The inconsistency of fuzzy relational equations is frequently encountered when modelling an inference system with fuzzy logic. A polynomial-time direct search procedure is designed to resolve the inconsistency for fuzzy relational equations on {\L}ukasiewicz algebra by minimizing the maximum absolute deviation to the right-hand side vector. Alternatively, a reformulation approach is proposed by minimizing instead the sum of absolute deviations. It results in a polynomial-sized mixed integer linear programming problem and hence, may be readily solved with the aid of an off-the-shelf optimization solver. This reformulation approach is flexible in nature and can be adjusted to deal with some other problems related to the inconsistency of fuzzy relational equations.
fuzzy relational equations, mixed integer linear programming, inconsistency, optimal approximation
08A72, 15B15, 90C11