Acoustic response discrimination of phulae pineapple maturity and defects using factor analysis of mixed data and machine learning algorithms

• Non-invasive technique for maturity and internal defect of Phulae pineapple was proposed. • Acoustics response integrated with machine learning (ML) was proposed. • The efficacy of the six ML classifiers was compared. • Random Forest (RF) provided the highest classification accuracy higher than 99.0 %. Acoustic response is non-destructive evaluation technique that replicates the conventional method for determining maturity by tapping the fruit. The physical (dimensions, color, firmness, and specific gravity) chemical (TSS, %TA, and TSS/TA), and acoustic properties of Phulae pineapple were determined and used to classify the maturity and defect, e.g. translucency flesh symptoms. Results showed that all physical parameters of the two maturity stages were not significantly different ( p > 0.05). Translucency flesh (TF) defects were observed in 23.5 % and 27.3 % of pineappls in the green and green-yellow stages, respectively. The dominant resonance frequency ( f n ) of Phulae pineapple ranged of 0.057 to 3.010 kHz. All the physical, chemical, and acoustic properties were used to classify for maturity and defects using the factor analysis (FA) technique and machine learning (ML). Results showed that maturity was correctly classified at 84.0 % by all parameters, while elected non-destructive parameters (color, specific gravity, and stiffness coefficients) showed lower results for distinguishing pineapples. Random Forest (RF) provided a better classification than other MLs with 99.93 % accuracy of maturity classification, while TF classification was 99.59 %. Results showed acoustic method integrated with ML was a fast reliable, and cost effective technique for assessing Phulae pineapple quality.