CIGR Ejournal Style and Format Guidelines ARID ZONE JOURNAL OF ENGINEERING, TECHNOLOGY & ENVIRONMENT AZOJETE - CIGR Section VI Special Issue: Innovation & Technologies for Sustainable Agricultural Production & Food Sufficiency AZOJETE, December, 2018. Vol. 14(SP.i4): 121-128 Published by the Faculty of Engineering, University of Maidiguri, Maidiguri, Nigeria. Print ISSN: 1596-2490, Electronic ISSN: 2545-5818 www.azojete.com.ng _____________________ *Corresponding author Email address: shuso@bpe.agr.hokudai.ac.jp 121 ORIGINAL RESEARCH ARTICLE NON-DESTRUCTIVE ONLINE REAL-TIME MILK QUALITY DETERMINATION IN A MILKING ROBOT USING NEAR-INFRARED SPECTROSCOPIC SENSING SYSTEM P. Iweka 1, S. Kawamura*1, T. Mitani2, and S. Koseki1 1 Laboratory of Agricultural and Food Process Engineering, Graduate School of Agricultural Science, Hokkaido University, Japan 2 Field Science Center for Northern Biosphere, Hokkaido University, Japan ARTICLE INFORMATION Received October, 2018 Accepted December, 2018 Keywords: Milk fat Protein Lactose Somatic cell count Dairy precision farming Japan ABSTRACT A near-infrared spectroscopic (NIRS) sensing system was developed on an experimental basis for the quality determination of three major milk constituents (fat, protein and lactose) and somatic cell count (SCC) of non-homogenized milk. The NIRS sensing system was used for acquiring NIR spectra of non-homogenized milk during milking in an automatic milking system (milking robot) over the wavelength range of 700 nm to 1050 nm. The three major milk constituents were analyzed for reference data using a MilkoScan instrument, while SCC was analyzed using a Fossomatic instrument. We developed calibration models using partial least square (PLS) regression analysis, and the precision and accuracy of the models was validated. The coefficients of determination (r2), standard errors of prediction (SEP) and bias were 0.98, 0.23% and 0.00% for fat, 0.72, 0.25% and 0.00% for protein, 0.54, 0.15% and 0.00% for lactose, and 0.63, 0.48 Log SCC/mL and 0.00 Log SCC/mL for SCC respectively. These results show that the NIRS sensing system developed in this study could be used for online real-time determination of milk quality in a milking robot. The system can provide dairy farmers with information on milk quality and physiological status of each cow and therefore, give them feedback control for improving dairy farm management. By using the NIRS system, dairy farmers will be able to produce high-quality milk and precision dairy farming will be realized. ©2018 Faculty of Engineering, University of Maiduguri, Nigeria. All rights reserve 1.0 Introduction Dairy farming involves a lot of work such as feeding, milking, livestock management, feed crop production and manure treatment. As usual, extensive dairy farmers manage their livestock in groups which is a system known as herd management (Svennersten-Sjaunja et al., 1997). However, a system known as individual cow management is essential for monitoring milk composition quality of each cow which is important for animal breeding, effective cow usage and feed management. Thus, this is the reason for the recent need for a technique that will enable dairy farmers to determine milk quality of individual cows during milking. http://www.azojete.com.ng mailto:shuso@bpe.agr.hokudai.ac.jp Iweka, et al,. Non-destructive online real-time milk quality determination in a milking robot using near-infrared spectroscopic sensing system. AZOJETE, 14(sp.i4):121-128. ISSN 1596-2490; e-ISSN 2545-5818, www.azojete.com.ng 122 The non-destructive, rapid, easy to use, time saving and pre-treatment free nature of near-infrared spectroscopy (NIRS) makes it an effective tool for analyzing milk quality during milking process, NIRS has been used to obtain qualitative and quantitative information of food and agricultural commodity such as rice (Kawamura et al., 2003a; Natsuga and Kawamura, 2006), wheat (Natsuga et al., 2001), and fruits and vegetables (Lakshmi et al., 2017). NIRS has been practically used in automatic rice-quality assessment in Japan (Kawamura et al., 2002; Kawamura et al., 2003a). NIRS has also been use for milk quality determination (Sato et al., 1987; Tsenkova et al., 2001; Kawamura et al., 2003b; Tsenkova et al., 2006; Kawamura et al., 2007; Kawasaki et al., 2008; Tsenkova et al., 2009; Iweka et al., 2016). However, the application of NIRS for online real-time monitoring of milk quality of individual cow has not been achieved. In this study, an experimental online near-infrared (NIR) spectroscopic sensing system was developed for milk quality determination. Iweka et al., (2016) reported that the NIR spectroscopic sensing system can be used for real-time determination of milk quality during milking with sufficient precision and accuracy. As a result of our findings, the NIR spectroscopic sensing system was installed in an automatic milking system. The objective of this study was to examine the accuracy of the NIR spectroscopic sensing system for milk quality determination in an automatic milking system. 2.0 Materials and Methods 2.1 Near-Infrared Spectroscopic Sensing System An experimental online near-infrared (NIR) spectroscopic sensing system was designed for analyzing milk quality of each cow during milking. The system consisted of an NIR instrument (NIR spectrum sensor and NIR spectrometer), milk flow meter, milk sampler and a laptop computer (Fig. 1). The system was installed in a milking robot system (GEA Farm Technologies, Westfaliasurge, Germany). Non-homogenized milk from the milking robot flowed continuously across a bypass into the milk chamber of the NIR spectrum sensor. Excess raw milk flowed past the milk flow meter and was then released through a line tube into the bucket. The volume of a milk sample in the chamber was about 30 mL. The optical axes of halogen lamps A and B and the optical fiber were set at the same level, but the optical axis for halogen lamp C was set at 5 mm higher the optical fiber (Fig. 2). The NIR instrument acquired absorbance spectra through the milk. Spectra were obtained in the wavelength range from 700 nm to 1050 nm at 1 nm intervals every 20 s during milking (Table 1). The milk flow rate was simultaneously recorded in the laptop computer. 2.2 Cows and Milk Samples Twenty six Holstein cows belonging to a dairy farm at Tochigi Prefecture, Japan were used for this study. These cows were at their different lactation stages. The experiment was conducted throughout the whole day for two consecutive days, that is; on the 22nd and 23rd of February 2018. Milking was automatically started as soon as a cow walked into the milking robot. Milk samples were collected from the milking sampler every 20 s during milking. http://www.azojete.com.ng Arid Zone Journal of Engineering, Technology and Environment, December, 2018; Vol. 14(sp.i4):121-128. ISSN 1596-2490; e-ISSN 2545-5818; www.azojete.com.ng 123 2.3 Reference Analyses Three major milk constituents (fat, protein and lactose) and somatic cell count (SCC) of non- homogenized milk were measured as milk quality items in this study. The milk constituents were determined using a MilkoScan instrument (Foss Electric, Hillerod, Denmark) and SCC were determined using Fossomatic instrument (Foss Electric, Hillerod, Denmark). The total number of samples used for reference analyses were 377 for milk constituents and SCC. Figure 1. Flow chart of an on-line near-infrared spectroscopic sensing system for determining milk quality in an automatic milking system Figure 2. Schematic diagram of the optical system of milk chamber of the NIR spectrum sensor 2.4 Chemometric Analyses Chemometric analyses were carried out to develop calibration models for milk quality parameters and to validate the precision and accuracy of the models. Spectra data analyses software (The Unscrambler ver. 10.3, Camo AS Trondheim, Norway) was used for the analyses. The total reference samples were used to develop calibration models. The calibration models were validated using full cross validation method. The statistical method of partial least squares (PLS) was used to develop ../../../user/Downloads/azojete143/www.azojete.com.ng Iweka, et al,. Non-destructive online real-time milk quality determination in a milking robot using near-infrared spectroscopic sensing system. AZOJETE, 14(sp.i4):121-128. ISSN 1596-2490; e-ISSN 2545-5818, www.azojete.com.ng 124 calibration models from the absorbance spectra and reference data. The best model was obtained when we used the original spectra data thus, pretreatment techniques such as multiplicative scatter correction, 2nd derivative and smoothing was not used. Table 1. Specifications of the near-infrared spectroscopic instrument Devices Specifications NIR spectrum sensor Absorbance spectrum sensor Light source Three halogen lamps Optical fiber Quartz Fiber Milk chamber surface Glass Volume of milk sample Approx. 30 mL Distance between optical axis and milk level 55 mm NIR spectrometer Diffraction grating spectrometer Optical density Absorbance Wavelength range 700 - 1050 nm, 1-nm internal Wavelength resolution Approx. 6.4 nm Photocell CMOS linear array, 512 pixels Thermal controller Heater and cooling fan Data processing computer Windows 7 A/D converter 16 bit Spectrum data acquisition Every 20 s 3. Results and Discussion 3.1 Near-infrared Spectra The original spectra of raw milk are shown in Fig. 3. The NIR spectra showed two bands peaks at around 740 nm and 840 nm indicate the overtone absorptions by C-H bands and C-C bands that are related to the distinctive absorption bands of milk constituents such as fat, protein and lactose. There was a strong absorption peak of O-H functional groups in water such that band around 960 nm were prominent spectra. Figure 3. Original spectra of non-homogenized milk from cow number 1 during milking on February 23, 2018 http://www.azojete.com.ng Arid Zone Journal of Engineering, Technology and Environment, December, 2018; Vol. 14(sp.i4):121-128. ISSN 1596-2490; e-ISSN 2545-5818; www.azojete.com.ng 125 3.2 Precision and Accuracy of Calibration Model The validation statistics of the NIR sensing system for milk quality determination are summarized in Table 2. Correlations between reference and NIR-predicted values of milk fat, protein, lactose and SCC are shown in Figures 4 to 7 respectively. The three major milk constituents are vital for milk quality determination. The quality of milk constituents are influenced by the physiological condition of each cow and cow feed. Thus, monitoring of milk constituents during milking everyday can be used for individual cow and feed management. The coefficient of determination (r2), standard error of prediction (SEP) and bias were 0.98, 0.23% and 0.00% for fat, 0.72, 0.25% and 0.00% for protein, and 0.54, 0.15% and 0.00% for lactose respectively. The high r2 values, small SEP values and the negligible bias values (zero) indicated that there were sufficient levels of precision and accuracy for predicting the three major milk constituents. The performance of calibration models for fat was excellent. The high performance of calibration model for fat was due to the fact that milk spectra had much information on fat content, starting from the scattering of light by fat globules to the absorption by C-H bands and C-C bands of triacylglycerol. The results showed that the NIR spectroscopic sensing system designed in our study can be used for online real-time milk constituent quality determination during milking by a milking robot. SCC is a recognized standard for mastitis diagnosis and it is a very important indicator for health and milk quality. Milk SCC can show the level of cow infection and it consequence on the mammary gland of dairy cows which is related to mastitis (Satu, 2003). Milk produced from the udder of a healthy cow contains less than 100,000 somatic cells per mL (i.e., 4logSCC/mL) while cows with subclinical mastitis produce milk containing more than 200,000 somatic cell per mL (i.e., 5.3logSCC/mL) (Satu, 2003). The values of r2, SEP and bias for SCC were 0.63, 0.48 Log SCC/mL and 0.00 Log SCC/mL respectively. The results obtained for SCC indicated that the precision and accuracy for predicting SCC was sufficiently high. Thus, the calibration model could be used for the diagnosis of subclinical mastitis 3.3 Dairy Precision Farming The installation of NIR spectroscopic sensing system developed in our study into a milking robot system would facilitate the monitoring of milk constituents and diagnosis of mastitis of individual cows in real-time during milking. The NIR sensing system could provide dairy farmers and veterinarians useful information on milk quality and physiological status of each cow and thus, give them assessment control for improving dairy farm management. The application of this NIR sensing system could take dairy farm management to the next level of dairy precision farming on the basis of individual cow information. 4.0 Conclusions The NIR spectroscopic sensing system developed in this study can be used for online real-time monitoring of fat, protein, lactose and SCC during milking by a milking robot with sufficiently high levels of precision and accuracy. By application, the NIR sensing system would enable dairy farmers to be able to produce high-quality milk and dairy precision farming will be actualized. ../../../user/Downloads/azojete143/www.azojete.com.ng Iweka, et al,. Non-destructive online real-time milk quality determination in a milking robot using near-infrared spectroscopic sensing system. AZOJETE, 14(sp.i4):121-128. ISSN 1596-2490; e-ISSN 2545-5818, www.azojete.com.ng 126 5.0 Acknowledgement This study was supported by a grant from the Project of NARO (National Agriculture and Food Research Organization, Japan) Bio-oriented Technology Research Advancement Institution (The Project for Development of New Practical Technology). The authors are grateful to NARO for the grant. Table 2. Validation statistics of the near-infrared sensing system for milk quality determination Milk quality items n Range r2 SEP Bias Regression Fat, % 377 0.98-8.54 0.98 0.23 0.00 y = 1.00 x + 0.00 Protein, % 377 2.73-4.46 0.72 0.25 0.00 y = 0.99 x + 0.04 Lactose, % 377 3.91-4.99 0.54 0.15 0.00 y = 0.98 x + 0.08 SCC, log SCC/mL 377 3.48-6.56 0.63 0.48 0.00 y = 0.98 x + 0.09 n: number of validation samples. r2: coefficient of determination. SEP: standard error of prediction. Regression line: Regression line from predicted value (x) to reference value (y) Figure 4. Correlation between reference fat content and NIRS-predicted fat content Figure 5. Correlation between reference protein content and NIRS-predicted protein content http://www.azojete.com.ng Arid Zone Journal of Engineering, Technology and Environment, December, 2018; Vol. 14(sp.i4):121-128. ISSN 1596-2490; e-ISSN 2545-5818; www.azojete.com.ng 127 References Iweka, P., Kawamura, S., Morita, A., Mitani, T., Okatani, T. and Koseki, S. 2016. Development of a near-infrared spectroscopic sensing system for milk quality evaluation during milking. In the 4th CIGR International Conference Proceedings of Agricultural Engineering, Aarhus, Denmark. Kawamura, S., Takekura, K. and Itoh, K. 2002. Accuracy of near-infrared transmission spectroscopy for determining rice constituent contents and improvement in the accuracy. ASAE Paper No. 023006. St. Joseph, Mich.: ASAE. Kawamura, S., Natsuga, M., Takekura, K. and Itoh, K. 2003a. Development of an automatic rice- quality inspection system. Computers and Electronics in Agriculture, 40(3): 115–126. Kawamura, S., Tsukahara, M., Natsuga, M. and Itoh, K. 2003b. On-line near infrared spectroscopic sensing technique for assessing milk quality during milking. ASAE Paper No. 033026. St. Joseph, Mich.: ASAE. Kawamura, S., Kawasaki, M., Nakatsuji, H. and Natsuga, M. 2007. Near-infrared spectroscopy sensing system for online monitoring of milk quality during milking. Sensing and Instrumentation for Food Quality and Safety, 1(1): 37–43. Kawasaki, M., Kawamura, S., Tsukahara, M., Morita, S., Komiya, M. and Natsuga, M. 2008. Near- infrared spectroscopic sensing system for on-line milk quality assessment in a milking robot. Computers and Electronics in Agriculture, 63(1): 22–27. Lakshmi S., Pandey, AK., Ravi, N., Chauhan, OP., Gopalan, N. and Sharma, RK. 2017. Non-destructive quality monitoring of fresh fruits and vegetables. Defence Life Science Journal, 2(2): 103–110. Natsuga, M., Kawamura, S. and Itoh, K. 2001. Determination of the quality of high-moisture wheat using near-infrared spectroscopy (Part 1). Journal of the Japanese Society of Agricultural Machinery, 63(1): 93–99. Figure 6. Correlation between reference lactose content and NIRS-predicted lactose content Figure 7. Correlation between reference SCC and NIRS-predicted SCC ../../../user/Downloads/azojete143/www.azojete.com.ng Iweka, et al,. Non-destructive online real-time milk quality determination in a milking robot using near-infrared spectroscopic sensing system. AZOJETE, 14(sp.i4):121-128. ISSN 1596-2490; e-ISSN 2545-5818, www.azojete.com.ng 128 Natsuga, M. and Kawamura, S. 2006. Visible and near-infrared reflectance spectroscopy for determining physicochemical properties of rice. Transactions of the ASABE, 49(4): 1069–1076. Sato, T. and Yoshino, M. 1987. Analysis of milk constituents by the near infrared spectrophotometric method. Nihon Chikusan Gakkaiho, 58(8): 698–706. Svennersten-Sjaunja, K., Sjaunja, LO., Bertilsson, J. and Wiktorsson, H. 1997. Use of regular milking recorded versus daily records for nutrition and other kinds of management. Livestock Production Science, 48(3): 167–174. Tsenkova, R., Atanassova, S., Kawano, S. and Toyoda, K. 2001. Somatic cell count determination in cow’s milk by near-infrared spectroscopy: A new diagnostic tool. Journal of Animal Science, 79(10): 2550–2557. Tsenkova, R., Atanassova, S., Morita, H., Ikuta, K., Toyoda, K., Iordanova, K. and Hakogi, E. 2006. Near infrared spectra of cows’ milk for milk quality evaluation: Disease diagnosis and pathogen identification. Journal of Near Infrared Spectroscopy, 14(6): 363–370. Tsenkova, R., Meilina, H., Kuroki, S. and Burns, D. 2009. Near infrared spectroscopy using short wavelengths and leave-one-cow-out cross-validation for quantification of somatic cells in milk. Journal of Near Infrared Spectroscopy, 17(6): 345–351. http://www.azojete.com.ng ARTICLE INFORMATION 1.0Introduction 2.0 Materials and Methods 2.2 Cows and Milk Samples 2.4 Chemometric Analyses 3. Results and Discussion 3.1 Near-infrared Spectra 3.2 Precision and Accuracy of Calibration Model 3.3 Dairy Precision Farming 4.0 Conclusions