Issue 51

G. Ramaglia et alii, Frattura ed Integrità Strutturale, 51 (2020) 288-312; DOI: 10.3221/IGF-ESIS.51.23 297 exp,i , exp,i th i f f e f   (19) where, exp,i f and ,i th f represent the experimental and theoretical values for the generic specimen, i respectively. In particular they represent the compressive strength of the confined masonry specimen derived by the experimental and theoretical results. They are generally normalized to the unconfined compressive strength of the masonry specimen. However, the approximation parameter provides local information without considering the entire sample. Therefore it cannot be used to assess the performance of mechanical models. Three statistical parameters can be used to assess the entire sample of the experimental results: the mean absolute percentage error (MAPE), mean square error (MSE) and coefficient of determination (R 2 ). These statistical parameters allow to compare the theoretical prediction with the experimental results on the entire statistical sample. The statistical parameters can be calculated as follow: exp,i , 1 exp,i n th i i f f f MAPE n     (20)   2 exp,i , 1 n th i i f f MSE n     (21)   2 exp,i , 2 1 , 1 1 n th i i n th i i f f R f                    (22) where, n is the number of specimens for each confinement test. The prediction models provide reliable results when: 2 0 0 1 MAPE MSE R    T HEORETICAL AND EXPERIMENTAL COMPARISON RESULTS he reliability of mechanical models previously derived by the available failure criteria has been discussed in the following section, by comparing the experimental results with the theoretical previsions. The analysis was carried out in terms of direct comparison between the experimental and theoretical results and in terms of confinement curves. A first step has been carried out by comparing the confined compressive strength, mc f normalized to the unconfined compressive strength, 0 m f , experimentally evaluated with the same ratio calculated using the theoretical models. No specific information is provided in original papers on tensile strength of masonry as a whole. Tensile strength is preferably retrieved by direct or indirect experimental tests [41] or correlations with other mechanical properties [40]. Compressive strength of the mortar only was reported and the tensile strength of masonry is assumed equal to 10% of compressive strength for lime mortar and 20% for cementitious mortars. The values of assumed for each experimental program are shown in Tab. 1. The direct comparison between the experimental results and the theoretical previsions is a useful tool to assess the response of the model on the entire experimental sample. It can be performed on a diagram where in the horizontal axIs is shown the normalized confined compressive strengths, theoretically assessed; while on the vertical axis is reported the T