Microsoft Word - 1murphy.docx CHEMICAL ENGINEERING TRANSACTIONS VOL. 58, 2017 A publication of The Italian Association of Chemical Engineering Online at www.aidic.it/cet Guest Editors: Remigio Berruto, Pietro Catania, Mariangela Vallone Copyright © 2017, AIDIC Servizi S.r.l. ISBN 978-88-95608-52-5; ISSN 2283-9216 Nondestructive Dropped Fruit Impact Test for Assessing Tomato Firmness Kubilay K. Vursavusa*, Zehan Kesilmisb, Y. Benal Oztekinc a Department of Agricultural Machinery and Technologies, Faculty of Agriculture, University of Çukurova, 01330, Adana, Turkey. b Department of Electrical and Electronics Engineering, Faculty of Engineering, University of Osmaniye Korkut Ata, Osmaniye, Turkey. c Department of Agricultural Machinery and Technologies, Faculty of Agriculture, University of Ondokuz Mayıs, Samsun, Turkey. kuvursa@cu.edu.tr A nondestructive method for assessing the firmness of tomato fruit was developed based on the mechanical properties of the fruit under the dropped fruit impact test. The tests were carried out on Bandita F1 greenhouse tomato variety at six maturity stages for getting a wide range of firmness stage in 2016 season. In the nondestructive dropped fruit impact measurements, impact force and contact time were sensed by a force sensor attached under the impact plate. Other impact parameters were derived from the impact force-contact time curves. Force-deformation ratio at rupture point was used in the measurements of destructive reference parameter and, it was expressed to be tomato firmness (FT). These nondestructive impact parameters were compared with destructive reference parameter for estimating FT. Ten nondestructive impact parameters were used and, the number of impact parameters being processed were reduced with correlation matrix and stepwise regression analyses. After these processes, simple linear regression (SLR) and multiple linear regression (MLR) were used for model development. Root mean squared error (RMSE), mean absolute error (MAE), mean absolute percentage error (MAPE) and coefficient of determination (R 2) were also used for performance evaluation of modelling approaches used to estimate the tomato firmness. The firmness levels of tomato samples were classified with cluster analysis and, classification performance of developed modelling approaches were tested for classification of tomato samples into three firmness levels. Average firmness values of 135 tomato samples were primarily separated to two groups. 70% and 30% of destructive reference and nondestructive impact parameters were used for calibration and validation data set, respectively. According to results of SLR and MLR statistical analysis, MLR model was found to be the most accurate model for firmness estimation with a RMSE of 0.19 N, MAPE of 5.35%, MAE of 0.10 N and R2 of 0.85 after validation. Therefore, it can be applied for firmness estimation of Bandita F1 greenhouse tomatoes with highest accuracy and success rate of 82.93% compared to SLR model in this study. 1. Introduction Fruit commercially harvested for fresh market must be in the early stages of horticultural maturity to minimize damage during harvest, storage, packing and shipping, and to insure acceptable quality at the retail market (Delwiche and Sarig., 1991). As other fruits, only physical properties such as color and size are not adequate for sorting and grading process of tomato fruits exhibiting visco-elastic, non-homogenous and anisotropic behavior. Flesh firmness is a critical handling parameter for tomato, peach, nectarine, pear, apple, avocado, and kiwi fruits (Delwiche et al., 1996). The loss of fruit firmness is a physiological process that occurs during fruit maturation/ripening on the tree, during cold storage and retail handling (Valero et al., 2007). The firmness of a fruit is an index of the mechanical, chemical and rheological properties of the fruit. It is negatively proportional to the maturity of the fruit, and can therefore be used as an alternative indicator to maturity in fruit grading and sorting (Lien et al., 2009). DOI: 10.3303/CET1758055 Please cite this article as: Vursavus K., Kesilmis Z., Oztekin B., 2017, Nondestructive dropped fruit impact test for assessing tomato firmness, Chemical Engineering Transactions, 58, 325-330 DOI: 10.3303/CET1758055 325 Magness-Taylor test, which is called as destructive measurement is a classical method and commonly used for measuring the fruit flesh firmness. This test is conducted by handheld penetrometer or a PC controlled material test device that records the force required to puncture the flesh with a cylindrical probe of fixed diameter and tip geometry. Destructive reference test measures the mechanical attitudes of fruits under the static loading. Several devices related to the classical penetrometer have been developed. While many of these proposed techniques result in reasonably accurate and reproducible estimates, they have a destructive nature, represent mechanical properties at the point of measurements only, and cannot be used as real time measurements for fruit sorting (Lien and Ting, 2014). At present, researchers have tried various nondestructive techniques such as acoustic, vibration, sonic transmission, ultrasonic transmission, near-infrared, micro-deformation and impact (low-mass impact and dropped fruit impact) for sensing fruit firmness. Researchers have shown interest in using impact techniques for predicting firmness of fruits. Previous studies have shown that dropped fruit impacting of fruits on a load- cell or force sensor can be used to evaluate the firmness of fruits successfully (Delwiche et al.1987; Delwiche et al. 1989; Gutierrez et al. 2007; Lien et al., 2009; Ragni et al. 2010; Lien and Ting, 2014; Mireei et al. 2015). Most of the research in using impact test for estimation of fruit firmness uses the impact parameters proposed by Delwiche et al. (1987). Prediction based on those parameters may not result in a significantly accurate classification of firmness. The accuracy of firmness estimation can be improved by combining a dropped fruit impact test and most dominant impact parameters extracted with adequate statistical analysis. This study investigated the feasibility of dropped fruit impact in evaluating tomato firmness via optimization of the impact parameters derived through adequate statistical analysis. Therefore, some researchers such as Gutierrez et al. (2007) for peach, Lien et al. (2009) for tomato, Ragni et al. (2010) for kiwifruit, Lien and Ting (2013) for guava fruit and Mireei et al. (2015) for date fruit have used adequate statistical analyses in order to improve the accuracy of maturity prediction of falling impact test on a force sensor. This study investigated the feasibility of dropped fruit impact in evaluating tomato firmness via optimization of the impact parameters derived through adequate statistical analysis. 2. Materials and methods Fresh greenhouse tomatoes (Bandita F1) that were sorted by color and size, free from disease and injury, and uniform in shape were harvested by hand from a commercial greenhouse in 2016 season. Tomatoes were classified at six different maturity stages (mature green, breaking, turning, pink, light red and red) according to the a*/b* ratio recommended by Batu (2004).Tomatoes at six maturity stages were used for getting a wide range of firmness stage depending on the maturity properties for destructive and nondestructive measurements. Color measurements were performed using Minolta CR-400 colorimeter; four replicates in the equatorial region were taken on each intact tomato. The L*, a* and b* values were obtained directly, and were used to calculate the a*/b* ratio. For mature green, breaking turning, pink, light red and red maturity stages, - 0.590.95 ratios were used, respectively. The reference destructive tests were conducted to define the firmness stage of tomato samples. Lloyd Testing Machine (Model LRX Plus Series) was used for the mechanical test to determine the firmness group of the test samples and, to compare with the nondestructive impact parameters. Puncture test was performed by using a flat ended probe with 4 mm diameter, at a deformation rate of 10 mm min -1 at two equatorial region of each tomato fruit. The load-cell admits a maximum force of 5000 N (resolution 0.005 N) and an error range of 0.03%. Destructive firmness measurements were taken after nondestructive measurements on exactly the same points as the other measurements. For destructive measurements, on each labeled place, puncture probe penetrated at least 11 mm into the flesh. Force-deformation ratio at maximum point was selected from the force-deformation curve and expressed to be tomato firmness in N mm-1 (FT). A nondestructive experimental setup, which is similar to test device developed by Lien et.al. (2009) have been manufactured and used in the experiment. The experimental setup consisted of a force sensor, charge amplifier, and a PC equipped with NI USB6009 DAQ module. Simplified illustration of the experimental setup and photograph of the firmness measurement test device was given in Figure 1. A fruit was held by a manually manipulated vacuum sucker cup and released to fall freely from an adjustable height onto the force sensor. It is well known phenomenal that falling of the fruit on the force sensor can produce deformation on fruit flesh. Therefore, the free falling height was adjusted to 20mm that does not cause deformation to the fruit flesh as recommended by Lien et al. (2009). Force sensor is piezo electric transducer that generates an analogue signal proportional to the applied drop force. In this work, Dytran model 1051V3 was used as force sensor, which has a 45 kg range and 110mV/kg accuracy. The force sensor response was sampled by using NI USB6009 DAQ at 20 kHz and 16 bit precision. 326 Figure 1: Simplified scheme and photograph of the firmness measurement test device Figure 2 shows a typical force response of tomato. Important parameters were marked in Figure 2. The recorded impact response data was analyzed to extract some mechanical indices such as first peak force (fp1), second peak force (fp2), first impact duration (tc1), second impact duration (tc2), latency of first (tP1) and second impact peak force (tP2), time between two peak force (tP1-2) and flying time between two impacts (tC1-2) by using MATLAB. Furthermore, impulse of the first (I1) and second (I2) impact were calculated from the area under the first and second impact force-time curve. Totally, ten nondestructive impact parameters were used to predict FT by single linear regression (SLR) and multiple linear regression (MLR) analyzes. Figure 2: Definitions of impact response The impact parameters provide neither qualitative nor quantitative information about firmness of tomatoes. Therefore, the impact test measurements were correlated with the results of the destructive reference test measurement. The reference destructive test is only used to indicate the firmness of the tomato samples. In order to classify the tomato samples into several firmness groups, cluster analysis (CA) was introduced to classify the tomatoes, according to test measurements, into different firmness stages. CA was used to search natural grouping trends among samples into soft, intermediate and hard firmness groups. SPSS 19.0 for Windows was used for CA data set. For analyzing the experimental data set, SLR and MLR were developed and compared to find the best results for prediction the fruit firmness. Experimental data set consisted of 135 samples. For the calibration data set, 70% of the samples was used and for the test set the remaining part. The calibration data set was used for model development, and the fruits from the validation set were reserved for model testing. After the dominant nondestructive impact parameters were selected according to correlation matrix and stepwise regression analysis, SLR and MLR methods were used to determine the relationship of selected parameters to tomato firmness. SLR and MLR analyses were conducted using SPSS Statistics 20 in order to evaluate tomato firmness models. The data recorded in the test conditions were statistically analysed using one way ANOVA to Force Sensor ���������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������� ���������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������� 20 m m Vacuum generator Plate Mounting Surface NI USB 6009 Charge Amplifier Fruit Vacuum suction cup USB PC High pressure air intake High pressure air out 0 1 0 2 0 3 0 4 0 5 0 6 0 7 0 8 0 9 0 0 5 1 0 1 5 2 0 2 5 3 0 T i m e ( m s ) Fo rc e (N ) 1pf F i r s t i m p a c t S e c o n d i m p a c t 1ct 2ct 1 2pt − f l y i n gt 1pt 2pt 2pf 327 study the effect of tomato maturity stages on tomato firmness and impact parameters. DUNCAN’s multiple range test was used to compare the means. In order to compare the simple linear regression (SLR) and multiple linear regression (MLR) models, the root mean square error (RMSE), the mean absolute error (MAE), the mean absolute percentage error (MAPE) and the coefficient of determination (R2) values were used. According to these criteria, the model that gives a higher value of R2 and lower values of RMSE, MAE and MAPE were determined as the optimal model. The equations for these criteria and the terms in the equations were expressed as below:  = −= n i est i act i YYn RMSE 1 2)( 1 (1)  = −= n i est i act i YYn MAE 1 1 (2) 100. 1 1           − =  = n i act i est i act i Y YY n MAPE (3) In this expressions, actiY is the ith measured value, est iY is the ith estimated value, and n is the total number of measurements. 3. Results and discussion The effect of tomato maturity stages on tomato firmness and ten nondestructive impact parameters were determined using one way ANOVA test. As seen in Table 1, tomato firmness decreased from 3.58 N to 1.16 N significantly during tomato ripening (P<0.01). Among the ten nondestructive impact parameters, I2 and FP2 were found to be the most sensitive parameters related to maturity stages. Maturity stage effect was not significant for tC1-2 and tP1-2. These results also showed that the soft tomatoes (light red and red stages) has a much less firmness than the intermediate (turning or pink stages) and the hard tomatoes (mature green and breaking stages). Hence, this leads to a prolonged total contact time (tc1 and tc2) in the soft tomatoes. Table 1. Destructive and nondestructive measurements of tomatoes in impact tests at different maturity stages Parameters Maturity stages Mature Green Breaking Turning Pink Light Red Red FT (N mm -1 ) 3.58±0.38 e 3.09±0.56 d 2.20±0.47 c 1.74±0.32 b 1.54±0.44 b 1.16±0.21 a I1 (N ms) 114.96±11.54 d 116.36±15.05 d 109.49±13.69c d 102.32±9.85b c 98.64±11.82 b 89.09±9.62 a I2 (N ms) 58.40±6.88 c 55.66±8.77 c 50.48±6.81 b 46.67±4.99 ab 47.16±4.70 ab 43.30±4.40 a fP1 (N) 35.79±2.27 e 34.34±4.49 e 30.73±3.41 d 27.79±2.27 c 24.30±4.04 b 20.55±2.48 a fP2 (N) 16.82±1.95 f 15.08±2.62 e 12.41±1.74 d 10.90±1.23 c 9.63±1.53 b 8.19±1.05 a tP1 (ms) 3.46±0.30 a 3.61±0.32 ab 3.91±0.54 b 3.90±0.46 b 4.29±0.49 c 4.65±0.45 d tP2 (ms) 4.00±0.57 ab 3.84±0.38 a 4.28±0.48 bc 4.46±0.57 c 5.07±0.56 d 5.61±0.73 e tC1 (ms) 8.17±0.49 a 8.51±0.60 a 9.17±0.86 b 9.35±0.85 b 10.43±1.25 c 11.65±1.22 d tC2 (ms) 8.79±0.84 a 8.79±0.76 a 9.87±1.01 b 10.42±1.25 b 12.02±1.55 c 13.38±1.81 d tC1-2 (ms) 84.14±6.18 a 82.21±2.39 a 80.64±3.22 a 73.17±8.30 c 76.80±6.06 b 77.06±4.67 b tP1-2 (ms) 92.80±6.62 a 90.89±2.59 ab 90.13±3.44 ab 83.03±8.16 c 87.96±6.16 b 89.58±5.18 ab The 135 tomato samples at different maturity stages used for firmness classification of tomatoes were first subjected to cluster analysis and, it was decided what the firmness groups should be. According to cluster analysis, the SPSS procedure classified the samples into 3 distinguishing clusters such as soft, intermediate and hard firmness groups. Consequently, 32 tomato samples with a mean of 1.39±0.28 N mm-1 in the soft group, 39 tomato samples with a mean of 2.23±0.30 N mm-1 in the intermediate group and 23 samples with a mean of 3.43±0.40 N mm-1 in the hard group were determined . Firmness ranges formed by cluster analysis was FT≤1.79 N mm-1, 1.80