Purpose - In outside constructions (e.g. aircraft frames, bridges, tanks and ships) real-life noises reduce significantly the capability of location and characterization of crack events. Among the most important types of noise is the rain, producing a signal similar to crack. This paper seeks to present a robust crack detection system with simultaneous raining conditions and additive white-Gaussian noise at 220 to 20 dB signal-to-noise rati (SNR). Design/methodology/approach - The proposed crack detection system consists of two sequentially, connected modules: the feature extraction module where 15 robust features are derived from the signal and a radial basis function neural network is built up in the pattern classification module to extract the crack events. Findings - The evaluation process is carried out in a database consisting of over 4,000 simulated cracks and drops signals. The analysis showed that the detection accuracy using the most robust 15 features ranges from 77.7 to 93 percent in noise-free environment. This is a promising method for nondestructive testing (NDT) by acoustic emission method of aircraft frame structures in extremely noisy conditions. Practical implications - Continuous monitoring of crack events in the field requires the development of advance noise reduction and signal identification techniques. Robust detection of crack signals in noisy environment, including raining drops, improves significantly the reliability of realtime monitoring systems in large and complex constructions and in adverse weather conditions. Originality/value - Asfar as is known this is the first time that an efficient system is presented and evaluated which deals with the problem of crack detection in adverse environment including both stationary and non-stationary noise components. Moreover, it provides further information on the engineering and efficiency problems associated with NDT techniques in the aircraft industry.