Multiple AI model integration strategy : Application to saturated hydraulic conductivity prediction from easily available soil properties
استراتژی یکپارچه سازی مدل هوش مصنوعی چندگانه: کاربرد در پیش بینی هدایت هیدرولیکی اشباع شده از خصوصیات خاک که به راحتی در دسترس است-2020
A multiple model integration scheme driven by artificial neural network (ANN) (MM-ANN) was developed and tested to improve the prediction accuracy of soil hydraulic conductivity (Ks) in Tabriz plain, an arid region of Iran. The soil parameters such as silt, clay, organic matter (OM), bulk density (BD), pH and electrical conductivity (EC) were used as model inputs to predict soil Ks. Standalone models including multivariate adaptive regression splines (MARS), M5 model tree (M5Tree), support vector machine (SVM) and extreme learning machine (ELM) were also implemented for comparative evaluation with MM-ANN model predictions. Based on several performance indicators such as Nash Sutcliffe Efficiency (NSE), results showed that the calibrated MMANN model involving the predictions of MARS, M5Tree, SVM and ELM models by considering all the soil parameters used in this study as inputs provided superior soil Ks estimates. The proposed hybrid model (MMANN) emerged as a reliable intelligence model for the assessment of soil hydraulic conductivity with an NSE=0.939 & 0.917 during training and testing, respectively. Accurate prediction of field-scale soil hydraulic conductivity is crucial from the view point of agricultural sustainability and management prospects.
Keywords: Saturated hydraulic conductivity | Extreme learning machine | Multiple model strategy | Multivariate adaptive regression splines | M5Tree | Support | vector machine | Prediction
Reinforcement learning based optimizer for improvement of predicting tunneling-induced ground responses
بهینه ساز مبتنی بر یادگیری تقویتی برای بهبود پیش بینی پاسخ های ناشی از tunneling-2020
Prediction of ground responses is important for improving performance of tunneling. This study proposes a novel reinforcement learning (RL) based optimizer with the integration of deep-Q network (DQN) and particle swarm optimization (PSO). Such optimizer is used to improve the extreme learning machine (ELM) based tunnelinginduced settlement prediction model. Herein, DQN-PSO optimizer is used to optimize the weights and biases of ELM. Based on the prescribed states, actions, rewards, rules and objective functions, DQN-PSO optimizer evaluates the rewards of actions at each step, thereby guides particles which action should be conducted and when should take this action. Such hybrid model is applied in a practical tunnel project. Regarding the search of global best weights and biases of ELM, the results indicate the DQN-PSO optimizer obviously outperforms conventional metaheuristic optimization algorithms with higher accuracy and lower computational cost. Meanwhile, this model can identify relationships among influential factors and ground responses through selfpracticing. The ultimate model can be expressed with an explicit formulation and used to predict tunnelinginduced ground response in real time, facilitating its application in engineering practice.
Keywords: Tunnel | Ground response | Reinforcement learning | Extreme learning machine | Optimization