An electronic nose based on acoustic wave sensors has been developed to detect spoilt fruit. Different varieties of fruits, edible and rotten, were analysed. Starting from six sensors, the minimum number of sensors capable of discriminating between spoiled and unspoiled fruit was found. The discrimination capability of the sensor array was studied separately for each fruit variety, as well as for the whole set. Mathematical models were built to classify the fruits within a fruit variety, in an objective and clear way. The models were able to distinguish between edible and rotten fruits with 100% success for New Hall oranges, Golden apples, Kiwis and William pears, and with 97.2% of success for the Starking apples. Without forming fruit variety subsets, discrimination between edible and rotten fruit was achieved with 95% success.