This paper presents a method for diagnosis of railway arrester considering uncertainty. Arresters, a protective device that prevents damage due to transient voltages, deteriorate due to the absorption of moisture, repetitive operation during over-voltages and manufacturing defects. Various diagnostic techniques are available for monitoring deterioration of arresters. The technique based on the amplitude of leakage current measures the root mean square or peak values of leakage current components. After measuring the total leakage current, harmonics of leakage current components are analyzed by using a microprocessor based device. The level of leakage current is indicative of the arrester conditions. Harmonics of leakage current components occur due to nonlinear characteristics of railway arrester. Since leakage current contains uncertainty characteristics of power source, the probability density functions of leakage current components can be obtained for ZnO arrester. This paper presents a probabilistic approximation method for the harmonic currents analysis in diagnosis for railway arresters. Mean and variance of harmonic currents in railway system are obtained based on leakage current components. These statistical measures can be helpful to reduce the diagnostic error for railway arrester.