• Journal of Korean Society for Quality Management
• /
• v.48 no.3
• /
• pp.511-520
• /
• 2020

Purpose: The purpose of this study was to devise an accurate machine learning model for predicting silica concentrations following the addition of impurities, through time series analysis of mining data. Methods: The mining data were preprocessed and subjected to time series analysis using the machine learning model. Through correlation analysis, valid variables were selected and meaningless variables were excluded. To reflect changes over time, dependent variables at baseline were treated as independent variables at later time points. The relationship between independent variables and the dependent variable after n point was subjected to Pearson correlation analysis. Results: The correlation (R2) was strongest after 3 hours, which was adopted as a dependent variable. According to root mean square error (RMSE) data, the proposed method was superior to the other machine learning methods. The XGboost algorithm showed the best predictive performance. Conclusion: This study is important given the current lack of machine learning studies pertaining to the domestic mining industry. In addition, using time series analysis in mining data will show further improvement. Before establishing a predictive model for the proposed method, predictions should be made using data with time series characteristics. After doing this work, it should also improve prediction accuracy in other domains.

• Food Science of Animal Resources
• /
• v.35 no.1
• /
• pp.108-113
• /
• 2015

The purpose of this study was to develop predictive models for the growth of Listeria monocytogenes in pork Bulgogi at various storage temperatures. A two-strain mixture of L. monocytogenes (ATCC 15313 and isolated from pork Bulgogi) was inoculated on pork Bulgogi at 3 Log CFU/g. L. monocytogenes strains were enumerated using general plating method on Listeria selective medium. The inoculated samples were stored at 5, 15, and $25^{\circ}C$ for primary models. Primary models were developed using the Baranyi model equations, and the maximum specific growth rate was shown to be dependent on storage temperature. A secondary model of growth rate as a function of storage temperature was also developed. As the storage temperature increased, the lag time (LT) values decreased dramatically and the specific growth rate of L. monocytogenes increased. The mathematically predicted growth parameters were evaluated based on the modified bias factor ($B_f$), accuracy factor ($A_f$), root mean square error (RMSE), coefficient of determination ($R^2$), and relative errors (RE). These values indicated that the developed models were reliably able to predict the growth of L. monocytogenes in pork Bulgogi. Hence, the predictive models may be used to assess microbiological hygiene in the meat supply chain as a function of storage temperature.

• Food Science of Animal Resources
• /
• v.29 no.1
• /
• pp.47-54
• /
• 2009

Predictive modeling was applied to study the growth of microorganisms related to spoilage in frankfurter sausage containing various levels of dietary fiber (0, 1, 2, and 3%) from rice bran and to estimate its shelf-life. Using the Baranyi model, total viable cells, anaerobic and psychrotrophic bacteria were measured during 35 days of cold storage ($<4{\pm}1^{\circ}C$). The lag times (LT) demonstrated by control and treatment groups were 6.28, 623, 6.24, and 6.25 days, respectively. The growth rate of total viable cells in each group were 0.95, 0.91, 0.92, and 0.91 (Log CFU/g/day), respectively. The anaerobic and psychrotrophic bacteria had lower initial ($y_0$) and maximal bacterial counts ($y_{max}$) than total viable cells. Also, the anaerobic and psychrotrophic bacteria possessed lower growth rate and longer lag time than total viable cells. The estimated shelf-life of frankfurter containing rice bran fiber by the growth rate of total viable cells was 7.8, 7.9, 7.9, and 7.7 days, respectively. There were no significant differences in shelf-life as a function of fiber content. In other words, the addition of dietary fiber in sausage did not show the critically hazardous results in growth of microorganism. The 12 predictive models were then characterized by high $R^2$, and small RMSE. Furthermore, $B_f$ and $A_f$ values showed a very close relationship between the predictive and observed data.

• Asian-Australasian Journal of Animal Sciences
• /
• v.33 no.12
• /
• pp.2039-2049
• /
• 2020

Objective: This study was conducted to provide models to accurately predict nitrogen (N) and phosphorus (P) excretion of dairy cows on smallholder farms in Indonesia based on readily available farm data. Methods: The generic model in this study is based on the principles of the Lucas equation, describing the relation between dry matter intake (DMI) and faecal N excretion to predict the quantity of faecal N (Q_FN). Excretion of urinary N and faecal P were calculated based on National Research Council recommendations for dairy cows. A farm survey was conducted to collect input parameters for the models. The data set was used to calibrate the model to predict Q_FN for the specific case. The model was validated by comparing the predicted quantity of faecal N with the actual quantity of faecal N (Q_FNACT) based on measurements, and the calibrated model was compared to the Lucas equation. The models were used to predict N and P excretion of all 144 dairy cows in the data set. Results: Our estimate of true N digestibility equalled the standard value of 92% in the original Lucas equation, whereas our estimate of metabolic faecal N was -0.60 g/100 g DMI, with the standard value being -0.61 g/100 g DMI. Results of the model validation showed that the R² was 0.63, the MAE was 15 g/animal/d (17% from Q_FNACT), and the RMSE was 20 g/animal/d (22% from Q_FNACT). We predicted that the total N excretion of dairy cows in Indonesia was on average 197 g/animal/d, whereas P excretion was on average 56 g/animal/d. Conclusion: The proposed models can be used with reasonable accuracy to predict N and P excretion of dairy cattle on smallholder farms in Indonesia, which can contribute to improving manure management and reduce environmental issues related to nutrient losses.

• Environmental Engineering Research
• /
• v.25 no.4
• /
• pp.479-487
• /
• 2020

Current assumptions are used in the formulation of pseudo-first (PFO) and second-order (PSO) models to describe the kinetic data of filtration based on ideal operating conditions. This paper presents a new model developed with pseudo n^th order and based on real assumption. A comparison was performed between PFO, PSO and the new model to highlight their performance and the optimisation of the pseudo-order equation, using MATLAB software. Adsorption characteristic of bovine serum albumin adsorption on the track-etched membrane are used as a medium based on protein filtration data were extracted from the literature for different concentrations to demonstrate the comparison between PFO/PSO and the new model. The pseudo first and second-order kinetic models were applied to test the experimental data and they did not provide reasonable values. The results show that the predicted values are consistent with experimental values giving a good correlation coefficient R² = 0.997 and a minimum root mean squared error RMSE = 0.0171. Indeed, the experimental results follow the new model and the optimal pseudo equation order n = 1.115, the most suitable curves for the new model. As a result, we used different experimental adsorption data from the literature to examine and check the applicability and validity of the model.

• Journal of Korean Society on Water Environment
• /
• v.29 no.1
• /
• pp.29-35
• /
• 2013

Recently, various Best Management Practices (BMPs) have been applied at a field to reduce soil erosion. Hourly Runoff and Sediment Model for Best Management Practices (HRSM4BMP) model could be used to evaluate soil erosion reduction for various agricultural BMPs at fields. Runoff and sediment yield from source areas have to be predicted with greater accuracies to evaluate sediment reduction efficiently with BMPs. To achieve this, the best parameters related with runoff and sediment modules of the HRSM4BMP model should be identified with proper calibration processes. Although manual calibration is often utilized in calibrating runoff and sediment using the HRSM4BMP, objective calibration method would be recommended. The purpose of the study was to develop an automatic calibration tool of the HRSM4BMP model with PARASOL method. This automatic calibration tool was applied to Bangdongri, Chuncheon-si to evaluate its calibration performance. The $R^2$, NSE and RMSE value for runoff estimation were 0.92, 0.92, $0.3m^3$, and for sediment yield estimation were 0.94, 0.94, 0.0027 kg. As shown in this result, automatic calibration tool of HRSM4BMP model would be used to determine the best parameters and can be used to simulate runoff and sediment yield with acceptable accuracies.

• Korean Journal of Agricultural and Forest Meteorology
• /
• v.22 no.3
• /
• pp.128-134
• /
• 2020

The daily gridded meteorological information and climatology with high resolution (30m and 270m) was produced from 94 Automated Surface Observing System (ASOS) of Korea Meteorological Administration (KMA) for the past 50 years (1971-current) by different downscaling methods. In addition, the difference between daily meteorological data and the mean state of past 30 years (1981-2010) was calculated for the analysis of climate change. These datasets with GeoTiff format are available from the web interface (https://agecoclim. agmet.kr). The performance of the data is evaluated using 172 Automatic Weather S tation (AWS ) of Rural Development of Administration (RDA). The data have biases lower than 2.0, and root mean square errors (RMSE) lower than 3.8. This data may help to better understand the regional climatic change and its impact on agroecosystem in S outh Korea.

• Advances in environmental research
• /
• v.5 no.3
• /
• pp.153-167
• /
• 2016

We applied multilayer perceptron (MLP) and radial basis function (RBF) neural network in upstream and downstream water quality stations of the Karaj Reservoir in Iran. For both neural networks, inputs were pH, turbidity, temperature, chlorophyll-a, biochemical oxygen demand (BOD) and nitrate, and the output was dissolved oxygen (DO). We used an MLP neural network with two hidden layers, for upstream station 15 and 33 neurons in the first and second layers respectively, and for the downstream station, 16 and 21 neurons in the first and second hidden layer were used which had minimum amount of errors. For learning process 6-fold cross validation were applied to avoid over fitting. The best results acquired from RBF model, in which the mean bias error (MBE) and root mean squared error (RMSE) were 0.063 and 0.10 for the upstream station. The MBE and RSME were 0.0126 and 0.099 for the downstream station. The coefficient of determination ($R^2$) between the observed data and the predicted data for upstream and downstream stations in the MLP was 0.801 and 0.904, respectively, and in the RBF network were 0.962 and 0.97, respectively. The MLP neural network had acceptable results; however, the results of RBF network were more accurate. A sensitivity analysis for the MLP neural network indicated that temperature was the first parameter, pH the second and nitrate was the last factor affecting the prediction of DO concentrations. The results proved the workability and accuracy of the RBF model in the prediction of the DO.

• Korean Journal of Soil Science and Fertilizer
• /
• v.50 no.5
• /
• pp.336-344
• /
• 2017

The soil water characteristics curve (SWCC) represents the relation between soil water potential and soil water content. The shape and range of SWCC according to the relation could vary depending on soil characteristics. The objective of the study was to estimate SWCC depending on soil types and layers and to analyze the trend among them. To accomplish this goal, the unsaturated three soils were considered: silty clay loam, loam, and sandy loam soils. Weighable lysimeters were used for exactly measuring soil water content and soil water potential. Two fitting models, van Genuchten and Campbell, were applied. Two models entirely fitted well the measured SWCC, indicating low RMSE and high $R^2$ values. However, the large difference between the measured and the estimated was found at the 30 cm layer of the silty clay loam soil, and the gap was wider as soil water potential increased. In addition, the non-linear decrease of soil water content according to the increase of soil water potential tended to be more distinct in the sandy loam soil and at the 10 cm layer than in the silty clay loam soil and at the lower layers. These might be seen due to the various factors such as not only pore size distribution, but also cracks by high clay content and plow pan layers by compaction. This study clearly showed difficulty in the estimation of SWCC by such kind of factors.

• Journal of Korean Society of Industrial and Systems Engineering
• /
• v.43 no.2
• /
• pp.48-55
• /
• 2020

Recently, a study of prognosis and health management (PHM) was conducted to diagnose failure and predict the life of air craft engine parts using sensor data. PHM is a framework that provides individualized solutions for managing system health. This study predicted the remaining useful life (RUL) of aeroengine using degradation data collected by sensors provided by the IEEE 2008 PHM Conference Challenge. There are 218 engine sensor data that has initial wear and production deviations. It was difficult to determine the characteristics of the engine parts since the system and domain-specific information was not provided. Each engine has a different cycle, making it difficult to use time series models. Therefore, this analysis was performed using machine learning algorithms rather than statistical time series models. The machine learning algorithms used were a random forest, gradient boost tree analysis and XG boost. A sliding window was applied to develop RUL predictions. We compared model performance before and after applying the sliding window, and proposed a data preprocessing method to develop RUL predictions. The model was evaluated by R-square scores and root mean squares error (RMSE). It was shown that the XG boost model of the random split method using the sliding window preprocessing approach has the best predictive performance.