Determination of Avocado Fruit Ripening Stage Using an Electronic Nose with Fuzzy Logic Algorithm

Determination of Persea Americana Fruit Ripening Stage using Electronic Nose with Fuzzy Logic Algorithm

Julie Ann A. Parañal¹

School of Electrical, Electronics, and Computer Engineering, Mapúa University

Abstract— Ethylene is a gaseous plant hormone that naturally occurs in fruits and helps speed up the ripening process. Persea Americana, or Avocado, is a climacteric fruit that does not produce large amounts of ethylene while still attached to the tree. Therefore, it does not ripen until harvested, and thus, its ripeness cannot be determined by the naked eye. Since fruit quality is judged by consumers primarily from their perception of the acceptability of fruits based on characteristics including visual appeal (lack of blemishes, color, size, and texture), relying on such methods is not applicable for Avocado in general. The use of an electronic nose - an intelligent sensing device that can sense aroma more effectively than the human sense of smell - would prove to be useful. It also possesses a non-destructive property, therefore resulting into it being selected to be the ideal digital, electronic device for identifying, characterizing, and grading fruits ripeness. This paper examines and measures the concentration of the metal oxide gas sensors in determining the ripeness using the ethylene gas in fruits. The preliminary performance of the electronic nose has been demonstrated at the ripening stage of the avocado fruit and is compared to the separation machine. Confusion matrix was used to show the accuracy of the system in detecting ripening stage. This study used the fuzzy logic algorithm to classify and achieve an accuracy rate of 82.5% of classifying unripe, ripe, and overripe of the fruits.

Index Terms— ethylene; ripeness; gas sensor; persea americana; avocado; fuzzy logic algorithm; electronic nose

I. Introduction

Ethylene is a small, naturally occurring hydrocarbon gas, and it is also a by-product of combustion and other processes. Research has since demonstrated that ethylene plays an important role in many plant development processes including seed germination, vegetative growth, leaf abscission, flowering, senescence, and fruit ripening. The determination of climacteric fruit ripening states once harvested, is becoming a very important issue for producers, sellers, food industries, and as well as consumers, due to the delicate nature of the fruit. Fruit quality is judged by consumers primarily from their perception of the acceptability of fruits based on characteristics including visual appeal (lack of blemishes, color, size, and texture) [1]. Some fruit varieties vary widely in aroma characteristics because aroma is often the most valued trait in determining fruit quality and consumer choice. A perfect example of which is the Persea Americana or Avocado, a climacteric fruit that does not produce large amounts of ethylene when still connected to the tree, ripening only when it is harvested. The advancement of technology, however, has offered electronic nose or e-nose as a new alternative way to evaluate and grade fruits based from an aroma called the ethylene gas. The electronic nose is an intelligent sensing device that can sense aroma more effectively than the human sense of smell, and due to its non-destructive property, it has been selected to be the ideal digital, electronic device for identifying, characterizing and grading fruits ripeness.

Several studies have focused on using electronic nose for different applications. A study classified various odor groups using electronic nose. Existing studies in determining the ripeness of the fruits include a new ultrasonic echo pulse method in order to study the feasibility of maturity assessment of orange fruits. This research is conducted in order to measure the ultrasonic parameters eventually velocity and attenuation in order to check the aptitude of this technique to determine the maturity degree of the fruit. Other researchers that determine the maturity of a fruit used near infrared spectroscopy (NIR), which is also a non-destructive method for determining the maturity degree of mango fruits during postharvest ripening. Detection methods used in other countries make use of the degree of ripening [2]. It then relies on an automatic determining system based on cameras with computer-based technology. The method has been widely explored for quality analysis and the grading of agricultural products in recent years, more commonly known as the computer vision system or computerized image analysis technique [3]. It was proven to be successful for objective measurement of various fruit crops. None of the previous researches, however, incorporated the idea of determining the ripeness of a climacteric fruit with the use of fuzzy logic algorithm, enabling the generation of precise solutions from approximate information, leading up to the ability to determine the persea americana’s maturity stages [4]. Varying on the specific chemical, the electronic nose will focus on have incorporated the idea of determining the ripeness of a climacteric fruits with the use of the electronic nose and fuzzy logic algorithm.

The main objective of this study is to measure ethylene and other metal oxide gases present in determining the ripeness stage of persea americana samples through electronic nose using fuzzy based classification algorithms and to verify the accuracy of results by comparing it to data from human expert graders of fruits. This study categorizes the metal oxide gas present in fruit aroma of the sample and utilizes the use of the fuzzy logic algorithm for the software. The response level of the metal oxide concentration in the fruits sample are classified as unripe, ripe and overripe, which is displayed in GUI as the output.

The study poses several advantages, including its non-destructive method, reliability, and repeatable operation of ripeness stage determination. The system also homogenizes the “ripeness determination” through analysis of ethylene emitted by a ripening fruit [5].

The electronic nose is composed of the following: a sensing mechanism that detects the type of gas entering in the system, gas sensors that detects metal oxide gas, fuzzy logic algorithm for data processing, and Raspberry Pi for the computation. The e-nose system is a tool that can sense the concentration of gas fruit sample. The system utilizes sensitive sensors that need to be in an enclosed container to avoid unnecessary gas in air to be examined for the accuracy of the sensor readings. The center of attention of this study is the detection of ethylene and other metal oxide gas concentration found in fruit samples [6].

II. Methodology

A. Conceptual Framework

The system focuses on the aroma of persea Americana characterized based on the ripening stage. With the aim to develop an electronic nose that determines whether persea americana is unripe, ripe, or overripe through the aroma of the fruit samples.

Abbildung in dieser Leseprobe nicht enthalten

Fig 1. Conceptual Framework

Based on Fig. 1, the input of the system is ethylene gas from the persea americana. Gas sensors were used to capture the concentration of metal oxide gas present in the fruit aroma. The system starts by placing the fruit sample to the closed vessel. When the array of gas sensors detects the aroma of the fruit, the acquired data is analysed and characterized using the fuzzy logic algorithm. The fuzzy inference system (FIS) classifies the gas present in the fruit based from the fuzzy rules. Then, the output of the system is displayed in the graphical user interface, classifying the fruit as either unripe, ripe, or overripe.

B. System Flowchart

Abbildung in dieser Leseprobe nicht enthalten

Fig 2. System Flowchart

The system flowchart of the study is shown in Fig 2. The gas sensor senses the metal oxide concentration from the fruit aroma. The data acquired by the sensors are sent to Raspberry Pi microcomputer. The system then checks if it detects the presence of gas. The data gathered then undergoes a process using Fuzzy Logic Algorithm [7]. Then, the data is classified through the Fuzzy Inference System. Lastly, the output of the system is then classified into unripe, ripe, or overripe.

C. Block Diagram

Abbildung in dieser Leseprobe nicht enthalten

Fig. 3. Block Diagram of E-nose

Fig. 3 shows the block diagram that represents the e-nose system sensing concept. E-nose is an enclosed vessel device that depends on the gas sensors and microcomputer. The E-nose block diagram comprises of functional blocks of power supply, sensor array that captures metal oxide gas data from the persea americana fruit. Finally, the GUI would then output the determination of the ripening stage based on the gas acquisition from fruit sample analysis [8]. The sensor array of the system is composed of seven commercial gas sensors: MQ-2, MQ-3, MQ-8, MQ-135, MQ-136, TGS-822 and TGS2600.

D. Gas Sensor

Gas sensors used in this study can detect and respond to a range of ethylene gas in persea americana fruit. The e-nose is comprised of seven commercial gas sensors: MQ-2, MQ-3, MQ-8, MQ-135, MQ-136, TGS-822 and TGS2600. A temperature sensor (LM35) is used to monitor the sensor chamber temperature and relative utility. The e-nose sensors specifications determine what gasses can be measured.

The basis of selecting the gas sensor used in this study was based on the selectivity and sensitivity of sensors on specific metal oxide gas produced in metal oxide gas sensor of determining the ripening stage of fruits.

TABLE I Sensors used in the E-nose

Abbildung in dieser Leseprobe nicht enthalten

All seven sensors are sensitive to gases. The TGS-822 gas sensor detects Ethanol or Ethylene gas, while the TGS-2600 is for the gaseous air contaminants. The MQ sensors are sensitive to combustive gas, ammonia, alcohol, hydrogen gas, and hydrogen sulphide [9]. The sensors, when triggered, there have a change in the electrical resistance when exposed to the aroma of the fruit.

Abbildung in dieser Leseprobe nicht enthalten

Fig 4. Array of Sensors

The array of sensors in the e-nose system is illustrated in Fig 4. Gas sensor array is placed in an enclosed chamber

E. Gas Data Analysis

The persea americana fruits contains ethylene gas which is a naturally occuring gas resulting from the ripening of the fruit [10]. In this study, the electronic nose was used to determine the gathered persea americana fruit samples ripening stage. For training set, there were 40 fruit samples used. The experiment procedure was done by placing each persea americana inside the chamber for at least 5 minutes.

TABLE II Composition of fruits for training

Abbildung in dieser Leseprobe nicht enthalten

Table II shows the number of fruit samples in the study. The gathered samples were categorized beforehand, according to their ripening stages. Tests were then performed to determine the baseline values for the gas sensors, and based on the gathered data, samples belonging to the overripe category produced higher level concentrations of VOCs as compared to samples labelled unknown, ripe and unripe.

TABLE III Variation of Ethylene to sample fruit

Abbildung in dieser Leseprobe nicht enthalten

Table III shows the variation in ethylene levels, measured in the calibration process of analyzing the different ripeness category of the samples, whether ripe, unripe, or overripe. Based on the result, fruits belonging to the overripe category were observed to possess higher levels of ethylene. This is due to the level of ethylene being directly linked to the premature aging and rotting of fruits and vegetables, including the wilting of flowers and leafy vegetables. Thus, the concentration of which is used as a basis in determining the ripening stage of fruits.

TABLE IV Variation of Other Volatile Organic Compound in Aroma of The Persea Americana Sample Fruit

Abbildung in dieser Leseprobe nicht enthalten

Table IV shows the variations present in the MQ sensors used in the fruit ripening stage test. In the training process, the unknown fruit ranges from 0 to 0.42, the unripe Persea Americana ranges from 0.43 to 0.50, the ripe 0.51 to 0.70, and the overripe is 0.71 to the highest value attained by the sensor. The MQ sensors are capable of sensing combustive gas, ammonia, alcohol, hydrogen sulphide, and hydrogen gas in the fruit samples. During the data gathering process, there was an observed increase in all levels of gas from samples belonging to the overripe category as compared to those in other categories.

The MQ sensors were combined to develop the distinguishing capability of the system. In this experiment, it was detected that the data acquired by the MQ-136 and TGS2600 did not show a significant change in values compared to TGS822 that clearly changed in values in all ripening stage. Hence, in the experiment, combining the values from the MQ-136 and TGS2600 sensors did not affect the results. The MQ-3 that detected alcohol content in fruit sample could be associated to the concentration of ripening stage of fruit as chemical formula analysis of ethylene, C2H4 that refers to 2 carbon atoms and 4 hydrogen atoms. MQ3 gas sensor is used for detection of alcohol and ethanal (Ethanol; C2H5OH) which its elements are ethylene (C2H4) and water (H2O) [11]. Also, the MQ-2, MQ-8 and MQ-135 detect the gas content of the fruit sample. Based on the research, a high gas content concentration in fruits can also predict its ripening stage [11]. This information can be used to verify the ripening stage of the fruit samples.

F. Membership Function of Fuzzy Logic

In this study, membership function of fuzzy logic was used to classify the aroma sample of the fruit into three groups [11]. The membership function of the fuzzy is a curve that defines how each input is mapped to a membership value (or degree of membership) between 0 and 1. The input values refer to the voltage levels given by the sensors to recognize the data for classification. The relationship between the input of the seven sensors and the ripening stage is determined by the fuzzy rules in the inference system. A fuzzy control rule refers to a fuzzy conditional statement in which the antecedent is a condition in the application domain and the consequent is a control action for the system [12].

The fuzzy divides the level concentration data as shown in Tables II and III, that is the input into three groups. Based on the table of ethylene and combination of MQ sensors range of concentration, the Fuzzy Logic Algorithm was used to determine the samples [13]. The ethylene concentration level and other gas acquisition were classified into their classes, namely, Unripe, Ripe and Overripe stage of the fruit.

In the fuzzy rules the MQ and TGS sensors were combined to enhance the sensing capability of the system [14]. In the experiment, it was observed that data acquired by the seven metal oxide sensors which are: MQ-2, MQ-3, MQ-8, MQ-135, MQ-136, TGS-822 and TGS2600 should be in low if it is unripe, medium for ripe and high for overripe MQ-3 and MQ-8 in three ripeness stage did not show a significant change in values compared to TGS822 that evidently changed in values in three ripeness stage [15]. Thus, in the experiment, combining the values from the MQ sensors does affect the results [16].

Abbildung in dieser Leseprobe nicht enthalten

Fig 5. Gas Voltage Reading

The raw data from Fig 5 are interpreted by the Fuzzy through signal value. The classification of the gases is done using fuzzy logic-based approach [17]. The member functions and rules of the fuzzy inference system are defined. The inputs applied to the eight sensors are combined by AND operations [18]. By applying the rules to the FIS, the outputs are obtained.

G. Experimental Procedure and Setup

1. The sample is placed inside the chamber.
2. The chamber should be closed for 5 minutes.
3. The array of sensor analyses the gas from the fruit samples. The sensors act as a sensing mechanism with their own specifications.
4. After the gas sensors detect the gases inside the device, the ADC module converts the analog signals to digital signals so that the raspberry pi can read the signals from the gas sensors.
5. The membership function of the fuzzy classifies the ethylene and other gas content of the sample fruit, whether it is unripe, ripe or overripe.
6. The Raspberry pi can output the diagnosis through the GUI.

Abbildung in dieser Leseprobe nicht enthalten

Fig 6. Experimental Setup of the System

The set-up for the system is shown in Fig 6. During testing, the sample fruit is placed inside the unit for at least 5 minutes. Once the sensor has analysed the breath samples, the array of sensors will send signals to the raspberry pi [19]. The data obtained can be displayed in the laptop to observe the behaviour of the sensor readings. The result is displayed to the GUI.

To perform another testing on the sample fruit, it must wait for 1 to 5 minutes to circulate an ambient air to reset the sensor readings.

III. Result and Discussion

The e-nose prototype system was developed to measure and sense the range of concentration of ethylene that could be present in fruit sample. An increase in ethylene and other VOC in the aroma of fruit is seen as an indication of ripeness stage. In this study, the classifications of the subjects based on the exhaled VOCs concentration used the Fuzzy Logic algorithm:

Abbildung in dieser Leseprobe nicht enthalten

Fig 7. Range Value for the Readings of Ethylene

In figure 7, these are the range values for the reading of the ethylene gas sensor. For the variation for reading the MQ Sensors, the fuzzy rules in the fuzzy logic was used, as seen on Figures 8 to 11. In table IV, the MQ sensors ranges is determined in the training process of the fruits. The voltage passing thru the gas sensors can be seen in the GUI, each of the gas sensors has a capability to transfer data by reacting to the chemical compounds present in a fruit.

Abbildung in dieser Leseprobe nicht enthalten

Fig 8. Sample Fuzzy Logic Algorithm for UNKNOWN

In figure 8, gas sensor MQ8 must have a reading at “medium” level and other sensors has a value of “superlow” level to say that the target sample is not an avocado.

Abbildung in dieser Leseprobe nicht enthalten

Fig 9. Sample Fuzzy Logic Algorithm for UNRIPE

In figure 9, there are three other rules of fuzzy logic for an unripe result. For documentation purposes, however, only one is presented in this study. In this rule, there are 2 gas sensors that are of super low level, 4 gas sensors with a low-level reading and 1 gas sensor with a high-level reading. The common reading in the four fuzzy logic rules for unripe sample is that, TGS822 has a low-level reading and it is for ethylene compounds which determine the ripeness of a fruit. The more ethylene present in the avocado, the riper the fruit.

Abbildung in dieser Leseprobe nicht enthalten

Fig 10. Sample Fuzzy Logic Algorithm for RIPE

In figure 10, there are 8 other rules for ripe result. In this rule, having a medium level reading for MQ3 and TGS822 will give a result of a ripe avocado.

Abbildung in dieser Leseprobe nicht enthalten

Fig 11. Sample Fuzzy Logic Algorithm for OVERRIPE

In figure 11, having a high-level reading in TGS822, MQ136 and MQ8 are leading into an overripe result because the target compounds of MQ136 and MQ8 are the foul odour of a fruit.

TABLE V Determination of Ripening Stage through Aroma analysis (testing)

Abbildung in dieser Leseprobe nicht enthalten

The results and data are shown in Table 5. In the table, based on the set threshold parameters discussed in the previous section, the data have been shown in labelling and classifying of ripening stage of the fruit samples. A total of 40 fruit samples were gathered by the researchers to be tested by using the prototype. There are 8 fruit samples that the system’s value detected that is not the true value of the fruit based on the farm’s data.

TABLE VI

Abbildung in dieser Leseprobe nicht enthalten

Confusion Matrix

In Table VI it shows the confusion matrix used to evaluate the performance of the e-nose system. The total fruit samples collected were 40 fruits, having 8 unknown fruits, 10 unripe avocado fruits, 12 ripe avocado fruits and 10 overripe avocado fruits. The confusion matrix is used to describe the performance and accuracy of the system on the true vales where the set of test data is known. To determine the accuracy of the prototype, the following formula was used:

Abbildung in dieser Leseprobe nicht enthalten

The accuracy of the system is determined using eq. 1. The accuracy of the electronic nose system is calculated as 82.5% accuracy. It also observes that the system has an error in determining ripening stage in some sample.

IV. Conclusions and Future Works

The researchers observed and were able to conclude that ripening stage of fruits that are overripe have higher amounts of ethylene and various VOCs compared to those unripe and overripe fruit samples . The fuzzy logic algorithm that was used achieves the accuracy of 82.5% of classifying unripe, ripe and overripe of the fruits. The analysis of VOCs in the aroma of fruit refines the conventional way of determining ripening stage among fruit samples through gas analysis using fuzzy logic algorithm. The created device helps agricultural industries with a non-destructive method of sample fruits fall under the three category which are unripe, ripe and overripe. The use of LM35 as temperature sensor for the humidity of the enclosed chamber is a big factor in maintaining the stable values of the gas sensors. Moreover, the study can also be a proof of concepts in pursuing further studies regarding to VOC’s analysis and electronic nose.

As for the long run upgrades, a much faster reaction of the gas sensors is needed in getting the amount of compounds and can produce more accurate results. The waiting time in getting the readings in the sensors might also be lessened if the gas sensors are much reactive in sensing their reactants, no need to increase the number of gas sensors, just buy the most efficient one that might be expensive but with a greater result.

References

[1] Baietto, M., & Wilson, AD., “Electronic-nose applications for fruit identification, ripeness and quality grading”, Center for Bottomland Hardwoods Research, January 2015.

[2] Kanade, A., Shaligram, A.D.D., “Development of an E-Nose Using Metal Oxide Semiconductor Sensors for the Classification of Climacteric Fruits”, International Journal of Scientific & Engineering Research, Volume 5, Issue 2.

[3] Nouri, F.G., Chen, Z., Maqbool, M., “Monitoring Mango Fruit Ripening after Harvest using Electronic Nose (zNoseTM) Technique”, International Conference on Food Engineering and Biotechnology, IPCBEE vol.65

[4] Padilla, D.A., Balbin, J.R., Caluyo, F.S., De Luna, C.Y.B., Orio, J.C.R., “Zigbee and Power Line Communications Interconnectivity Applied to Fuzzy Logic Controlled Automated Lighting system”, 6th IEEE Region 10 Annual International Conference, Proceedings/TENCON, Volume 2018-5 April 2017, 25 November 2016.

[5] Sese, J.T., Balbin, J.R., Babaan, C.V.R., Poblete, D.M.M., Panganiban., R.P., Poblete, J.G., “Detection and Classification of Bacteria in Common Street Foods Using Electronic Nose and Support Vector Machine”, 7th IEEE Region 10 Annual International Conference, Proceedings/TENCON, Volume 2017-7 November 2017, 7 February 2018.

[6] Zhang, H., Wang, J., “Evaluation of Peach Quality Attribute Using an Electronic Nose”, Sensors and Materials, Vol. 21, No. 8, pp419-431.

[7] Dimatira, J.B.U., et al., “Application of Fuzzy Logic in Recognition of Tomato Fruit Maturity in Smart Farming”, 2016 IEEE Region 10 Conference (TENCON), 2016.

[8] Nandi, C.S., Tudu, B., Koley, C., “Machine Vision Based Techniques for Automatic Mango Fruit Sorting and Grading Based on Maturity Level and Size”, Sensing Technology: Current Status and Future Trends II, pp 27-46, 2013.

[9] Szulczynski, B., Gebicki, J., Namiesnik, J., “Application of fuzzy logic to determine the odour intensity of model gas mixtures using electronic nose”, E3S Web of Conferences, Volume 28, 2018.

[10] Mulyani, E.D.S., Susanto, J.P., “Classification of Maturity level of Fuji Apple Fruit with Fuzzy Logic Method”, 2017 5th International Conference on Cyber and IT Service Management (CITSM), 2017.

[11] Udomsak P., et al., “The Analysis of Fruit Ripening Level from Diffusion of Gases by using Data Mining Techniques”, International Journal of Pure and Applied Mathematics, 2018

[12] Geethapriya, M.N., Praveena, S.M.D., “Evaluation of Fruit Ripeness using Electronic Nose”, International Journal of Advanced Information Science and Technology Vol.6, No.5, May 2017.

[13] Sharma, J., et al., “Fuzzy Logic based Odour Classification System in Electronic Nose”, International Journal of Computer Applications, Volume 78 – No.15, September 2013.

[14] Beaulieu, J.C., Grimm, C.C., “Identification of Volatile Compounds in Cantaloupe at Various Developmental Stages Using Solid Phase Microextraction”, J. Agric. Food Chem, pp 1345–1352, February 2001.

[15] Kanade, A., Shaligram, A., “Ripening state determination of guava fruit (Psidium guajava) using e-nose with fuzzy logic as pattern recognition tool”, International Journal of Scientific Research Engineering & Technology (IJSRET), ISSN 2278 – 0882 Volume 7 Issue 4, April 2018.

[16] Lim, J.V.M., Linsangan, N.B., Cruz, F.R.G., Chung, W., “International Journal of Computer and Communication Engineering, Volume 5, Number 5”, December 2015.

[17] Dela Cruz, J.C., et.al., “Ripeness classification of cocoa through acoustic sensing and machine learning”, 2017 IEEE 9th International Conference on Humanoid, Nanotechnology, Information Technology, Communication and Control, Environment and Management (HNICEM), January 2018.

[18] Rajin, S.M.A., Samad, S.A., Muad, A.M. “Post-harvest Quality Evaluation of Grapes using Non-destructive Electronic Nose”, Journal of Electrical and Electronics Engineering, October 2015.

[19] Taghoy, D.M., Villaverde, J.F., “A Fuzzy Logic Approach for the Determination of Cavendish Banana Shelf Life”, IEEE Region 10 Annual International Conference, Proceedings/TENCON, Volume 2018-October, 22 February

Recommendation of Further Study

Possible future works include; adding a small touch-screen monitor having compatibility with the sensors used. This includes a user-friendly interface for easy access and tracking of obtained results. The researchers also suggest increasing the number of gas sensors to lessen the waiting time in getting the readings of the sensors and to produce more accurate results.