A low-mass non-destructive impact device was tested and compared with destructive tests to evaluate the firmness of peach [Prunus persica (L.) Batsch var. persica]. The tests were carried out on four peach varieties, namely, Royal Glory, Caterina, Tirrenia and Suidring. In the measurements, impact acceleration and contact time were sensed by an accelerometer attached on impact head, and the primitive impact parameters such as maximum acceleration (A_max), time required to reach peak acceleration (t_max) and contact time during impact (t_contact) were determined by using impact acceleration-contact time curves. Other parameters were derived by using the theory of elasticity. These non-destructive impact parameters were compared with the destructive reference parameter (Magness-Taylor force). The levels of firmness of the four peach varieties were classified with discriminant analyses based on the primitive impact parameters and their derivatives.

The accuracy of classification was improved with linear discriminant analysis, and the number of parameters being processed was reduced with stepwise regression analysis. The correlations between destructive reference and non-destructive impact parameter test results were statistically significant. Discriminant analysis results showed that the accuracy of classification was 65.40% by the primary impact parameters and 76.50% by all 10 impact parameters.

Results of the four most dominant parameters [t_contact, A_max^2.5, A_max/(t_max)², A_max/(t_contact)²] extracted with stepwise regression analysis showed that overall accuracy of classification was improved to 77.90%. Because the four most dominant parameters gave better classification accuracy than the 10 impact parameters, they could be practical in static application using a low-mass lateral impactor. To improve the accuracy of classification, the four impact parameters approach may be adopted for extracting the soft peaches (81.6%) and the primary impact parameters approach for the hard peaches (83.3%).