A typical process control plant consists of hundreds or thousands of control loops. Only one third of these controllers provide acceptable performance. Hence Performance monitoring and control (Plant-wide oscillation assessment) is gaining huge focus in today's world to achieve substantial economical benefits. This work presents a methodology for detection and diagnosis of plant-wide oscillations. The proposed methodology is based on routine operating data and detects the problematic control loops by improved spectral factorization based on Genetic Algorithm (GA) search technique. The advantage of proposed GA based factorization lies in its ability to search the solution space globally. The proposed technique (GA-based factorization), along with existing techniques (Independent Component Analysis and Non-negative Matrix Factorization), are tested on two industrial case studies. Performance comparison of GA based factorization with that of ICA and NMF verifies the improved and efficient results of the proposed method for plant-wide oscillation detection. Higher order statistics (HOS) is used to identify the reasons behind the poor performance of the loops detected as root causes.