Artificial Neural Network and Grid Search Aided Optimization of Temperature Profile of Temperature Gradient Reactor for Dimethyl Ether Synthesis from Syngas

Abstract

The temperature setting of a fixed bed reactor with a temperature gradient (TGR, temperature gradient reactor) was optimized using an artificial neural network (ANN) and grid search to attain high one-pass CO conversion for one-step dimethyl ether (DME) synthesis from syngas (3CO + 3H 2 f DME + CO 2 ). In the TGR, the catalyst bed was divided into 5 zones in series, and the temperature of each zone was optimized. Experiments were designed using an orthogonal array, and the experimental result was used for training the ANN to correlate the temperature setting and CO conversion. A grid search on the trained ANN was applied to find the optimum temperature setting. TGR was effective in overcoming both the equilibrium limit of the reaction at high temperature and the low activity of the catalyst at low temperature. To attain high CO conversion, Cu-Zn-Al-Ti-Nb-V-Cr catalysts with the optimized composition for each reaction temperature and γ-alumina were packed into the 5 zones of the TGR. As a result, a high one-pass conversion of CO at 82% was attained at 1 MPa, W/F ) 50 g-cat•h/mol by means of the combination of the optimum catalyst and TGR. The CO conversion is much higher in comparison to the 72% found in TGR with a standard Cu catalyst, and to 69.5% in the isothermal reactor at 523K with a standard Cu catalyst.

Advances in Machine Learning for Monitoring, Control, and Optimization of Temperature of Reactors

“…Omata and Yamada [68] faced the challenge of optimizing the temperature profile in a temperature gradient reactor (TGR) for dimethyl ether synthesis. To address this challenge, they employed an ANN and grid search approach.…”

Probabilistic Forecasting Method of Metro Station Environment Based on Autoregressive LSTM Network

“…For the proposed method, an external environmental parameter will be retained as an input variable when its influence degree, w i , is larger than g. So, g is regarded as a hyperparameter to determine the input variables of the network. In general, a grid search and a manual search are the most widely used strategies for hyperparameter optimization [46][47][48]. Because most of the influence degrees are less than 0.3 according to Figure 8, all available values of g are set as 0.1, 0.2, and 0.3, respectively.…”

Recognition of Free Tryptophan in Water by Synthetic Pseudopeptides: Fluorescence and Thermodynamic Studies

“…The most difficult step in the synthetic scheme is the macrocyclization reaction to obtain macrocycle 2 b. In this case, careful control of the temperature, with the use of a temperature gradient [47,52] and under conditions similar to those used for related pseudopeptidic macrocycles, provided significantly better yields than by performing the reaction at reflux. Thus, macrocycle 2 b was prepared from the reaction of bis(amino amide) 1 b with 1,3-bis(bromomethyl)benzene in acetonitrile in the presence of tetrabutylammonium bromide (TBABr) and N,N-diisopropylethylamine (DIPEA) with a temperature gradient of 50-81 8C.…”