuracy and computational efficiency of the technique.

1 Introduction

Various methodologies have been developed over the last six decades in the field of stochastic engineering dynamics for solving the governing stochastic differential equations of motion and for determining response statistics of diverse structural and mechanical systems; see Refs. [14] for a broad perspective. Nevertheless, in many cases these methodologies are limited to treating only conventional forms of the equations of motion based on classical mechanics theories. In fact, generalizing the aforementioned solution approaches to account for systems exhibiting time- and space-localized behaviors described by operators based on wavelets and/or noninteger order (fractional) derivatives can be a rather challenging task (e.g., Refs. [57]).

Specifically, fractional calculus can be construed as a generalization of ordinary calculus (e.g., Ref. [8]) that has been successfully employed, indicatively, for developing nonlocal continuum mechanics theories (e.g., Refs. [9,10]), and for modeling viscoelastic materials (e.g., Ref. [11]). Further, a wide range of methodologies, including extensions of the standard statistical linearization solution scheme [2], have been developed over the past few years for determining the stochastic response of nonlinear systems endowed with fractional derivative terms (e.g., Refs. [1216]). Note, however, that in many cases the performance of the aforementioned methodologies is hindered by the associated significant computational cost and/or the quite high degree of the involved approximations. For example, a statistical linearization solution treatment was proposed in Ref. [16], where the original single-degree-of-freedom (single-DOF) oscillator with a fractional derivative element was cast, equivalently, into a multi-DOF system with integer-order derivatives only. This was done by augmenting the response vector and by considering additional auxiliary state variables and equations; see also Ref. [17]. Inevitably, the increased dimensionality of the problem due to the augmentation of the response vector translated into nontrivial additional computational cost. Further, due to the Gaussian response assumption [2], a statistical linearization solution treatment exhibits satisfactory accuracy in determining only first- and second-order response statistics (i.e., mean vector and covariance matrix), and not the complete joint probability density function (PDF).

Over the past few years, Kougioumtzoglou and coworkers have developed a Wiener path integral (WPI) technique for determining the stochastic response of diverse nonlinear dynamical systems (e.g., Refs. [1820]). The fundamental concept of the WPI technique, which exhibits both low computational cost and high accuracy, relates to treating the system response joint transition PDF as a functional integral over the space of all possible paths connecting the initial and the final states of the response vector. Further, the functional integral is evaluated, ordinarily, by resorting to an approximate approach that considers the contribution only of the most probable path. This corresponds to an extremum of the functional integrand and is determined by solving a functional minimization problem that takes the form of a deterministic boundary value problem (BVP).

Notably, the WPI technique has been extended recently to treat systems endowed with fractional derivative elements; see Refs. [2123]. Specifically, a single-DOF nonlinear oscillator with a fractional derivative element was considered in Ref. [21] that was cast, equivalently, into a multi-DOF nonlinear system with integer-order derivatives only. Next, to circumvent the challenge of increased dimensionality of the problem due to the augmentation of the response vector, an efficient reduced-order WPI variational formulation was employed that was originally developed in Ref. [20]. This reduced-order formulation can be construed as an approximation-free dimension reduction approach that renders the associated computational cost independent of the total number of stochastic dimensions of the problem. Concisely stated, the technique in Ref. [20] can determine directly any lower-dimensional joint response PDF corresponding to a subset only of the response vector components by utilizing an appropriate combination of fixed and free boundary conditions. Thus, the original single-DOF oscillator joint PDF corresponding to the response displacement and velocity was determined efficiently, while circumventing the rather challenging task of treating directly equations of motion involving fractional derivatives. Further, a conceptually different and more direct approach was proposed in Ref. [22], where the derived closed-form expression for the WPI-based joint response PDF involved the fractional derivative terms of the original equation of motion. This led to a fractional variational problem that was solved numerically based on a Rayleigh-Ritz scheme to determine the most probable path, and ultimately, to obtain the system response joint PDF. Furthermore, based on a functional change of variables treatment, it was proved in Ref. [23] that the mathematical formulation of Ref. [22] can be generalized to treat a broader class of systems with non-Markovian response processes. In fact, it was shown that the WPI technique can treat nonlinear systems with a history-dependent state, such as hysteretic structures or oscillators endowed with fractional derivative elements, in a straightforward manner; that is, without resorting to ad hoc modifications pertaining, typically, to employing additional auxiliary filter equations and state variables.

In this paper, the reduced-order WPI formulation of Ref. [20] is extended to account for systems endowed with fractional derivative elements. Alternatively, the herein developed technique can be construed as a generalization of the results in Ref. [22] to account for mixed boundary conditions in the variational formulation. An indicative numerical example is considered pertaining to a stochastically excited tuned mass-damper-inerter nonlinear system with a fractional derivative element. Comparisons with relevant Monte Carlo simulation (MCS) data are included as well for demonstrating the accuracy and computational efficiency of the reduced-order formulation.

2 Preliminaries

2.1 Stochastic Differential Equations of Motion.

In this section, some basic aspects of the theory of stochastic differential equations (SDEs) are presented for completeness. The interested reader is also directed to Refs. [24] and [25] for a broader perspective.

Consider a multidimensional first-order SDE in the form
(1)
where α=[αj]n×1 is an ndimensional stochastic vector process whose components αj, j{1,,n}, are scalar stochastic processes; A(α,t) is the drift vector; B(α,t) is the diffusion matrix; and W denotes the Wiener process, which is a Markov process with independent increments; i.e., W(tl+1)=W(tl)+ΔW(tl,tl+1),l0, with ΔW(tl,tl+1) being statistically independent from any other increment corresponding to different time instants. Further, W has continuous, nowhere differentiable, sample paths. Notwithstanding some loss of mathematical rigor, Eq. (1), is written alternatively as
(2)

where the dot above a variable denotes differentiation with respect to time t, and η denotes a zero-mean and delta-correlated Gaussian white noise stochastic process of intensity one. That is, E[ηj(t)]=0 and E[ηj(tl)ηk(tl+1)]=δjkδ(tltl+1) for any j,k{1,,n}, where δjk is the Kronecker delta and δ(t) is the Dirac delta function. Regarding the relation between the Wiener and the white noise processes, η(t) can be defined as an infinitesimal jump of the Wiener process, i.e.,, η(t)dt=dW. Thus, it is often, informally, written as the time derivative of the Wiener process in the form η(t)=dWdt; see also Refs. [24] and [26] for a more detailed discussion on the topic.

Further, in the field of stochastic engineering dynamics (e.g., Ref. [3]), the equations of motion pertaining to diverse mDOF structural and mechanical systems take the form of second-order SDEs given by
(3)
where x denotes the displacement vector process (x=[x1,,xm]T); M is the m × m mass matrix; and g(·) represents a nonlinear vector function of arbitrary form that can also account for possible dependence of the state of the system on its history. Furthermore, w(t) is a white noise stochastic vector process with E[w(t)]=0 and E[w(t)wT(t+τ)]=Swδ(τ), where Swm×m is a nonsingular symmetric matrix. Next, it is readily seen that Eq. (3) can be cast, equivalently, in the form of Eq. (2) by employing a state variable formulation with n=2m, i.e.
(4)
(5)
and
(6)

where the square root of matrix Sw is given by SwSwT=Sw.

2.2 Wiener Path Integral Formalism.

In this section, the salient aspects of a recently developed novel formalism of the WPI technique for determining the stochastic response of diverse dynamical systems are presented concisely for completeness; see Ref. [23] for more details. In comparison to alternative derivations in the literature, which resort to the Chapman–Kolmogorov equation as the starting point, the novel formalism circumvents the Markovian assumption for the system response process. In this regard, nonlinear systems with a history-dependent state, such as hysteretic structures or oscillators endowed with fractional derivative elements, can be treated in a direct manner; that is, without resorting to ad hoc modifications of the WPI technique pertaining, typically, to employing additional auxiliary filter equations and state variables (e.g., Ref. [21]).

Specifically, it has been shown (e.g., Ref. [27]) that the probability of a path corresponding to the ndimensional Wiener vector process of Eq. (1) with W(t0)=W0,W(tf)=Wf and ΔWl=Wl+1Wl is given by
(7)
where the time domain is discretized into L +2 points ε apart (with L as ε0), i.e., ti=t0<<tL+1=tf. Further, considering Eq. (7) and accounting for the probabilities of all possible paths that the Wiener process W can follow, the corresponding transition PDF is given as the limit of an Ldimensional integral in the form
(8)

Next, taking into account the relationship between the system response process α and the Wiener process W described by Eq. (1), the transition PDF p(αf,tf|αi,ti) is obtained by applying a functional change of variables to Eq. (8); see Ref. [23] for more details.

In this regard, denoting the set of all paths with initial state αi at time ti and final state αf at time tf by C{αi,ti;αf,tf}, the joint transition PDF p(αf,tf|αi,ti) takes the form of a functional integral over C{αi,ti;αf,tf}, i.e.,
(9)
where the stochastic action S[α̇,α] is given by
(10)
and the functional measure D[α(t)] takes the form
(11)
Further, the Lagrangian functional L[α̇,α] in Eq. (10) is expressed as
(12)
where B̃(α,t)=B(α,t)BT(α,t). Furthermore, considering Eq. (3) and expressing α in terms of (x,ẋ), the transition PDF of Eq. (9) becomes
(13)
where
(14)
The Lagrangian functional L[x,ẋ,ẍ] in Eq. (14) takes the form
(15)
and
(16)

2.3 Most Probable Path Approximation.

It is remarked that the analytical evaluation of the WPI of Eq. (13) is, in general, an impossible task. Thus, alternative approaches are typically pursued in the literature for evaluating approximately Eq. (13), such as the most probable path approach (e.g., Ref. [27]). Note that the most probable path approximation has exhibited a quite high degree of accuracy in various diverse engineering mechanics applications (e.g., Refs. [28] and [29]).

Specifically, the largest contribution to the functional integral of Eq. (13) relates to the trajectory xc(t) for which the stochastic action of Eq. (14) becomes as small as possible. This leads to the variational (functional minimization) problem
(17)
Following the solution of Eq. (17) and determination of xc(t), the functional integral of Eq. (13) is evaluated approximately as
(18)
where C is a constant to be determined by the normalization condition
(19)

In passing, it is remarked that in cases where an even higher degree of accuracy is required (e.g., reliability assessment applications), fluctuations around the most probable path can be also accounted for (e.g., Ref. [19,30]). Further, various methodologies can be employed for treating the optimization problem of Eq. (17) and for determining xc(t). These range from standard Rayleigh-Ritz type numerical solution schemes (e.g., Refs. [28,31]) to more recently developed techniques relying on computational algebraic geometry tools [32]. Alternatively, considering Eq. (17) and resorting to calculus of variations yields the corresponding Euler–Lagrange equations, which take the form of a BVP to be solved for obtaining the most probable path xc(t) (e.g., Ref. [33]).

3 Mathematical Formulation

In this section, the WPI technique is generalized by developing a variational formulation with mixed fixed/free boundary conditions for determining the stochastic response of diverse nonlinear systems endowed with fractional derivative elements. The herein developed technique can be construed as an extension of the methodology proposed in Ref. [20] to account for fractional derivatives in the system equations of motion.

Specifically, the developed technique is capable of determining directly (i.e., without computing first the 2mdimensional joint response PDF and subsequently marginalizing) any lower-dimensional joint response PDF p(u,tf|xi,ẋi,ti), where u denotes a pdimensional vector with 1p2m and
(20)

In Eq. (20), U0 and U1 are sets containing the indices of the coordinates of x and ẋ, respectively, that participate in the lower-dimensional joint response PDF. In this regard, the boundary conditions at tf are considered fixed for these indices, while the rest are treated as free. Notably, the associated variational formulation of the WPI technique allows for determining the pdimensional, marginalized, joint response PDF p(u,tf|xi,ẋi,ti) in a direct manner at a reduced computational cost. In fact, the cost depends only on the reduced dimension p of the target PDF and not on the dimension 2 m of the original system.

In the ensuing analysis, consider an mDOF dynamical system endowed with fractional derivative elements. In particular, Eq. (3) takes the form
(21)
where h(·) denotes an arbitrary nonlinear vector function, C is a constant parameter matrix, and tiCDtax represents the left Caputo fractional derivative of order a defined as [8,34]
(22)

with Γ(·) denoting the Gamma function.

3.1 Variational formulation and Euler–Lagrange Equations With Mixed Fixed/Free Boundary Conditions.

Considering Eq. (21), Eq. (14) takes the form
(23)
and the Lagrangian functional of Eq. (15) becomes
(24)
Further, the most probable path xc(t), being an extremal of S[x,tiCDtax,ẋ,ẍ], can be determined by resorting to calculus of variations [35] and enforcing the necessary condition that the first variation δS[x,tiCDtax,ẋ,ẍ] vanishes, i.e.,
(25)
In this regard, x(t) is expressed as (e.g., Refs. [3638])
(26)
where ε denotes a parameter of relatively small magnitude, and y(t) is an arbitrary function defined so that the boundary conditions of Eq. (20) are satisfied. Substituting Eq. (26) into Eq. (23) yields
(27)
where the linearity of the fractional derivative operator has been utilized [8]. Note that
(28)
is a necessary condition for S to have an extremal for x=xc and for all admissible functions y(t). Further, considering Eq. (27), Eq. (28) becomes
(29)
Next, integrating by parts the last two terms in Eq. (29) yields
(30)
and
(31)
whereas applying integration by parts to the second term in Eq. (29) yields (e.g., Refs. [34,37,39])
(32)
In Eq. (32), tRLDtfa denotes the right Riemann–Liouville fractional derivative defined, for an arbitrary function f(t), as
(33)
Next, substituting Eqs. (30)(32) into Eq. (29) yields
(34)
Further, considering fixed boundary conditions at ti and mixed fixed/free conditions at tf as in Eq. (20) yields for j=1,,m
(35)
Taking into account Eq. (35) in conjunction with the representation of Eq. (26) leads to
(36)
Note that the boundaries xj(tf),jU0, and ẋj(tf),jU1, in Eq. (35) or, equivalently, yj(tf),jU0, and ẏj(tf),jU1, in Eq. (36) are considered free. It is readily seen that for the special case where U0=U1={1,,m}, all final conditions at tf are considered fixed. Thus, Eq. (36) becomes yj(ti)=ẏj(ti)=yj(tf)=ẏj(tf)=0 for j=1,,m and the first summation in Eq. (34) becomes zero, which, combined with the fundamental lemma of calculus of variations, yields the system of Euler–Lagrange equations for j=1,,m
(37)
Note that according to the fundamental lemma of calculus of variations the derived Euler–Lagrange Eq. (37) is satisfied also by the general class of functions x(t) with arbitrary boundary conditions. In this regard, for the general case of mixed fixed/free boundary conditions described by Eq. (35) or, equivalently, Eq. (36), there are terms in the first summation of Eq. (34) that do not vanish for jU0 and jU1. In summary, Eq. (34) leads to the Euler–Lagrange Eq. (37) in conjunction with the boundary conditions
(38)
Lastly, following the solution of the BVP of Eqs. (37) and (38), a specific point of the (lower-dimensional) system response joint transition PDF is determined based on Eq. (18) as
(39)

It is remarked that the herein developed technique is an extension of the reduced-order WPI formulation in Ref. [20] to account for systems endowed with fractional derivative elements. Alternatively, the technique can be construed as a generalization of the results in Ref. [22] to account for mixed boundary conditions in the variational formulation. Indeed, it can be readily seen that for the special case of all boundary conditions being fixed, i.e., U0=U1={1,,m}, the technique reduces to the formulation in Ref. [22].

3.2 Computational Aspects.

The BVP of Eqs. (37) and (38) is not, in general, amenable to analytical solution treatment. In fact, the problem exhibits a relatively high degree of mathematical complexity, particularly due to the presence of both left Caputo and right Riemann–Liouville fractional derivatives. Although there have been relevant efforts in the literature for treating this kind of problem analytically by resorting to solution representations based on appropriate series expansions (e.g., Ref. [40]), numerical schemes are usually more versatile (e.g., Ref. [41]). Alternatively, the complexity of the BVP of Eqs. (37) and (38) can be reduced drastically by approximating the left Caputo fractional derivative in the Lagrangian functional of Eq. (24) by the following expression involving integer-order derivatives only [42], i.e.,
(40)
where
(41)

Truncating the expression in Eq. (40) after a number of terms and substituting into the Lagrangian of Eq. (24) yields a BVP to be solved for the most probable path that involves integer order derivatives only. This can be readily solved by a plethora of well-established numerical schemes (e.g., Ref. [43]).

Further, as noted in Sec. 3.1, for the special case where U0=U1={1,,m}, all final conditions at tf in Eq. (38) become fixed, and Eq. (39) corresponds to the complete 2mdimensional joint PDF. In this regard, for a given time instant tf, a standard brute-force numerical implementation of the technique entails the discretization of the PDF effective domain into N2m points, where N is the number of points in each dimension, and the evaluation of the PDF is performed point-wise on the discretized lattice. In other words, N2m BVPs in the form of Eqs. (37) and (38) need to be solved numerically, yielding an exponential increase of the computational cost with the dimension m of the system.

Remarkably, the variational formulation with mixed fixed/free boundaries developed in Sec. 3.1, which extends the technique in Ref. [20] to account for fractional derivative elements, can reduce the associated computational cost drastically by determining directly any lower-dimensional (p2m) joint response PDF. This capability is particularly advantageous for problems where the interest lies in few only specific DOF whose stochastic response is critical for the design and optimization of the overall mDOF system. In this regard, the N2m BVPs required to be solved by the standard formulation of the technique decrease to Np problems only, where 30<N<50 is a reasonable range of values for various diverse engineering dynamics applications (e.g., Refs. [22,23,29]). Note that for small values of p relative to 2 m, it has been shown in Ref. [20] that the reduced-order WPI technique becomes orders of magnitude more efficient than an alternative MCS solution treatment.

3.3 Mechanization of the Numerical Implementation.

Concisely stated, the mechanization of the numerical implementation of the herein developed technique comprises the following steps:

  • For a given time instant tf, select sets U0 and U1 corresponding to the system response variables of interest.

  • Discretize the pdimensional domain, where p=|U0|+|U1|, with N points in each dimension.

  • For each point of the discretized domain, determine the most probable path xc(t) by solving numerically Eq. (37) in conjunction with the boundary conditions of Eq. (38).

  • Obtain the pdimensional joint response PDF p(u,tf|xi,ẋi,ti) by employing Eq. (39).

4 Numerical Example

An indicative numerical example pertaining to a stochastically excited tuned mass-damper-inerter (TMDI) nonlinear system with a fractional derivative element is considered next for demonstrating the efficiency and accuracy of the herein developed reduced-order WPI technique.

Specifically, TMDI systems can be construed as generalizations of the classical tuned mass-damper (TMD) concept for passive vibration control applications. In fact, TMDI configurations have demonstrated enhanced performance compared to standard TMD systems by exploiting the mass amplification effect of the inerter, a two-terminal flywheel device developing forces proportional to the relative acceleration of its terminals. The interested reader is directed to the recent review papers [44] and [45] for a broad perspective.

Notably, the majority of research efforts for the optimization of stochastically excited TMDI structures refer to linear systems for which convenient closed-form input–output relationships exist in the time or frequency domains (e.g., Refs. [46,47]). However, in many structural engineering applications, the primary structure (or the combined system) exhibits nonlinear/hysteretic response behavior (e.g., Refs. [45,48,49]). In such cases, employing a standard MCS-based solution treatment of the equations of motion translates into performing a relatively large number of computationally demanding time-domain response time-history analyses. In fact, the computational cost as a function of the number of MCS-based response analyses increases significantly and can become prohibitive when constraints pertaining to low probability events (e.g., failures) are considered in the TMDI optimization problem. Further, alternative, computationally efficient approximate techniques, such as statistical linearization [2], exhibit typically a low degree of accuracy and are incapable of treating optimization problems with low probability constraints. Indeed, as shown in Refs. [29] and [50] where a class of nonlinear electromechanical harvesters was considered, statistical linearization was strikingly outperformed in terms of accuracy by a WPI-based solution treatment in determining the system response PDF.

In the following, to demonstrate that the reduced-order WPI technique developed herein is capable of addressing the aforementioned challenges, the 2-DOF TMDI system shown in Fig. 1 is considered next. Its equation of motion is given by Eq. (21) with x=[x1,xTMD]T,Da[x]=[ẋ1,tiCDtfaxTMD]T
(42)
(43)
(44)
and
(45)

where the values m1=2000kg,c1=21980Ns/m,k1=35×105N/m,mTMD=1800kg,cTMD=15716Ns/m,kTMD=4.49×105N/m,b=2000kg,S0=0.153, a =0.5 and ε=0.1 are used for the parameters in Eqs. (42)(45). The restoring force of the primary structure exhibits cubic stiffness nonlinearities, whereas the secondary system is connected to the primary structure with a viscoelastic element whose force is modeled by a fractional derivative term as in Ref. [51].

Fig. 1
Stochastically excited tuned mass-damper-inerter (TMDI) nonlinear system with a fractional derivative element
Fig. 1
Stochastically excited tuned mass-damper-inerter (TMDI) nonlinear system with a fractional derivative element
Close modal

Next, in the ensuing implementation of the reduced-order WPI technique, only the response displacement of the primary structure, i.e., x1, is considered fixed at the final time instant tf. Thus, according to Eq. (39), the one-dimensional response PDF p(x1,f,tf|x1,i,xTMD,i,ẋ1,i,ẋTMD,i,ti) with fixed zero initial conditions at ti is determined in a direct manner by solving only N =41 BVPs of the form of Eqs. (37) and (38). For the numerical solution of the BVPs, the approximation of Eq. (40) is adopted and the expansion is truncated at n =2. Also, the initial time instant is set at ti=0.6s, and results are obtained and plotted for the relative final time instant tf=tfti. In particular, Fig. 2 shows the evolution in time of the marginal response displacement PDF p(x1,f,tf|x1,i,xTMD,i,ẋ1,i,ẋTMD,i,ti). Comparisons with MCS-based estimates (50,000 realizations) demonstrate a satisfactory degree of accuracy. For the MCS analyses, the L1 algorithm (e.g., Ref. [52]) has been used for integrating numerically Eq. (21) and for determining response realizations. Further, the quite high accuracy exhibited by the WPI technique is also shown in Fig. 3, where the response displacement PDF at various time instants is compared with pertinent MCS-based results.

Fig. 2
Evolution in time of the response displacement PDF of the primary structure of a stochastically excited TMDI nonlinear system with a fractional derivative element: (a) WPI-based estimates and (b) MCS data—50,000 realizations
Fig. 2
Evolution in time of the response displacement PDF of the primary structure of a stochastically excited TMDI nonlinear system with a fractional derivative element: (a) WPI-based estimates and (b) MCS data—50,000 realizations
Close modal
Fig. 3
Wiener path integral (WPI)-based response displacement PDF of the primary structure of a stochastically excited TMDI nonlinear system with a fractional derivative element at various time instants; comparisons with MCS data—50,000 realizations
Fig. 3
Wiener path integral (WPI)-based response displacement PDF of the primary structure of a stochastically excited TMDI nonlinear system with a fractional derivative element at various time instants; comparisons with MCS data—50,000 realizations
Close modal

Remarkably, compared to a standard implementation of the WPI technique with all boundary conditions fixed, which requires N4=414=2,825,761 BVPs to be solved and yields the four-dimensional joint PDF p(x1,f,xTMD,f,ẋ1,f,ẋTMD,f,tf|x1,i,xTMD,i,ẋ1,i,ẋTMD,i,ti), the proposed reduced-order implementation minimizes the associated computational cost by obtaining the marginal PDF p(x1,f,tf|x1,i,xTMD,i,ẋ1,i,ẋTMD,i,ti) directly by solving only N =41 BVPs. Notably, for this particular example, MCS based on 50,000 realizations requires approximately 15 min of computation time, whereas the marginal PDF of x1 at a specific time instant is determined via the proposed reduced-order WPI technique in approximately 4 s on the same computer.

Lastly, it can be argued that the reduced-order WPI technique is ideally suited for the considered engineering dynamics problem of vibration control, where the fundamental objective relates to determining and controlling, specifically, the DOF corresponding to the response displacement of the primary structure.

5 Concluding Remarks

In this paper, a reduced-order WPI technique has been developed for determining the stochastic response of nonlinear systems with fractional derivative elements. The technique can be construed as an extension of the formulation in Ref. [20] to account for fractional derivative elements in the system equation of motion. Alternatively, it can be viewed as a generalization of the formulation in Ref. [22] to yield enhanced computational efficiency.

Specifically, the developed technique constitutes an approximation-free dimension reduction approach that renders the associated computational cost independent of the total number of stochastic dimensions of the problem. In this regard, it can determine directly any lower-dimensional joint response PDF corresponding to a subset only of the response vector components. This is done by utilizing an appropriate combination of fixed and free boundary conditions in the related functional minimization problem. An indicative numerical example pertaining to a stochastically excited TMDI nonlinear system with a fractional derivative element has been considered for demonstrating the accuracy and efficiency of the technique compared to relevant MCS data.

Overall, it has been shown that the reduced-order WPI technique is ideally suited for problems where the objective relates to determining the response of few only specific DOF that are critical for the design and optimization of the overall system.

Acknowledgment

I. A. Kougioumtzoglou gratefully acknowledges the support through his CAREER award by the CMMI Division of the National Science Foundation (Award No. 1748537).

Funding Data

  • Directorate for Engineering (Award No. 1748537; Funder ID: 10.13039/100000084).

Data Availability Statement

The datasets generated and supporting the findings of this article are obtainable from the corresponding author upon reasonable request.

References

1.
Lin
,
Y. K.
,
1967
,
Probabilistic Theory of Structural Dynamics
,
McGraw-Hill
, New York.
2.
Roberts
,
J. B.
, and
Spanos
,
P. D.
,
2003
,
Random Vibration and Statistical Linearization
,
Dover Publications
, Mineola, NY.
3.
Li
,
J.
, and
Chen
,
J.
,
2009
,
Stochastic Dynamics of Structures
,
Wiley, Ltd
.,
Chichester, UK
.
4.
Grigoriu
,
M.
,
2012
,
Stochastic Systems: Uncertainty Quantification and Propagation
, Springer Series in Reliability Engineering,
Springer
,
London
.
5.
Spanos
,
P. D.
,
Kong
,
F.
,
Li
,
J.
, and
Kougioumtzoglou
,
I. A.
,
2016
, “
Harmonic Wavelets Based Excitation–Response Relationships for Linear Systems: A Critical Perspective
,”
Probab. Eng. Mech.
,
44
, pp.
163
173
.10.1016/j.probengmech.2015.09.021
6.
Kougioumtzoglou
,
I. A.
, and
Spanos
,
P. D.
,
2016
, “
Harmonic Wavelets Based Response Evolutionary Power Spectrum Determination of Linear and Non-Linear Oscillators With Fractional Derivative Elements
,”
Int. J. Non-Linear Mech.
,
80
, pp.
66
75
.10.1016/j.ijnonlinmec.2015.11.010
7.
Pirrotta
,
A.
,
Kougioumtzoglou
,
I. A.
,
Di Matteo
,
A.
,
Fragkoulis
,
V. C.
,
Pantelous
,
A. A.
, and
Adam
,
C.
,
2021
, “
Deterministic and Random Vibration of Linear Systems With Singular Parameter Matrices and Fractional Derivative Terms
,”
J. Eng. Mech.
,
147
(
6
), pp.
1
12
.10.1061/(ASCE)EM.1943-7889.0001937
8.
Oldham
,
K. B.
, and
Spanier
,
J.
,
1974
,
The Fractional Calculus Theory and Applications of Differentiation and Integration to Arbitrary Order
,
Academic Press
, New York.
9.
Di Paola
,
M.
,
Failla
,
G.
,
Pirrotta
,
A.
,
Sofi
,
A.
, and
Zingales
,
M.
,
2013
, “
The Mechanically Based Non-Local Elasticity: An Overview of Main Results and Future Challenges
,”
Philos. Trans. R. Soc. A
,
371
(
1993
), epub.10.1098/rsta.2012.0433
10.
Tarasov
,
V. E.
,
2017
, “
Fractional Mechanics of Elastic Solids: Continuum Aspects
,”
J. Eng. Mech.
,
143
(
5)
, pp.
1
8
.10.1061/(ASCE)EM.1943-7889.0001074
11.
Di Paola
,
M.
,
Pirrotta
,
A.
, and
Valenza
,
A.
,
2011
, “
Visco-Elastic Behavior Through Fractional Calculus: An Easier Method for Best Fitting Experimental Results
,”
Mech. Mater.
,
43
(
12
), pp.
799
806
.10.1016/j.mechmat.2011.08.016
12.
Huang
,
Z. L.
, and
Jin
,
X.
,
L.
,
2009
, “
Response and Stability of a SDOF Strongly Nonlinear Stochastic System With Light Damping Modeled by a Fractional Derivative
,”
J. Sound Vib.
,
319
(
3–5
), pp.
1121
1135
.10.1016/j.jsv.2008.06.026
13.
Spanos
,
P. D.
, and
Evangelatos
,
G. I.
,
2010
, “
Response of a Non-Linear System With Restoring Forces Governed by Fractional Derivatives–Time Domain Simulation and Statistical Linearization Solution
,”
Soil Dyn. Earthquake Eng.
,
30
(
9
), pp.
811
821
.10.1016/j.soildyn.2010.01.013
14.
Spanos
,
P. D.
,
Kougioumtzoglou
,
I. A.
,
dos Santos
,
K. R. M.
, and
Beck
,
A. T.
,
2018
, “
Stochastic Averaging of Nonlinear Oscillators: Hilbert Transform Perspective
,”
J. Eng. Mech.
,
144
(
2
), p.
04017173
.10.1061/(ASCE)EM.1943-7889.0001410
15.
Fragkoulis
,
V. C.
,
Kougioumtzoglou
,
I. A.
,
Pantelous
,
A. A.
, and
Beer
,
M.
,
2019
, “
Non-Stationary Response Statistics of Nonlinear Oscillators With Fractional Derivative Elements Under Evolutionary Stochastic Excitation
,”
Nonlinear Dyn.
,
97
(
4
), pp.
2291
2303
.10.1007/s11071-019-05124-0
16.
Spanos
,
P. D.
, and
Zhang
,
W.
,
2022
, “
Nonstationary Stochastic Response Determination of Nonlinear Oscillators Endowed With Fractional Derivatives
,”
Int. J. Non-Linear Mech.
,
146
, p.
104170
.10.1016/j.ijnonlinmec.2022.104170
17.
Di Paola
,
M.
,
Failla
,
G.
, and
Pirrotta
,
A.
,
2012
, “
Stationary and Non-Stationary Stochastic Response of Linear Fractional Viscoelastic Systems
,”
Probab. Eng. Mech.
,
28
, pp.
85
90
.10.1016/j.probengmech.2011.08.017
18.
Kougioumtzoglou
,
I. A.
, and
Spanos
,
P. D.
,
2012
, “
An Analytical Wiener Path Integral Technique for Non-Stationary Response Determination of Nonlinear Oscillators
,”
Probab. Eng. Mech.
,
28
, pp.
125
131
.10.1016/j.probengmech.2011.08.022
19.
Psaros
,
A. F.
, and
Kougioumtzoglou
,
I. A.
,
2020
, “
Functional Series Expansions and Quadratic Approximations for Enhancing the Accuracy of the Wiener Path Integral Technique
,”
J. Eng. Mech.
,
146
(
7
), p.
04020065
.10.1061/(ASCE)EM.1943-7889.0001793
20.
Petromichelakis
,
I.
, and
Kougioumtzoglou
,
I. A.
,
2020
, “
Addressing the Curse of Dimensionality in Stochastic Dynamics: A Wiener Path Integral Variational Formulation With Free Boundaries
,”
Proc. R. Soc. A: Math., Phys. Eng. Sci.
,
476
(
2243
), epub.10.1098/rspa.2020.0385
21.
Zhang
,
Y.
,
Kougioumtzoglou
,
I. A.
, and
Kong
,
F.
,
2023
, “
A Wiener Path Integral Technique for Determining the Stochastic Response of Nonlinear Oscillators With Fractional Derivative Elements: A Constrained Variational Formulation With Free Boundaries
,”
Probab. Eng. Mech.
,
71
, p.
103410
.10.1016/j.probengmech.2022.103410
22.
Di Matteo
,
A.
,
Kougioumtzoglou
,
I. A.
,
Pirrotta
,
A.
,
Spanos
,
P. D.
, and
Di Paola
,
M.
,
2014
, “
Stochastic Response Determination of Nonlinear Oscillators With Fractional Derivatives Elements Via the Wiener Path Integral
,”
Probab. Eng. Mech.
,
38
, pp.
127
135
.10.1016/j.probengmech.2014.07.001
23.
Mavromatis
,
I. G.
,
Psaros
,
A. F.
, and
Kougioumtzoglou
,
I. A.
,
2023
, “
A Wiener Path Integral Formalism for Treating Nonlinear Systems With Non-Markovian Response Processes
,”
J. Eng. Mech.
,
149
(
1)
, pp.
1
11
.10.1061/JENMDT.EMENG-6873
24.
Gardiner
,
C.
,
1985
,
Handbook of Stochastic Methods for Physics, Chemistry, and the Natural Sciences
, 3rd ed.,
Springer-Verlag
, Berlin.
25.
Grigoriu
,
M.
,
2002
,
Stochastic Calculus
,
Birkhäuser
,
Boston, MA
.
26.
Øksendal
,
B.
,
2003
,
Stochastic Differential Equations
, 5th ed.,
Universitext, Springer
,
Berlin, Heidelberg
.
27.
Chaichian
,
M.
, and
Demichev
,
A.
,
2001
,
Path Integrals in Physics
, Vol.
1
,
IOP Publishing Ltd
., Bristol, UK.
28.
Kougioumtzoglou
,
I. A.
,
2017
, “
A Wiener Path Integral Solution Treatment and Effective Material Properties of a Class of One-Dimensional Stochastic Mechanics Problems
,”
J. Eng. Mech.
,
143
(
6
), pp.
1
12
.10.1061/(ASCE)EM.1943-7889.0001211
29.
Petromichelakis
,
I.
,
Psaros
,
A. F.
, and
Kougioumtzoglou
,
I. A.
,
2018
, “
Stochastic Response Determination and Optimization of a Class of Nonlinear Electromechanical Energy Harvesters: A Wiener Path Integral Approach
,”
Probab. Eng. Mech.
,
53
, pp.
116
125
.10.1016/j.probengmech.2018.06.004
30.
Zhao
,
Y.
,
Psaros
,
A. F.
,
Petromichelakis
,
I.
, and
Kougioumtzoglou
,
I. A.
,
2022
, “
A Quadratic Wiener Path Integral Approximation for Stochastic Response Determination of Multi-Degree-of-Freedom Nonlinear Systems
,”
Probab. Eng. Mech.
,
69
, p.
103319
.10.1016/j.probengmech.2022.103319
31.
Petromichelakis
,
I.
,
Psaros
,
A. F.
, and
Kougioumtzoglou
,
I. A.
,
2020
, “
Stochastic Response Determination of Nonlinear Structural Systems With Singular Diffusion Matrices: A Wiener Path Integral Variational Formulation With Constraints
,”
Probab. Eng. Mech.
,
60
, p.
103044
.10.1016/j.probengmech.2020.103044
32.
Petromichelakis
,
I.
,
Bosse
,
R. M.
,
Kougioumtzoglou
,
I. A.
, and
Beck
,
A. T.
,
2021
, “
Wiener Path Integral Most Probable Path Determination: A Computational Algebraic Geometry Solution Treatment
,”
Mech. Syst. Signal Process.
,
153
, p.
107534
.10.1016/j.ymssp.2020.107534
33.
Psaros
,
A. F.
,
Zhao
,
Y.
, and
Kougioumtzoglou
,
I. A.
,
2020
, “
An Exact Closed-Form Solution for Linear Multi-Degree-of-Freedom Systems Under Gaussian White Noise Via the Wiener Path Integral Technique
,”
Probab. Eng. Mech.
,
60
, p.
103040
.10.1016/j.probengmech.2020.103040
34.
Podlubny
,
I.
,
1999
,
Fractional Differential Equations
,
Academic Press
, San Diego, CA.
35.
Gelfand
,
I. M.
, and
Fomin
,
S. V.
,
1963
,
The Calculus of Variations
,
Prentice Hall
, Upper Saddle River, NJ.
36.
Agrawal
,
O. P.
,
2002
, “
Formulation of Euler-Lagrange Equations for Fractional Variational Problems
,”
J. Math. Anal. Appl.
,
272
, pp.
368
379
.10.1016/S0022-247X(02)00180-4
37.
Almeida
,
R.
, and
Torres
,
D. F.
,
2009
, “
Calculus of Variations With Fractional Derivatives and Fractional Integrals
,”
Appl. Math. Lett.
,
22
(
12
), pp.
1816
1820
.10.1016/j.aml.2009.07.002
38.
Malinowska
,
A. B.
, and
Torres
,
D. F.
,
2010
, “
Generalized Natural Boundary Conditions for Fractional Variational Problems in Terms of the Caputo Derivative
,”
Comput. Math. Appl.
,
59
(
9
), pp.
3110
3116
.10.1016/j.camwa.2010.02.032
39.
Samko
,
S. G.
,
Kilbas
,
A. A.
, and
Marichev
,
O. I.
,
1993
,
Fractional Integrals and Derivatives: Theory and Applications
,
Gordon and Breach
, Philadelphia, PA.
40.
Baleanu
,
D.
, and
Trujillo
,
J. J.
,
2008
, “
On Exact Solutions of a Class of Fractional Euler–Lagrange Equations
,”
Nonlinear Dyn.
,
52
(
4
), pp.
331
335
.10.1007/s11071-007-9281-7
41.
Di Matteo
,
A.
, and
Pirrotta
,
A.
,
2015
, “
Generalized Differential Transform Method for Nonlinear Boundary Value Problem of Fractional Order
,”
Commun. Nonlinear Sci. Numer. Simul.
,
29
(
1–3
), pp.
88
101
.10.1016/j.cnsns.2015.04.017
42.
Atanacković
,
T. M.
,
Konjik
,
S.
, and
Pilipović
,
S.
,
2008
, “
Variational Problems With Fractional Derivatives: Euler–Lagrange Equations
,”
J. Phys. A: Math. Theor.
,
41
(
9
), p.
095201
.10.1088/1751-8113/41/9/095201
43.
Kierzenka
,
J.
, and
Shampine
,
L. F.
,
2008
, “
A BVP Solver That Controls Residual and Error
,”
J. Numer. Anal., Ind. Appl. Math.
,
3
(
1–2
), pp.
27
41
.10.1145/502800.502801
44.
Smith
,
M. C.
,
2020
, “
The Inerter: A Retrospective
,”
Annu. Rev. Control, Rob., Auton. Syst.
,
3
, pp.
361
391
.10.1146/annurev-control-053018-023917
45.
Wagg
,
D. J.
,
2021
, “
A Review of the Mechanical Inerter: Historical Context, Physical Realisations and Nonlinear Applications
,”
Nonlinear Dyn.
,
104
(
1
), pp.
13
34
.10.1007/s11071-021-06303-8
46.
Marian
,
L.
, and
Giaralis
,
A.
,
2014
, “
Optimal Design of a Novel Tuned Mass-Damper–Inerter (TMDI) Passive Vibration Control Configuration for Stochastically Support-Excited Structural Systems
,”
Probab. Eng. Mech.
,
38
, pp.
156
164
.10.1016/j.probengmech.2014.03.007
47.
De Angelis
,
M.
,
Giaralis
,
A.
,
Petrini
,
F.
, and
Pietrosanti
,
D.
,
2019
, “
Optimal Tuning and Assessment of Inertial Dampers With Grounded Inerter for Vibration Control of Seismically Excited Base-Isolated Systems
,”
Eng. Struct.
,
196
, p.
109250
.10.1016/j.engstruct.2019.05.091
48.
Pietrosanti
,
D.
,
De Angelis
,
M.
, and
Giaralis
,
A.
,
2021
, “
Experimental Seismic Performance Assessment and Numerical Modelling of Nonlinear Inerter Vibration Absorber (IVA)-Equipped Base Isolated Structures Tested on Shaking Table
,”
Earthquake Eng. Struct. Dyn.
,
50
(
10
), pp.
2732
2753
.10.1002/eqe.3469
49.
Moghimi
,
G.
, and
Makris
,
N.
,
2021
, “
Seismic Response of Yielding Structures Equipped With Inerters
,”
Soil Dyn. Earthquake Eng.
,
141
, p.
106474
.10.1016/j.soildyn.2020.106474
50.
Petromichelakis
,
I.
,
Psaros
,
A. F.
, and
Kougioumtzoglou
,
I. A.
,
2021
, “
Stochastic Response Analysis and Reliability-Based Design Optimization of Nonlinear Electromechanical Energy Harvesters With Fractional Derivative Elements
,”
ASCE-ASME J. Risk Uncertainty Eng. Syst., Part B: Mech. Eng.
,
7
(
1
), pp.
1
13
.10.1115/1.4049232
51.
Rüdinger
,
F.
,
2006
, “
Tuned Mass Damper With Fractional Derivative Damping
,”
Eng. Struct.
,
28
(
13
), pp.
1774
1779
.10.1016/j.engstruct.2006.01.006
52.
Koh
,
C. G.
, and
Kelly
,
J. M.
,
1990
, “
Application of Fractional Derivatives to Seismic Analysis of Base-Isolated Models
,”
Earthquake Eng. Struct. Dyn.
,
19
(
2
), pp.
229
241
.10.1002/eqe.4290190207