Differential Equations With Fuzzy Parameters

Let \[ \dot x(t)= F(t,x(t), b),\quad x(0)= b,\tag{1} \] be an \(n\times n\)-system of ordinary differential equations with \(a\in\mathbb{R}^n\), \(b\in\mathbb{R}^m\) and \(F\) be a smooth and globally Lipschitz map with respect to the second variable uniformly for \(t\) in compact sets. Then global unique solutions to the system always exist. The authors can interpret them as the operator \[ L: \mathbb{R}^n\times \mathbb{R}^m\to C^1(\mathbb{R}^n) \] mapping each parameter \((a,b)\) to the solution \(x(t)\) to (1). In a lot of applications the authors deal with models where the provided information is uncertain and subjective, for example in information theory, data analysis, artificial intelligence, decision theory, control, etc. In those cases the fuzzy set theory is being applied as a tool to evaluate and measure the influence of imprecise knowledge of the parameters. In the situation given by (1), the parameters \(a\), \(b\) are considered as fuzzy parameters. Applying the Zadeh extension principle [Inform. and Control 8, 338-353 (1965; Zbl 0139.24606)] to the equations is possible to introduce the notions of fuzzy solutions and componentwise fuzzy solutions. Given a set \(X\), a fuzzy set \(A\) on \(X\) is defined as a map \(m_A: X\to [0,1]\). The set of fuzzy sets in \(X\) will be denoted by \(F(X)\). Given \(f: X\to Y\), the Zadeh extension principle allows \(f\) to be extended to a map \(f: F(X)\to F(Y)\) following the rule that to every fuzzy set \(A\) in \(X\), the fuzzy set \(f(A)\) in \(Y\) is assigned with the membership function \[ m_{f(A)}(y)= \sup\{m_A(x): x\in f^{-1}(\{y\})\} \] for \(y\in Y\) with the restriction \(m_{f(A)}= 0\) if \(f^{-1}(\{y\})= \varnothing\). Associated to (1) one introduces the operators: \(E:C^1(\mathbb{R}^n)\to C(\mathbb{R}^n)\) given by \(E(x(t))= [t\to \dot x(t)- F(t,x(t))]\), \(R: C^1(\mathbb{R}^n)\to \mathbb{R}^n\) by \(R(x(t))= x(0)\) and \(L: \mathbb{R}^n\to C^1(\mathbb{R}^n)\) by \(L(a)= x(t)\). Using the extension principle it follows \(E: F(C^1(\mathbb{R}^n))\to F(C(\mathbb{R}^n))\); \(R: F(C^1(\mathbb{R}^n))\to F(\mathbb{R}^n)\) and \(L: F(\mathbb{R}^n)\to F(C^1(\mathbb{R}^n))\). \(X\in F(C^1(\mathbb{R}^n))\) is called a fuzzy solution with initial value \(A\in F(\mathbb{R}^n)\) if \(E(X)= 0\) in \(F(C(\mathbb{R}^n))\), \(R(X)= A\) in \(F(\mathbb{R}^n)\). Given \(A\in F(\mathbb{R}^n)\), \(X= L(A)\) is a fuzzy solution to problem (1). In the paper, a numerical algorithm is presented to obtain the fuzzy point value \(X(t)\) which allows to process subjective information of the fluctuations of parameters in models involving ordinary differential equations. Finally, the authors present, including pictures, two interesting engineering applications: a queueing model for earthwork and a model of oscillations for bell-towers.

Recommendations

Cited in

(70)

This page was built for publication: Differential Equations With Fuzzy Parameters

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q4700084)