Solving fuzzy linear systems in Gaussian PDMF space

Abstract: In the first part of the paper, we solve the semi-fuzzy linear system (SFLS) $A \mathbf{\tilde{x}}= \mathbf{\tilde{b}}$, where $A\in \mathbb{R}^{m\times n}$ is a real-valued matrix, $\mathbf{\tilde{b}}$ is a fuzzy number vector, and $\mathbf{\tilde{x}}$ is the unknown fuzzy number vector. The elements of both $\mathbf{\tilde{b}}$ and $\mathbf{\tilde{x}}$ belong to a fuzzy number space $\mathcal{X}$, namely the Gaussian probability density membership function (Gaussian-PDMF) space. We present the Cramer's rule to calculate the solution with square matrix $A$ and find out that its solution set is a $5(n-R(A))$ dimensional affine space with $A\in \mathbb{R}^{m\times n}$ and $R(A)$ being the rank of $A$. The explicit form of the solution for RREF matrix $A$ is stated to ensure usability for modeling. In the second part of the paper, we solve the fully-fuzzy linear system (FFLS) $\mathbf{\tilde{A}}\mathbf{\tilde{x}}=\mathbf{\tilde{b}}$, where $\mathbf{\tilde{A}}$ is a fuzzy matrix with all components in $\mathcal{X}$. We analyze its solution set and present the parametric form of solutions under the fuzzy RREF matrix. We then adapt Gaussian elimination method to fuzzy matrices and systems by restricting it to the unit group of ring $\mathcal{X}$, proving the equivalence of solution sets after elementary row operations. We also establish the connection between FFLS and SFLS by confining elements of $\mathbf{\tilde{A}}$ to a subset of $\mathcal{X}$ that forms a field. In the third part, two numerical examples are given to illustrated our method. All results in this paper are explicit since the Gaussian-PDMF space $\mathcal{X}$, to which the membership function of the fuzzy number belongs, possesses a complete algebraic structure. The proposed framework offers a feasible and systematical tool for solving the mathematical models using fuzzy linear systems with uncertainty and fuzziness.