Artificial Neuron Network Model Research Articles

Application of machine learning models for green andsimultaneous determination of asiaticoside andmadecassoside in Centella asiatica

The artificial neuron network (ANN), which is a subfield of machine learning, has been widely applied in analytical chemistry for classification/pattern recognition, prediction, and modeling. ANN combined with UV spectrometry can be used to tackle the problem of overlapping spectra of a complicated matrix of herbal medicine. In the present study, ANN has been used for simultaneous determination of asiaticoside andmadecassoside in Centella asiatica collected from various provinces in Viet Nam based on UV spectra of standard reference and spiked samples. The absorption spectra of 108 C. asiatica samples were recorded in (190-250) nm of wavelength with an interval of 1 nm (61 variables) were used for data acquisition. An ANN model using traincgb function with 40 neural hidden layers was trained. The correlation coefficients were all higher than 0.9999. The concentration of asiaticoside and madecassoside in all tested samples revealed a good recovery, as compared to the referenced values. TheANN model can be considered as effective, time-saving and environment-friendly quantitative analysis tool for herbal medicine ® 2024 Journal of Science and Technology - NTTU

Developing interpretable machine learning-Shapley additive explanations model for unconfined compressive strength of cohesive soils stabilized with geopolymer.

This paper seeks to develop an interpretable Machine Learning (ML) model for predicting the unconfined compressive strength (UCS) of cohesive soils stabilized with geopolymer at 28 days. Four models including Random Forest (RF), Artificial Neuron Network (ANN), Extreme Gradient Boosting (XGB), and Gradient Boosting (GB) are built. The database consists of 282 samples collected from the literature with three different types of cohesive soil stabilized with three geopolymer categories including Slag-based geopolymer cement, alkali-activated fly ash geopolymer and slag/fly ash-based geopolymer cement. The optimal model is selected by comparing their performances with each other. The values of hyperparameters are tuned by Particle Swarm Optimization (PSO) algorithm and K-Fold Cross Validation. Statistical indicators show the superior performance of the ANN model with three metrics performance such as coefficient of determination R2 = 0.9808, Root Mean Square Error RMSE = 0.8808 MPa and Mean Absolute Error MAE = 0.6344 MPa. In addition, a sensitivity analysis was performed to determine the influence of different input parameters on the UCS of cohesive soils stabilized with geopolymer. The order of feature effect can be ordered in descending order using the Shapley additive explanations (SHAP) value as follows: Ground granulated blast slag content (GGBFS) > Liquid limit (LL) > Alkali/Binder ratio (A/B) > Molarity (M) > Fly ash content (FA) > Na/Al > Si/Al. The ANN model can obtain the best accuracy using these seven inputs. LL has a negative correlation with the growth of unconfined compressive strength, whereas GGBFS has a positive correlation.

Assessing the accuracy of sensitivity analysis: an application for a cellular automata model of Bogota’s urban wetland changes

This study analyzes the outcomes of Cellular Automata (CA) with different neighborhood sizes and spatial resolution configurations on the performance of the Future Land Use Simulation (FLUS) model. The analysis is executed using three analogic images to extract the land use/land cover in Bogota, Colombia, for three years: 1998, 2004, and 2010. The FLUS model has an Artificial Neuronal Network model, which was used for calculating the relationships between the land uses and the associated drivers and to estimate the probability of occurrence of each land use. Whenever a CA is used to model and simulate, sensitivity analysis (SA) becomes a crucial step in CA modeling to understand better the influence of parameters’ changes in the simulation outcomes. Therefore, the SA is conducted by varying the neighborhood sizes between 3 × 3, 5 × 5, and 7 × 7 for 5 and 30 meters. In addition, cross-classification maps, Area Under the Curve (AUC) of the Total Operating Characteristic, landscape metrics, the figure of merit, Fuzzy Kappa, and disagreement metrics were calculated to assess how well the model performed. High AUC values and low disagreement results show that, in general, the model performed well, and the accuracy of the outputs improves with a 3 × 3 neighborhood size and 5 meters spatial resolution. This study provides a broad assessment approach to the different methods that must be considered to evaluate the sensitivity of CA models in the simulation of urban wetlands’ spatial-temporal evolution.

Forecast of Precipitation in Ai-Petri Area Based on Artificial Neuron Network Model

Forecast of Precipitation in Ai-Petri Area Based on Artificial Neuron Network Model

Analysis of the Performance of Cutting Tools of Tunnel Boring Machine (TBM) in Silty-Sand Soils Using Artificial Neural Network (ANN) – Case Study: Tabriz Metro Line 2 Project

The choice of cutting tools for soil excavation has a significant impact on the performance of tunnel boring machines. In this study, the data of two excavated sections of the Tabriz metro line 2 project have been used, and the crossed soils were sandy silts. Two sections have investigated the performance of two kinds of cutting tools (disk cutters and rippers). The cutting tools used in the first section of the drilling route are disc cutters, while in the second section, rippers have been used. To predict and evaluate the performance of the two types of cutting tools in silty-clay soils, artificial neuron networks were used, and the Levenberg-Marquardt algorithm was chosen for training the ANNs. Three important TBM parameters (torque, thrust force, and speed) were considered as input parameters, and the TBM penetration rate is chosen as the output variable in the developed artificial neuron network models. The obtained results showed that the best-predicted value obtained from the ANN was obtained with one hidden layer that contains four neurons. Finally, considering the performance parameter observed in silty soils, the ripper’s performance is better compared to the disc cutter one.

Artificial Neural Network Modeling to Predict the Effect of Milling Time and TiC Content on the Crystallite Size and Lattice Strain of Al7075-TiC Composites Fabricated by Powder Metallurgy

In the study, Al7075-TiC composites were synthesized by using a novel dual step blending process followed by cold pressing and sintering. The effect of ball milling time on the microstructure of the synthesized composite powder was characterized using X-ray diffraction measurements (XRD), scanning electron microscopy (SEM), energy dispersive spectroscopy (EDS), and transmission electron microscopy (TEM). Subsequently, the integrated effects of the two-stage mechanical alloying process were investigated on the crystallite size and lattice strain. The crystallite size and lattice strain of blended samples were calculated using the Scherrer method. The prediction of the crystallite size and lattice strain of synthesized composite powders was conducted by an artificial neural network technique. The results of the mixed powder revealed that the particle size and crystallite size improved with increasing milling time. The particle size of the 3 h-milled composites was 463 nm, and it reduces to 225 nm after 7 h of milling time. The microhardness of the produced composites was significantly improved with milling time. Furthermore, an artificial neuron network (ANN) model was developed to predict the crystallite size and lattice strain of the synthesized composites. The ANN model provides an accurate model for the prediction of lattice parameters of the composites.

Comparison of Arrhenius model and artificial neuronal network for predicting quality changes of frozen tilapia (Oreochromis niloticus)

The objectives of this study were to study the quality changes (ice crystal morphology, Ca2+-ATPase activity, total sulfhydryl [SH] content, intrinsic fluorescence intensity [IFI], and K value [freshness determination]) of tilapia at different storage temperatures for 112 days, and kinetic models and artificial neuronal network (ANN) were developed to predict the changes. There was a dramatic increase in cross-section area and equivalent diameter and a sharp decrease in Ca2+-ATPase activity and SH content during the first 4 weeks (p < 0.05). IFIλmax decreased by 43.95%, 29.77%, 28.97% and 18.58% after 16 weeks at 265 K, 259 K, 253 K, and 233 K. The kinetic model established by IFIλmax could be accurately described the quality changes during storage at 233–265 K. However, the prediction accuracy established by other indices decreased at later stages (14–16 weeks). The ANN model was superior to Arrhenius models and performed better for all indicators.

Investigation of ANN Model Containing One Hidden Layer for Predicting Compressive Strength of Concrete with Blast-Furnace Slag and Fly Ash

The prediction accuracy of concrete compressive strength is important and considered a challenging task, aiming at reducing costly and time-consuming experiments. Moreover, compressive strength prediction of concrete using blast-furnace slag (BFS) and fly ash (FA) is more difficult due to the complex mix design of a composition. In this investigation, an approach using the artificial neuron network (ANN), one of the most powerful machine learning algorithms, is applied to predict the compressive strength of concrete containing BFS and FA. The ANN models with one hidden layer containing 13 neuron number cases are proposed to determine the best ANN structure. Under the effect of random sampling strategies and the network structures selected, Monte Carlo simulations (MCS) are introduced to statistically investigate the convergence of results. Next, the evaluation of the model is concluded over 100 simulations for the convergence analysis. The results show that ANN is a highly efficient predictor of the compressive strength using BFS and FA, with maximum values of the coefficient of determination (R2), root mean square error (RMSE), and mean absolute error (MAE) of 0.9437, 3.9474, and 2.9074, respectively, on the training part and 0.9285, 4.4266, and 3.2971, respectively, for the testing part. The best-defined structure of ANN is [8-24-1], with 24 neurons in the hidden layer. Partial Dependence Plots (PDP) are also performed to investigate the dependence of the prediction results of input variables used in the ANN model. The age of sample and cement content are found to be the two most crucial factors that affect the compressive strength of concrete using BFS and FA. The ANN algorithm is practical for engineers to reduce costly experiments.

Development of freight travel demand model with characteristics of vehicle tour activities

The actual movements of freight vehicles demonstrate comprehensive tour activities instead of a single trip. Freight vehicle tours activity represent trip chaining relationship between shippers, receivers, and carriers as a journey of tours. In India, identifying and modeling tour activity is complex because disaggregated freight tour activity data are generally not available to access. Hence, there is a need to identify the most suitable and sufficient modeling framework, which explains the non-linear behavior and complexity of tour-based freight travel demand. Thus, the objectives of this study are to identify the driver, vehicle, and journey characteristics of freight vehicle movement and their tour activities using the conventional non-linear and soft computing methods to estimate the number of tours and to identify the most suitable and effective modeling method. The questionnaire form designed, and study locations were selected in Tiruchirappalli, India. The selected survey locations were identified based on the availability of fright drivers who were making tours. An extensive face to face data collection process was performed, and a total of 600 drivers' data were collected from survey. A detailed description of the collected data was obtained through descriptive analysis. Next, significant relationship between characteristics of freight vehicle drivers and their tour activities was identified based on various correlation tests. Further, a non-linear regression method was adopted for explaining non-linear behavior and complexity of tour-based freight travel demand. Followed by soft computing models using Support-vector machines (SVM), k-Nearest Neighbors (k-NN), and Artificial Neuron Network (ANN) algorithms were developed to predict tour-based freight travel demand, which performed better than conventional non-linear model. Among all, ANN model outperformed and explained the variations with better accuracy. Finally, the outcome of this study will be helpful to transport planners and researchers for understanding the characteristics of freight vehicle tour activity and estimating tour-based freight travel demand.

Influence of water-to-binder ratio on the optimum percentage of nano-TiO2 addition in terms of compressive strength of mortars: A laboratory and virtual experimental study based on ANN model

This research aims to study the effect of water-to-binder ratio (w/b) as a potential key factor that changes the optimum percentage of nano-TiO2 addition in terms of compressive strength of mortars. Data from previous literature were collected and analyzed. An artificial neuron network (ANN) model was trained and validated to predict the compressive strength of mortars as a function of the level of nano-TiO2, w/b, cement content, aggregate-to-binder ratio, and age. Then, a virtual experimental campaign evaluating 45,015 case scenarios, and using the ANN model, was performed to study the effect of the w/b on the optimum percentage of nano-TiO2. Results suggest that the higher the w/b and the lower the age, the higher the optimum percentage of nano-TiO2. Laboratory experimental compressive strength results are in agreement with these trends. Thermogravimetric analysis and 3D X-Ray scans show that the use of nano-TiO2 presents a double positive effect on compressive strength, reducing the porosity and increasing the hydration product content when the initial porosity of the reference cement paste is high (due to high w/b). However, when the porosity of reference samples is very low (due to low w/b), the use of nano-TiO2 may lead to a reduction of hydration products content due to the lack of space for the CH to grow. Since the initial porosity was already very low, the porosity reduction due to the use of nano-TiO2 would not be enough to overcome the negative effect on hydration. Therefore, it may lead to a decrease in the compressive strength of mortars. These results explain the dependence of the optimum percentage of nano-TiO2 on w/b and age.

Application of Artificial Neural Network for Prediction of Key Indexes of Corn Industrial Drying by Considering the Ambient Conditions

Uncontrollable ambient conditions are the main factors limiting the self-adaption control of an industrial drying system. To achieve the goal of accurate control of the drying process, the influence of the ambient conditions on the drying behavior should be taken into consideration when modeling the drying process. Present work introduced an industrial drying system with a loading capacity of 50 t, two artificial neural network prediction models with (IANN) and without (OANN) considering the ambient conditions were established using artificial neural network modeling approach. The ambient conditions on the moisture content (MC), exergy efficiency of the heat exchanger (ηex,h) and specific recovered radiant energy (Er) of the drying process were also investigated. The results showed that the ηex,h and Er increase with the increase of ambient temperature while the drying time decrease with the increase of the ambient temperature. The IANN model has a better prediction performance that that of OANN model. An optimal architecture of 9-2-12-3 artificial neuron network model was developed and the best prediction performance of the artificial neural network (ANN) model were found at a training epoch number of 30, and a momentum coefficient of 0.4, where the coefficient of determination of moisture content, exergy efficiency of heat exchanger, and the specific recovered radiant energy, respectively are 0.998, 0.992, and 0.980, indicating that the model has an excellent prediction performance and can be used in engineering practice.

606-P: Artificial Neuronal Network Model for Three-Month Prognosis in Diabetic Foot Syndrome

Background: To make a reliable prognosis on wound healing in patients with diabetic foot syndrome (DFS) is extremely difficult. We used the artificial neuronal network (ANN) to identify the most significant variables which affect DFS healing process. We also aimed at providing data for designing a digital application which would help practitioners to predict the course of DFS. Methods: We enrolled 213 DFS patients and examined them using a variety of diagnostic tests. The patients were followed for three months. In the initial model, we assessed 35 clinical and biochemical variables. Subsequently, with the help of traditional statistics we reduced their number to twelve, and after conducting the sensitivity analysis, we found out that the final number of significant variables is as low as six. Results: The most significant variables in predicting the outcome of DFS treatment were: probe-to-bone test result, presence of blood flow in Doppler probe, prior amputation within the foot, erythrocyte sedimentation rate, and the area and duration of the ulceration. The area under the ROC curve was 0.87 (Figure). The total accuracy was 85%, sensitivity 94.6%, specificity 66% and F1 score 89%. Conclusions: ANN can be used in the prediction of the DFS course. The algorithm, which is the source of a digital application, is particularly useful in identifying individuals with diabetic foot ulcerations who fail to be healed in three months. Disclosure A.A. Poradzka: Other Relationship; Self; Boehringer Ingelheim (Canada) Ltd., Urgo Medical. L. Czupryniak: None.

Coupling an artificial neuron network daylighting model and building energy simulation for vacuum photovoltaic glazing

Window plays an essential role in the indoor environment and building energy consumption. As an innovative building integrated photovoltaic (BIPV) window, the vacuum PV glazing was proposed to provide excellent thermal performance and utilize renewable energy. However, the daylighting performance of the vacuum PV glazing and the effect on energy consumption have not been thoroughly investigated. Most whole building energy simulation used the daylighting calculation based on Daylight Factor (DF) method, which fails to address realistic calculation for direct sunlight through complex glazing materials. In this study, a RADIANCE model was developed and validated to adequately represent the daylight behaviour of a vacuum cadmium telluride photovoltaic glazing with a three-layer structure. However, RADIANCE will consume too many computational resources for a whole year simulation. Therefore, an artificial neuron network (ANN) model was trained based on the weather conditions and the RADIANCE simulation results to predict the interior illuminance. Subsequently, a preprocessing coupling method is proposed to determine the lighting consumption of a typical office with the vacuum PV glazing. The performance evaluation of the ANN model indicates that it can predict the illuminance level with higher accuracy than the daylighting calculation methods in EnergyPlus. Therefore, the ANN model can adequately address the complex daylighting response of the vacuum PV glazing. The proposed coupling method showed a more reliable outcome than the simulations sole with EnergyPlus. Furthermore, the computational cost can be reduced dramatically by the ANN daylighting prediction model in comparison with the RADIANCE model. Compared with the lighting consumption determined by the ANN-based coupling method, the two approaches in EnergyPlus, the split-flux method and the DElight method, tend to underestimate the lighting consumption by 5.3% and 9.7%, respectively.

Models for estimating daily photosynthetically active radiation in oceanic and mediterranean climates and their improvement by site adaptation techniques

In this work Photosynthetically Active Radiation (PAR) in oceanic and mediterranean climates is modeled. Twenty-two different models have been developed and tested: eleven Multilinear Regression (MR) models and eleven Artificial Neuron Network (ANN) models, using combinations of variables such as Global Horizontal Irradiance (GHI), Global Extraterrestrial Irradiance (G0), Temperature (T) and Relative Humidity (RH). Data provided by Satellite Application Facility on Climate Monitoring (CM SAF) are used to develop and train the models, while the models have been validated using field data from four stations located in Spain, covering the different study climates.According to the results, zones with different climate conditions need different models, both for the case of MR and ANN. The results show the need of including the GHI in all models in order to obtain accurate estimates; in fact, the presence of more variables only improves slightly the results in mediterranean climate, while in oceanic climate no improvement is observed.On the other hand, comparing MR and ANN models, ANN models did not show better results than those of MR models in no one of the cases studied. Regarding the climate, both types of models are clearly better for the mediterranean case than for the oceanic one. In order to improve the performance of the model for oceanic climate a correction based on the site adaptation technique was carried out. The good results obtained by this technique fully justify its use.The best proposed models provide better performance than other models which are restricted to certain locations. Besides, the clustering technique based on the PAR variable, used in this work, allows obtaining useful models for a whole region. Finally, another advantage of this methodology is that there is no need of ground measurements for its development, except for the site adaptation technique.

Micro-combined heat and power using dual fuel engine and biogas from discontinuous anaerobic digestion

The modeling of the Micro-CHP unit operating in dual-fuel mode is performed based on experimental results carried out at the laboratory scale. The engine tests were performed on an AVL engine, with a maximum power of 3.5 kW, using conventional diesel as pilot fuel and synthetic biogas as primary fuel. The biogas flow rate is evaluated using the experimental results from the literature, based on the anaerobic digestion in batch reactor of a mixture of 26% of Oat Straw and 74% of Cow Manure, diluted to contain only 4% of volatile solid.The engine operation was modeled using the Artificial Neuron Network (ANN) method. Experimental engine tests were used as a database for training and validation phases of ANN models. Three different ANN models are developed to model respectively the pilot fuel flow rate, the airflow rate and the exhaust gas temperature. Engine power output, biogas flow rate and biogas methane content were used as the same input layer.Given that the evolution of the biogas flow evolves along the entire digestion duration (50 days), the simulation work is performed by varying the number of digesters to be used in parallel mode. It is obtained that the optimal operation condition, minimizing the number of digesters and using less than 10% of the energy from diesel fuel, is to use 5 digesters and run the engine under load of 70%. It is concluded that a micro-CHP unit of 1 kWe, requires a dual fuel generator with a nominal power of 1 kWe, five digesters and a daily availability of effluents of 171 kg/day, consisting of 45 kg/day of oat straw and 126 kg/day of cow manure. It can also produce up to 2.45 kW of thermal power from the exhaust.

Heat stress risk in European dairy cattle husbandry under different climate change scenarios – uncertainties and potential impacts

Abstract. In the last decades, a global warming trend was observed. Along with the temperature increase, modifications in the humidity and wind regime amplify the regional and local impacts on livestock husbandry. Direct impacts include the occurrence of climatic stress conditions. In Europe, cows are economically highly relevant and are mainly kept in naturally ventilated buildings that are most susceptible to climate change. The high-yielding cows are particularly vulnerable to heat stress. Modifications in housing management are the main measures taken to improve the ability of livestock to cope with these conditions. Measures are typically taken in direct reaction to uncomfortable conditions instead of in anticipation of a long-term risk for climatic stress. Measures that balance welfare, environmental and economic issues are barely investigated in the context of climate change and are thus almost not available for commercial farms. Quantitative analysis of the climate change impacts on animal welfare and linked economic and environmental factors is rare. Therefore, we used a numerical modeling approach to estimate the future heat stress risk in such dairy cattle husbandry systems. The indoor climate was monitored inside three reference barns in central Europe and the Mediterranean regions. An artificial neuronal network (ANN) was trained to relate the outdoor weather conditions provided by official meteorological weather stations to the measured indoor microclimate. Subsequently, this ANN model was driven by an ensemble of regional climate model projections with three different greenhouse gas concentration scenarios. For the evaluation of the heat stress risk, we considered the number and duration of heat stress events. Based on the changes in the heat stress events, various economic and environmental impacts were estimated. The impacts of the projected increase in heat stress risk varied among the barns due to different locations and designs as well as the anticipated climate change (considering different climate models and future greenhouse gas concentrations). There was an overall increasing trend in number and duration of heat stress events. At the end of the century, the number of annual stress events can be expected to increase by up to 2000, while the average duration of the events increases by up to 22 h compared to the end of the last century. This implies strong impacts on economics, environment and animal welfare and an urgent need for mid-term adaptation strategies. We anticipated that up to one-tenth of all hours of a year, correspondingly one-third of all days, will be classified as critical heat stress conditions. Due to heat stress, milk yield may decrease by about 2.8 % relative to the present European milk yield, and farmers may expect financial losses in the summer season of about 5.4 % of their monthly income. In addition, an increasing demand for emission reduction measures must be expected, as an emission increase of about 16 Gg of ammonia and 0.1 Gg of methane per year can be expected under the anticipated heat stress conditions. The cattle respiration rate increases by up to 60 %, and the standing time may be prolonged by 1 h. This causes health issues and increases the probability of medical treatments. The various impacts imply feedback loops in the climate system which are presently underexplored. Hence, future in-depth studies on the different impacts and adaptation options at different stress levels are highly recommended.

Cystic duct cancer: Should it be deemed as a type of gallbladder cancer?

BACKGROUNDAccording to the latest American Joint Committee on Cancer and Union for International Cancer Control manuals, cystic duct cancer (CC) is categorized as a type of gallbladder cancer (GC), which has the worst prognosis among all types of biliary cancers. We hypothesized that this categorization could be verified by using taxonomic methods.AIMTo investigate the categorization of CC based on population-level data.METHODSCases of biliary cancers were identified from the Surveillance, Epidemiology, and End Results 18 registries database. Together with routinely used statistical methods, three taxonomic methods, including Fisher’s discriminant, binary logistics and artificial neuron network (ANN) models, were used to clarify the categorizing problem of CC.RESULTSThe T staging system of perihilar cholangiocarcinoma [a type of extrahepatic cholangiocarcinoma (EC)] better discriminated CC prognosis than that of GC. After adjusting other covariates, the hazard ratio of CC tended to be closer to that of EC, although not reaching statistical significance. To differentiate EC from GC, three taxonomic models were built and all showed good accuracies. The ANN model had an area under the receiver operating characteristic curve of 0.902. Using the three models, the majority (75.0%-77.8%) of CC cases were categorized as EC.CONCLUSIONOur study suggested that CC should be categorized as a type of EC, not GC. Aggressive surgical attitude might be considered in CC cases, to see whether long-term prognosis could be immensely improved like the situation in EC.

Detection of Common Causes between Air Traffic Serious and Major Incidents in Applying the Convolution Operator to Heinrich Pyramid Theory

Heinrich’s pyramid theory is one of the most influential theories in accident and incident prevention, especially for industries with high safety requirements. Originally, this theory established a quantitative correlation between major injury accidents, minor injury accidents and no-injury accidents. Nowadays, researchers from different fields of engineering also apply this theory in establishing quantitatively the correlation between accidents and incidents. In this work, on the one hand, we have detected the applicability of this theory by studying incident reports of different severities occurred in air traffic management. On the other hand, we have deepened the analysis of this theory from a qualitative perspective. For this purpose, we have applied the convolution operator in identifying correlations between contributing causes to different incident severities, also known as precursors to accidents, and system failures. The results suggested that system failures are mechanisms by which the causes are manifested. In particular, the same underlying cause can be manifested through different failures which contribute to incidents with different severities. Finally, deriving from this result, an artificial neuronal network model is proposed to recognize future causes and their possible associated incident severities.

Development of short chain fatty acid-based artificial neuron network tools applied to biohydrogen production

The biological production of biohydrogen through dark fermentation is a very complex system where the use of an artificial neuron network (ANN) for prediction, controlling and monitoring has a great potential. In this study three ANN models based on volatile fatty acids (VFA) production and speciation were evaluated for their capacity to predict (i) accumulated H2 production, (ii) hydrogen production rate and (iii) H2 yield. Lab-scale biohydrogen and VFA production kinetics from a previous study were used for training and validation of the models. The input parameters studied were: time and acetate and butyrate concentrations (model 1), time and lactate, acetate, propionate and butyrate concentrations (model 2), time and the sum of all VFA (model 3) and time and butyrate/acetate (model 4). All models could predict biohydrogen accumulated production, hydrogen production rate and H2 yield with high accuracy (R2 > 0.987). VFAT is the input parameter indicated for processes using pure cultures, while for complex/mixed cultures a model based on acetate and butyrate is recommended.

Machine learning models for the estimation of monthly mean daily reference evapotranspiration based on cross-station and synthetic data

AbstractThe estimation of reference evapotranspiration (ET0) is important in hydrology research, irrigation scheduling design and water resources management. This study explored the capability of eight machine learning models, i.e., Artificial Neuron Network (ANN), Random Forest (RF), Gradient Boosting Decision Tree (GBDT), Extreme Gradient Boosting (XGBoost), Multivariate Adaptive Regression Spline (MARS), Support Vector Machine (SVM), Extreme Learning Machine and a novel Kernel-based Nonlinear Extension of Arps Decline (KNEA) Model, for modeling monthly mean daily ET0 using only temperature data from local or cross stations. These machine learning models were also compared with the temperature-based Hargreaves–Samani equation. The results indicated that the estimation accuracy of these machine learning models differed in various scenarios. The tree-based models (RF, GBDT and XGBoost) exhibited higher estimation accuracy than the other models in the local application. When the station has only temperature data, the MARS and SVM models were slightly superior to the other models, while the ANN and HS models performed worse than the others. When there was no temperature data at the target station and the data from adjacent stations were used instead, MARS, SVM and KNEA were the suitable models. The results can provide a solution for ET0 estimation in the absence of complete meteorological data.