This paper presents a diagnosis system in which, an algorithm automatically finds all minimal structurally overdetermined (MSO) sets in a structural model of a system. It combines a first set of MSO sets to get the additional MSO sets, which were obtained after determining a complete matching between equations and unknown variables. Finding the complete set is useful for the diagnosis task increasing the fault isolability due to the fact that it can provide different signatures to each fault. The efficiency of the algorithm for finding all MSO sets is illustrated using the DAMADICS benchmark, which is a pneumatically-actuated control valve.