Search

Li3VO4 (LVO) is a highly promising anode material for lithium-ion batteries, owing to its high capacity and stable discharge plateau. However, LVO faces a significant challenge due to its poor rate capability, which is mainly attributed to its low electronic conductivity. To enhance the kinetics of lithium ion insertion and extraction in LVO anode materials, a conductive polymer called poly(3,4-ethylenedioxythiophene):poly(styrenesulfonate) (PEDOT:PSS) is applied to coat the surface of LVO. This uniform coating of PEDOT:PSS improves the electronic conductivity of LVO, thereby enhancing the corresponding electrochemical properties of the resulting PEDOT:PSS-decorated LVO (P-LVO) half-cell. The charge/discharge curves between 0.2 and 3.0 V (vs. Li+/Li) indicate that the P-LVO electrode displays a capacity of 191.9 mAh/g at 8 C, while the LVO only delivers a capacity of 111.3 mAh/g at the same current density. To evaluate the practical application of P-LVO, lithium-ion capacitors (LICs) are constructed with P-LVO composite as the negative electrode and active carbon (AC) as the positive electrode. The P-LVO//AC LIC demonstrates an energy density of 107.0 Wh/kg at a power density of 125 W/kg, along with superior cycling stability and 97.4% retention after 2000 cycles. These results highlight the great potential of P-LVO for energy storage applications.

The inventory level has a significant influence on the cost of process scheduling. The stochastic cutting stock problem (SCSP) is a complicated inventory-level scheduling problem due to the existence of random variables. In this study, we applied a model-free on-policy reinforcement learning (RL) approach based on a well-known RL method, called the Advantage Actor-Critic, to solve a SCSP example. To achieve the two goals of our RL model, namely, avoiding violating the constraints and minimizing cost, we proposed a two-stage discount factor algorithm to balance these goals during different training stages and adopted the game concept of an episode ending when an action violates any constraint. Experimental results demonstrate that our proposed method obtains solutions with low costs and is good at continuously generating actions that satisfy the constraints. Additionally, the two-stage discount factor algorithm trained the model faster while maintaining a good balance between the two aforementioned goals.

Large-scale solid oxide fuel cell (SOFC) systems create operational challenges owing to their size and scale. An SOFC dynamic digital twin model can help users understand the system status and determine the operating action. In this study, a dynamic digital twin of a 25 kW SOFC plant was presented as a simulator to help operators safely and stably determine the operating conditions for a real commercial SOFC plant. In the scale-up process, the data from the 1 kW SOFC pilot plant were used to analyze the reliability and electrode parameters and develop a digital twin model that matches actual conditions at the plant. The scale-up digital twin model is applied to a 25 kW SOFC system, in which the SOFC is connected with upstream reformers, downstream burners, and multiple heat exchangers. The 25 kW SOFC digital twin model was validated by steady-state data and applied to on-site operation prediction with very high accuracy. The digital twin was used as an operating reference in the field, and the results showed that it could effectively help operators determine operation strategies using simulations to safely and stably execute the process.

In this study, a computational fluid dynamics (CFD) model of a sieve plate scrubber was built to predict its particle-removal efficiency and predict the U-shaped curve of the particle-removal efficiency as particles became smaller. Due to the complexity of particle tracking, it takes considerable time to simulate the model by using a three-dimensional (3D) structure, which is not conducive to finding the appropriate setting of particle forces. Instead, we presented a dimension-reduction method to estimate the particle force setting by using a two-dimensional (2D) structure. The rationality of the dimension-reduction method was validated by the consistency in froth density for both 2D and 3D models at various air-inlet velocities. Furthermore, the result of the particle forces setting showed that besides the drag force, other forces, such as the lift force, pressure-gradient force, gravity force, and virtual mass force, should be employed in the CFD model to predict the particle-removal efficiency of the sieve plate scrubber. The prediction results of the 2D model remarkably matches the particle-removal efficiency results of experimental data from the literature for various gas velocities and particle sizes. In addition, the model predicts the U-shaped curve of the particle-removal efficiency for the particle-diameter range from 0.1 to 1.0 μm. Furthermore, a 3D model with the setting of the particle forces as in the 2D model was used to validate the consistency between the 2D and 3D models. The result showed that the particle-removal efficiency of the 3D model was considerably close to the prediction results of the 2D model.

The mechanism of classification of the RNNs was revealed in this study. The benchmark of the Tennessee Eastman process was used to demonstrate the performance of the RNN-based fault diagnosis model. This study investigated the fault diagnosis performance of the entire and early detections for complicated chemical processes using the time-series recurrent neural networks (RNNs). The investigation included various layers and neuron nodes in RNNs using lean and rich training datasets and compared these RNNs with the artificial neural networks (ANNs). The results showed that the RNNs had better classification accuracies than the ANNs regardless of using lean or rich training datasets. The general classification mechanism was a priori classification that used normal operating data as the center so that it was incapable of separating the fault types having similar features of the normal operating data. RNNs drove the normal operating data out of the center and created the even spatial distribution of fault types, leading that RNNs were effective in classifying the fault types with subtle features when there is sufficient data.

Soft sensors attempt to predict the key quality variables that are infrequently available using the sensor and manipulated variables that are readily available. Since only limited amount of labeled data are available, there is always the concern whether the underlying physics were captured so that the model can be reasonably extrapolated. A sequence-to-sequence model in the form of a nonlinear state-observer/encoder and predictor/decoder was proposed. The observer can be trained using a large amount of unlabeled data, but in a supervised manner in which the process dynamics is tracked. The encoder output and manipulated variables are used to train the quality predictor. The model is applied to the product impurity predictions of an industrial column. Results show that good predictions and excellent consistency in the sign of estimated gains can be achieved even with limited amount of data. These findings indicated that the proposed sequence-to-sequence data-driven approach is able to capture the underlying physics of the process.

In the development of soft sensors for industrial processes, the availability of data for data-driven modeling is usually limited, which led to overfitting and lack of interpretability when conventional deep learning models were used. In this study, the proposed soft sensor development methodology combining first-principle simulations and transfer learning was used to address these problems. Source-domain models were obtained using a large amount of data generated by dynamic simulations. They were then fine-tuned by a limited amount of real plant data to improve their prediction accuracies on the target domain and guaranteed the models with correct domain knowledge. An industrial C4 separation column operating at a refining unit was used as an example to illustrate the effectiveness of this approach. Results showed that fine-tuned networks could obtain better accuracy and improved interpretability compared to a simple feedforward network with or without regularization, especially when the amount of actual data available was small. For some secondary effects, such as interaction gain, its interpretability is mainly based on the interpretability of the corresponding source models.

Guidelines for optimal operation strategy of cooling water temperature are presented to find the maximum net electrical energy of a cogeneration system for turbine generators (TGs) and cooling towers (CTs). In this study, a real cogeneration plant in Taiwan, consisting of variable-frequency drive fan-based CTs and a generator, was chosen as a case study. Statistical linear models were used to build TGs and CTs. A multi-linear model was established to approach a nonlinear system in the cooling tower using a physically meaningful index, the cooling capability index, as an indicator of data clustering. The results showed that the cooling capability index was a good clustering index for the CTs. The cooling tower multi-linear model regressed by the clustering data performed well in predicting the cooling water outlet temperature. The simulated case verified that the online operating guidelines were consistent with real plant data and that it has the ability to find the optimal cooling water temperature. Two cases of plant test prove that the online operating guidelines can assist operators in operating cooling water temperature to the maximum net electrical energy.

Computational fluid dynamics (CFD) simulations of countercurrent gas–liquid flow in random packings of Raschig rings were carried out in this study. This model used gravity simulation to construct the random packing structure, and a volume expansion-recovery method to improve the meshing quality. A simple feedback control scheme was applied to control the gas inlet flow rate so that pressure drop can be estimated. The generated characteristics of the packing structure such as the number of packing elements, dry surface area and porosity were found to be close to experiments. CFD predictions of hydrodynamics properties are also validated by the experimental data using a small column section. The results indicate that our CFD model was able to capture the essential hydrodynamics behavior of the gas–liquid countercurrent flow. Hence, the approach presented in this study can be used as a basis for studying the effect of detailed packing-geometry design on hydrodynamic and mass transfer characteristics.

Data-driven operation monitoring and optimization of chemical plants can be performed using Digital Twin, along with intelligent algorithms. This study proposed a sequence-to-sequence rolling training algorithm to overcome the challenge of rolling predictions. The data were generated through dynamic simulation of the vapor-recompression C3 process using Aspen Plus. Studies showed that StS with rolling training could better fit the real data than the StS model. Moreover, StS with rolling training was able to present efficient long-term predictions as Digital Twin.

Background We present a dynamic model for designing a rotating packed bed (RPB) demonstrating volatile organic compound absorptions with circulation. The developed RPB model can be applied for industrial process simulations and scale-up. Methods The industrial applicability of the model was validated using in-house steady-state and dynamic RPB experimental data. The absorption of acetic acid (AA) in gas stream by an RPB, evaluated using reverse osmosis (RO) water and 1 wt.% NaOH aqueous solution, established the model. The operational variables’ effects, such as rotational speed and gas to liquid ratio, on the AA absorption efficiency are discussed. Significant findings The height of transfer unit (HTU) of the RPB was less than 3 cm, indicating the superior mass-transfer efficiency of the RPB than that of conventional packed bed column. The RPB model could predict the HTU in a steady-state well for the AA absorption, and the average absolute deviations (AAD) of HTU for the absorbent using RO water and 1 wt.% NaOH solution were 19.2% and 15.2%, respectively. Furthermore, the RPB model accurately predicted the AA absorption in circulation well for 1 wt.% NaOH with initial AA loadings of 1.7, 1.9, and 2.0 wt.% and AADs of 3.5%, 1.7%, and 1.9%, respectively.

Recurrent neural networks (RNNs), particularly those with gated units, such as long short-term memory (LSTM) and gated recurrent unit (GRU), have demonstrated clear superiority in sequence modeling. In this study, the fault diagnosis performance and classification mechanisms of basic LSTM and GRU were investigated to provide comparative information about suitable fault diagnosis models in chemical processes. Visualization techniques were used to interpret the behavior of LSTM and GRU when performing fault diagnosis in the Tennessee Eastman process (TEP). The analysis indicated that input and output gates were the two main gates of the LSTM model to filter information, while the forget gate of LSTM might be insignificant because it was usually opened to allow information to pass through. Therefore, the GRU model separated the faults better, especially Fault 15, and it provided more promising fault diagnosis performance compared to the LSTM model. The diagnosis accuracy for Fault 15 increased from 63%, while using the LSTM model, to 76% while using the GRU model. The simulation results of the TEP indicated that the GRU neural network in this study was superior to the LSTM neural network.

A monoethanolamine (MEA) aerosol growth model was developed to quantify the aerosol growth factor in an amine-based CO2 capture absorber that considers the gas-liquid interactions, and it is empirically validated by measuring the aerosol particle size and concentration. The aerosol growth model, using sucrose as the aerosol nuclei instead of sulfuric acid to prevent the corrosion of the test equipment, accurately predicted that the outlet aerosol size increased to the same level regardless of the sucrose concentration. It also found that particle concentration was the primary factor affecting aerosol growth and amine emissions. We found an inverse relationship between aerosol particle concentration and the aerosol size, while the MEA emissions were proportional to particle concentration.

In this study, a monoethanolamine aerosol growth model was developed to investigate the aerosol growth factor. Interactions among the internal conditions in an absorber were considered in this aerosol model. Additionally, an experiment was conducted to measure aerosol particle size, for collecting in-house validation data. Sucrose was used as the aerosol nuclei instead of sulfuric acid to prevent the corrosion of equipment used in the experiment. Experimental results showed that the outlet aerosol sizes increased to the same size regardless of the sucrose concentrations. The aerosol growth model was validated using the in-house experimental data. The aerosol growth model efficiently predicted the aerosol size. For investigating aerosol growth effects, particle number concentration was determined to be the primary factor affecting aerosol growth and amine emissions. When the particle number concentration increased, the aerosol size decreased, whereas the MEA emission increased.

This work presents a novel artificial data assisted machine learning modelling approach to guarantee the correct physical gain signs between manipulated variables (MV)/controlled variable (CV) in a multi-input-single-output (MISO) model predictive control system (MPC). These industrial systems, such as polymerization reacting systems, are basically controlled based on the operators’ experiences due the lack of the trustworthy models. The purpose of this work is to provide an implementable machine learning model for MPC purposed by the aid of artificial data. This approach is shown valid through a numerical problem, real plant test data and plant test.

this study presented a competitive electrochemical mechanism of the hydrogen and carbon monoxide to describe the real electrochemical reactions in SOFC. The SOFC models employed the competitive electrochemical mechanism was validated with a variety of fuels as the feedstocks, such as methane, mixing gas of H2 and CO, blast furnace gas (BFG), and coke oven gas (COG), via pilot-scale SOFC modules of 1 kW and 50 kW. The models were implemented on Aspen Custom Modeler platform to simulate steady-state results. Further, the SOFC of the 50 kW pilot simulation considering the whole pilot process including pre-reformers, heat exchangers, and fuel recycling. The results showed the SOFC employed the competitive electrochemical mechanism has the better performances of the power generation and voltage predictions regardless of the what fuel used in SOFC of the 1 kW scale, compared to the SOFC only employed hydrogen electrochemical mechanism. The mean voltage error of the SOFC with the competitive electrochemical mechanism using various fuels was 1.52%. the minimum error was 0.73% using COG as the feedstock. The validation of the SOFC model of the 50 kW scale showed that only a 5% error with the experiment in steady-state and 6.91% error in transition.

In this paper, a universal probability-based method is proposed for health stage division based on the feature exacted from different rolling element bearings. A new monitoring feature called Amplitude Difference of Strong and Carpet Impulse (ADSCI) is adopted to obtain the degradation information of bearings. Specifically, the high frequency information is utilized to get the degradation status, and a new impulse extractor method is developed to quantify the degree of the degradation. The Exponential Weibull distribution (EWD) is introduced to describe the probability distribution of the ADSCI feature in healthy stage and degradation stage, respectively. Then, a method named Likelihood Probability Comparison (LPC) is put forward to online detect the degradation point which is located between the healthy and degradation stage. For the LPC method, the EWD is utilized to calculate two likelihood probabilities regarding the healthy and degradation stage, respectively. The comparison result of the two likelihood probabilities is used to detect the degradation point. The performance of the proposed method is evaluated on the PRONOSTIA platform, and the comparison experiment with other features is performed. The result shows that the proposed ADSCI feature behaves stable for different bearings and the proposed LPC method is universal and effective for health monitoring of different bearings.

This study has developed a multi-layer model to estimate concentration gradients inside aerosol and aerosol growth. The results show that piperazine (PZ) aerosol grows from 0.1 um to 3.2 μm in the absorber and up to 9.6 μm in the water wash with 10 7 particles/cm 3 . Aerosol can be assumed to be well-mixed in the absorber and water wash. The PZ gradient in the aerosol is negligible because of the small aerosol diameter. Larger particle number concentration results in smaller drops. Aerosol growth is controlled by PZ mass transfer in the absorber and by water mass transfer in the water wash. More volatile amine should provide faster growth in the absorber.

Amine aerosol emissions in carbon capture by amine scrubbing result in solvent loss and environmental impact. Management of the amine aerosol will require understanding the aerosol growth mechanism and quantitative modeling of aerosol growth in the absorber and water wash. An aerosol model was developed by the authors in a separate paper to quantify the aerosol growth inside the absorber and the water wash, and was validated at bench-scale absorber conditions. This work simulates amine aerosol growth at the conditions of the National Carbon Capture Center (NCCC), and will provide guidance for test plans for the April 2017 NCCC pilot plant campaign. The rate-based absorber was modeled using Independence in Aspen Plus ® RateSep TM with rigorous thermodynamics and kinetics. Aerosol growth was modeled using gPROMS ® ModelBuilder. The solvent used was 5 molal (m) piperazine (PZ). Multiple process configurations were evaluated, including intercoolers, dry bed, and two-stage water wash. At NCCC conditions, with particle number concentration of 10 7 particle/cm 3 , aerosol can grow from 0.1 μm to 5 μm in the absorber and up to 12 μm in the water wash. Absorber and water wash operating conditions will have a significant effect on aerosol growth. Amine aerosol growth inside the absorber is driven by amine-limited diffusion, and aerosol is practically at equilibrium with bulk gas CO 2 , water, and temperature. Higher temperature and higher CO 2 driving force result in higher PZ driving force, resulting in faster aerosol growth. In a dry bed, aerosol rapidly approach equilibrium with the bulk gas. In the water wash, aerosol tends to approach equilibrium with bulk gas water activity, and grows rapidly by picking up water. Additional packing results in longer residence time, which allows further aerosol growth. With excessive absorber packing, aerosol growth is inhibited by the mass transfer pinch.

In this study, both packed-bed (PB) and rotating PB (RPB) absorbers with a packed-bed stripper were used in a pilot plant for the removal of carbon dioxide (CO2) with 30 wt% monoethanolamine (MEA) as the solvent. The flue gas feed is the combustion product of the blast furnace gas from a steel mill containing approximately 30% of CO2. Both PB and RPB exhibit the same performance for the capture of CO2 with the same amount of energy consumption, but the RPB has a volume of approximately one-third that of PB absorber. The results obtained from experiments using ∼20 wt% MEA showed good agreement with those obtained from simulation using an Aspen rate-based model for PB and an Aspen custom modeler for RPB. These results demonstrated the feasibility of using a RPB for decreasing the size of absorber under relevant process conditions for capturing CO2.

In this paper, a machine learning method is proposed to extract molecular features as floating-point numbers in a high dimensional space from the language-like description Simplified Molecular Input Line Entry Specification (SMILES). Principle component analysis showed that this method can successfully classify alkanes and alcohols and also the chain lengths of the molecular by their location in a three-dimensional feature space. A neural network model is build using the location of a compound in this high dimensional space as input to predict the “sigma-profile”, the charge distribution of the molecule near a perfect infinite conductor, which is calculated by quantum mechanics. The sigma-profile can be used in the COSMOSAC model for predicting thermodynamic properties such as activity coefficient. Preliminary results showed that an accurate neural work model with generalization ability can be developed.

Computer-aided molecular design requires the ability to predict different molecular properties of interesting from using molecular structure. Traditional quantitative structural property relations were developed by extracting molecular features for predicting various properties. Hence domains of molecular features are different for predictions of different properties. In this work, the concept of a universal translator was used to develop a universal digital chemical space by translating and projecting the chemical representation SMILES to a high-dimensional space that can be collapsed into different molecular fingerprints. We demonstrated different kinds of pure component properties, such as electrical and thermodynamic properties can be predicted by a simple input of molecular structure, SMILES. This method eliminates the need to manually extract different molecular features for predicting different properties. The ability of model to predict sigma profiles also pave the way of prediction phase equilibria of mixtures using molecular structure only.

In this work, a model for the absorption of carbon dioxide using 3 wt% dilute aqueous ammonia solution in rotating packed bed and packed bed was developed and validated using experimental data. The model was then employed to design absorbers for different throughput rates, ranging from 0.01 to 1000 ton-CO 2 -capture/day. The solvent used is 3 wt% ammonia solution operating in the lean-to-rich loading range of 0.2–0.4. The results were compared to those obtained with a similar model using 30 wt% monoethanolamine solution operating in the lean-to-rich loading range of 0.35–0.49. It was found that the apparent intensification effect of the rotating packed bed is more significant for ammonia than for monoethanolamine solution. This is explained by the fact that the operating loading of the monoethanolamine solution is closed to saturation and the average reaction rate is low at this end. Improving mass transfer will increase the amount of CO 2 transferred from gas to liquid resulting in an accumulation of free CO 2 , thus reducing the driving force. For the dilute ammonia process, there is still sufficient reaction rate to covert the free CO 2 because the operating lean-to-rich loading is far from saturation.

This study compares the carbon dioxide capture efficiency of packed bed (PB) column and rotating packed bed (RPB) absorber by mono-ethanolamine (MEA) solution. Carbon dioxide absorption experiments were carried out in laboratory scale packed bed column and rotating packed bed at various lean carbon dioxide loadings of the MEA solution They were modelled successfully by an ASPEN Plus advanced customer model (ACM). Using these models, we found that substantial savings of packing volume can be achieved using a series arrangement of RPB and PB for the CO2 capture process using 30wt% MEA solution.

A robust parameter estimation procedure was proposed to obtain accurate, confident parameters for the atmospheric residue desulfurization (ARDS) model in this work. A comprehensive ARDS model includes too many parameters such as deactivation and loading parameters. However, past studies have used data fitting in the early phases of reactor system life testing to obtain the parameters. This procedure was found to be too sensitive to the initial estimates used in fitting. In order to use the ARDS model for designing reactor systems and evaluating operating strategies, a robust parameter estimation procedure must be available. In this work, we demonstrate that the kinetic and deactivation parameters should be obtained from kinetic tests and accelerated test data, respectively for individual catalysts. Only loading parameters such as contact efficiencies should be obtained using early life test data. The procedure was validated with long-term experimental data. The results showed that the procedure could be a...

This study presented computational fluid dynamics (CFD) and experimental validation of hydrodynamic behavior of helical rings random packing. Random packings have been widely used in chemical industries in absorbers, strippers etc. Up till now, development of novel random packing structure has essentially been empirical. Even with the increased computing power, CFD simulations of random packings are hard to find in the literature. The random nature of the packing structures, and the stacked geometry make acceptable grid generation and convergence very difficult because the structure of random packing is complicated when large amounts of it are stacked in a column. In this work, a CFD model was first time developed to simulate countercurrent gas-liquid flow in random packings formed by helical rings. Gravity simulation was used to generate stacking structures. A simple feedback control scheme was applied to control the gas inlet flow rate so that a particular pressure. Multiphase model was employed to compute the gas and liquid interaction in which the surface tension and wall contact angle were found as key factors. The predictions of the CFD model were validated with a lab-scale packed-bed absorber. It was found that the helical structure did increase the interfacial area, liquid hold-up when compared to Raschig rings, and such predictions can be validated by our in-house experiment. The model also showed that helical rings will have lower pressure drop and can sustain a larger liquid-gas ratios compared to Raschig rings. In summary, our study found that CFD simulations can obtain reasonable predictions of hydrodynamic behaviour of packings, and the inclusion of microstructures into a packing element will improve its hydrodynamic properties. These results showed that CFD can be used as a basis for rational packing design.

CDQ (coke dry quenching) is a widely used method for recovering waste heat in the steel industry. We have developed a novel, data driven modeling approach and model based control for a CDQ unit to increase steam generation in a cogeneration system. First, the correlation between steam generation and TCGB (the temperature of circulation gas entering the associated boiler) was confirmed. Subsequently, a nonlinear variable selection method was employed to build models of TCGB and the carbon monoxide concentration of the circulation gas. The models obtained were implemented to achieve MPC (model predictive control) for regulating the supplementary gas to maximize steam generation in an existing steelmaking plant. Upon comparison of the original process and the proposed modified operation, the effectiveness of the implementation of MPC was justified. The results showed that steam generation was increased by 7%. In our approach, the large amount of available operational data stored electronically was used to establish the models. Modification of the established system is not required. Taking into account that no capital investment is required, the process improvement is remarkable in terms of its return on investment.

This study provided a method to simulate flow field of an irregular arrangement of random packing in an absorber using commercial software, Fluent®. Volume of fluid model was employed in this study to deal with gas and liquid phases contacting issue. Volume of fluid model has validated that it is able to simulate gas and liquid interface. 10 flow fields of small-scale absorbers in which several raschig rings were stacked randomly, were established to evaluate average values of gas-liquid contact area and liquid holdup. The results showed that both of values are increased as liquid velocity increases. The simulated gas-liquid contact area has the same trend with the correlations provided from literature. The simulated liquid holdup completely matches the liquid holdup correlation for traditional absorbers.

This study presented an aerosol growth model which added the effect of the gas phase amine concentration to investigate aerosol growth and amine aerosol emission for MEA-CO 2 capture process. The results showed that aerosol grew to 8 μm with 10 7 #/cm 3 of the particle number concentration for a CO 2 capture pilot plant. The model also showed that the temperature of aerosol was very fast to reach equilibrium with the temperature of gas. The aerosol collected over 10 3 ppm of amine to emission into the atmosphere.

This study provided a method to simulate flow field of an irregular arrangement of random packing in an absorber using commercial software, Fluent®. Volume of fluid model was employed in this study to solve the gas and liquid phases’ interaction. Volume of fluid model has validated that it is able to simulate gas and liquid interface. 10 flow fields of small-scale absorbers in which several raschig rings were stacked randomly, were established to evaluate average values of the gas-liquid contact area and the liquid holdup. The results showed that the gas-liquid contact area and the liquid holdup are increased as liquid velocity increases. The simulated gas-liquid contact area has the same trend with the correlations provided from literature. The simulated liquid holdup completely matches the liquid holdup correlation for traditional absorbers.

This paper developed a new variable selection method for soft sensor applications using the nonnegative garrote (NNG) and artificial neural network (ANN). The proposed method employs the ANN to generate a well-trained network, and then uses the NNG to conduct the accurate shrinkage of input weights of the ANN. This paper took Bayesian information criterion as the model evaluation criterion, and the optimal garrote parameter s was determined by v -fold cross-validation. The performance of the proposed algorithm was compared to existing state-of-art variable selection methods. Two artificial dataset examples and a real industrial application for air separation process were applied to demonstrate the performance of the methods. The experimental results showed that the proposed method presented better model accuracy with fewer variables selected, compared to other state-of-art methods.

This study presented a model of the absorption of carbon dioxide by mono-ethanolamine solution in a rotating packed bed absorber using two-film theory for mass-transfer. It was found that temperature bulge can also be found in rotating packed bed if large amount of carbon dioxide was removed. Proper accounting of changes in liquid holdup is critical to accurate prediction of changes in capture efficiencies with gas and liquid flow rate. Moreover, use of a modified expression for the apparent kinetic constant in the enhancement factor is necessary to predict experiment results using high amine concentrations. Since the model have been comprehensively validated using data provided by various authors in the literature with various gas–liquid flow rates, rotator speeds, and solvent conditions, it could be used as a tool for scale up studies.

The logistic growth model combined with the Luedeking-Piret equation was adopted in this study to model the batch production of CoQ(10) in the cultivation of Rhodobacter sphaeroides. The simulation results indicated that CoQ(10) production was a primary metabolite. As being a primary metabolite, a longer cell growing stage would tend to accumulate more biomass and lead to a higher CoQ(10) concentration being produced. In this context, a fed-batch operation by molasses feeding was performed to increase the biomass and subsequent CoQ(10) production. Three different molasses feeding strategies were operated in this study. Results suggested that the fed-batch operation with molasses controlled at 10 + or - 1 g/l could increase the cell mass and CoQ(10) concentration to reach their maximum values of 18.6 g/l and 83.8 mg/l, respectively, nearly 2.2 times and 1.9 times their respective values obtained in the batch cultivation.

This paper developed a new variable selection method for soft sensor applications using the nonnegative garrote (NNG) and artificial neural network (ANN). The proposed method employs the ANN to generate a well-trained network, and then uses the NNG to conduct the accurate shrinkage of input weights of the ANN. This paper took Bayesian information criterion as the model evaluation criterion, and the optimal garrote parameter s was determined by v -fold cross-validation. The performance of the proposed algorithm was compared to existing state-of-art variable selection methods. A real industrial application for air separation process were applied to demonstrate the performance of the methods. The experimental results showed that the proposed method presented better model accuracy with fewer variables selected, compared to other state- of-art methods.

This study presented a model of the absorption of carbon dioxide by mono-ethanolamine solution in a rotating packed bed absorber using two-film theory for mass-transfer. It was found that temperature bulge can also be found in rotating packed bed if large amount of carbon dioxide was removed. Proper accounting of changes in liquid holdup is critical to accurate prediction of changes in capture efficiencies with gas and liquid flow rate. Moreover, use of a termolecular model for enhancement factor is necessary to predict experiment results using high amine concentrations. Since the model have been comprehensively validated using data provided by various authors in the literature with various gas-liquid flow rates, rotator speeds, and solvent conditions, it could be used as a tool for scale up studies.

The plant input variables are transformed into the derived variables, which are physically and statistically meaningful for modeling. Then, this study proposed an adaptive modeling approach by employing nonnegative garrote variable selection and auto-regression integrated moving average correction. The high accuracy of the proposed modeling approaches makes the implementation of the model-based control of ratio of air to fuel for improving boiler's efficiency readily practicable.

Based on the batch results, we constructed a simplified simultaneous saccharification and fermentation (SSF) model for the simulation of lactic acid production directly from unhydrolyzed potato starch using Lactobacillus amylophilus. The results of batch operation at different initial starch concentrations (20, 40 and 60 g/l) indicated that a higher initial starch concentration would lead to a slightly lower productivity, but would largely decrease the yield. Among that, the batch with 20 g/l of initial starch had the maximum productivity and the maximum yield, which would be 0.31 g/(l h) and 98% (g/g), respectively. In view of increasing the productivity and the final lactic acid concentration, a starch-controlled fed-batch operation with 20 g/l of initial starch was performed. It showed the fed-batch operation with starch controlled at 8 ± 1 g/l by adjusting the starch-feeding rate led to the maximum productivity of 0.75 g/(l h) and the yield of 69%.

This study presented experimental data of packed bed and rotating packed bed adsorption of carbon dioxide by dilute aqueous ammonia solutions. The heights of transfer units of packed bed were found to be 0.35 to 1.96 m; while the heights of a transfer unit of the rotating packed bed were in the range 0.08 to 0.40 m. Predictive models of the two processes were developed (model details and implications discussed elsewhere) based on the same set of thermodynamic and kinetic data for carbon dioxide into dilute aqueous ammonia solutions and existing mass transfer correlations for packed bed and rotating packed bed. The average absolute deviations between model predictions and experiments were found to be 9.4% for rotating packed bed data and 12% pack bed data.