Output-relevant Variational autoencoder for Just-in-time soft sensor modeling with missing data

Main challenges for developing data-based models lie in the existence of high-dimensional and possibly missing observations that exist in stored data from industry process. Variational autoencoder (VAE) as one of the deep learning methods has been applied for extracting useful information or features from high-dimensional dataset. Considering that existing VAE is unsupervised, an output-relevant VAE is proposed for extracting output-relevant features in this work. By using correlation between process variables, different weight is correspondingly assigned to each input variable. With symmetric Kullback–Leibler (SKL) divergence, the similarity is evaluated between the stored samples and a query sample. According to the values of the SKL divergence, data relevant for modeling are selected. Subsequently, Gaussian process regression (GPR) is utilized to establish a model between the input and the corresponding output at the query sample. In addition, owing to the common existence of missing data in output data set, the parameters and missing data in the GPR are estimated simultaneously. A practical debutanizer industrial process is utilized to illustrate the effectiveness of the proposed method.     

基于GFU和分层LSTM的组群行为识别研究方法

提出一种以"关键人物"为核心，使用门控融合单元（GFU，Gated Fusion Unit）进行特征融合的组群行为识别框架，旨在解决两个问题：①组群行为信息冗余，重点关注关键人物行为特征，忽略无关人员对组群行为的影响；②组群内部交互行为复杂，使用GFU有效融合以关键人物为核心的交互特征，再通过LSTM时序建模成为表征能力更强的组群特征.最终，通过softmax分类器进行组群行为类别分类.该算法在排球数据集上取得了86.7%的平均识别率.     

Neural network-based hybrid models developed for free radical polymerization of styrene

In the present work, the free radical polymerization of styrene is modeled by considering the phenomenology of the process (a simplified model, which does not include the diffusional effects, gel, and glass effects) in combination with an empirical model represented by an artificial neural network. Differential evolution (DE) algorithm, belonging to the class of evolutionary algorithms, is applied for developing the neural models in optimal forms. For improving the results—predicted conversion and molecular weights as function of time, temperature, and initiator concentration—different combinations between phenomenological model and neural network are tested; also, individual and stacked neural networks have been developed for the polymerization process. This methodology based on hybrid models, including neural networks aggregated in stacks, provides accurate results.     

An integrated dimensionality reduction and surrogate optimization approach for plant-wide chemical process operation

With liquefied natural gas becoming increasingly prevalent as a flexible source of energy, the design and optimization of industrial refrigeration cycles becomes even more important. In this article, we propose an integrated surrogate modeling and optimization framework to model and optimize the complex CryoMan Cascade refrigeration cycle. Dimensionality reduction techniques are used to reduce the large number of process decision variables which are subsequently supplied to an array of Gaussian processes, modeling both the process objective as well as feasibility constraints. Through iterative resampling of the rigorous model, this data-driven surrogate is continually refined and subsequently optimized. This approach was not only able to improve on the results of directly optimizing the process flow sheet but also located the set of optimal operating conditions in only 2 h as opposed to the original 3 weeks, facilitating its use in the operational optimization and enhanced process design of large-scale industrial chemical systems.     

Frequency domain analysis of the mirror-modified filtered-x least mean squares algorithm with low ambient noise

The mirror-modified filtered-x least mean squares (MMFxLMS ) algorithm is a variation of the FxLMS algorithm with online secondary path modeling that cannot diverge due to secondary path modeling errors. However, problems may occur when the ambient noise is not limited due to insufficient modeling power. This work shows that under a frequency domain analysis without ambient noise, the MMFxLMS algorithm is always stable, and expressions for the maximum residual noise level at any given time are obtained. It is also shown that, under the same context, convergence to the minimum residual noise is guaranteed. Still, convergence can be much slower for high secondary path modeling errors than that of the LMS or MFxLMS algorithms. Simulations confirm these results.     

Thermodynamic modeling of gas solubility in aqueous sodium chloride solution

An electrolyte Equation of State is presented by combining the Cubic Plus Association Equation of State,Mean Spherical Approximation and the Born equation.This new model uses experimental relative static permittivity,intend to predict well the activity coefficients of individual ions (ACI) and liquid densities of aqueous solutions.This new model is applied to model water + NaCl binary system and water + gas +NaCl ternary systems.The cation/anion-water interaction parameters of are obtained by fitting the exper-imental data of ACI,mean ionic activity coefficients (MIAC) and liquid densities of water + NaCl binary system.The cation/anion-gas interaction parameters are obtained by fitting the experimental data of gas solubilities in aqueous NaCl solutions.The modeling results show that this new model can correlate well with the phase equilibrium and volumetric properties.Without gas,predictions for ACI,MIAC,and liquid densities present relative average deviations of 1.3％,3.6％ and 1.4％ compared to experimental ref-erence values.For most gas-containing systems,predictions for gas solubilities present relative average deviations lower than 7.0％.Further,the contributions of ACI,and salting effects of NaCl on gases are ana-lyzed and discussed.     

Impact of Lake Michigan water level rise on complex bidirectional flow in the Chicago Area Waterway System (CAWS)

In the past decade, continuously rising water levels in Lake Michigan have been threatening lakefront areas, especially in metropolitan regions like the Greater Chicago area. This provides the motivation to analyze the impact that high lake levels have on the Chicago Area Waterway System (CAWS). As the only primary free connection between the CAWS and Lake Michigan, the Calumet Area waterway subsystem plays a key and unique role. In this work, a numerical model covering the Calumet subsystem and having Lake Michigan as a boundary condition, is set up, calibrated, and validated using limited field observations. It is found that the Calumet subsystem has become bidirectional, where both discharge and flow directions are controlled by lake levels. When lake levels are below −0.15 m (-0.5 ft, Chicago City Datum, CCD), the discharge in the Grand Calumet River is around zero, with water flowing along its east branch towards Indiana. When lake levels are above +0.46 m (+1.5 ft, CCD), the flow reverses direction and drains west into Illinois. In 2020, the mean lake-level was at +1.07 m (+3.5 ft, CCD), and the base discharge in the Grand Calumet River was approximately 8.5 m³/s (300 ft³/s). The higher Lake Michigan’s level is, the larger the discharge would be into Illinois. Potential impact of this extra discharge on Lake Michigan Diversion Accounting (LMDA) of the State of Illinois and flood management in the Chicago Sanitary and Ship Canal (CSSC), is analyzed; while the nature of the bidirectional flows is characterized with the intent of shedding light on this complex phenomenon.     

Thermodynamic modeling of hydrogen refueling for heavy-duty fuel cell buses and comparison with aggregated real data

The foreseen uptake of hydrogen mobility is a fundamental step towards the decarbonization of the transport sector. Under such premises, both refueling infrastructure and vehicles should be deployed together with improved refueling protocols. Several studies focus on refueling the light-duty vehicles with 10 kg_H2 up to 700 bar, however less known effort is reported for refueling heavy-duty vehicles with 30–40 kg_H2 at 350 bar. The present study illustrates the application of a lumped model to a fuel cell bus tank-to-tank refueling event, tailored upon the real data acquired in the 3Emotion Project. The evolution of the main refueling quantities, such as pressure, temperature, and mass flow, are predicted dynamically throughout the refueling process, as a function of the operating parameters, within the safety limits imposed by SAE J2601/2 technical standard. The results show to refuel the vehicle tank from half to full capacity with an Average Pressure Ramp Rate (APRR) equal to 0.03 MPa/s are needed about 10 min. Furthermore, it is found that the effect of varying the initial vehicle tank pressure is more significant than changing the ambient temperature on the refueling performances. In conclusion, the analysis of the effect of different APRR, from 0.03 to 0.1 MPa/s, indicate that is possible to safely reduce the duration of half-to-full refueling by 62% increasing the APRR value from 0.03 to 0.08 MPa/s.     

Hydrogen production by silica membrane reactor during dehydrogenation of methylcyclohexane: CFD analysis

A comprehensive computational fluid dynamic model has been developed using COMSOL Multiphysics 5.4 software to predict the behavior of a membrane reactor in dehydrogenation of methylcyclohexane for hydrogen production. A reliable reaction kinetic of dehydrogenation reaction and a permeation mechanism of hydrogen through silica membrane have been used in computational fluid dynamic modeling. For performance comparison, an equivalent traditional fixed bed reactor without hydrogen removal has been also modeled. After model validation, it has been used to evaluate the operating parameters effect on the performance of both the silica membrane reactor and the equivalent traditional reactor as well. The operating temperature ranged between 473 and 553 K, pressure between 1 and 2.5 bar, sweep factor from ?6.22 to 25 and feed flow rate from 1 to 5 × 10^?6 mol/s. The membrane reactor performed better than the equivalent traditional reactor, achieving as best result complete methylcyclohexane conversion and 96% hydrogen recovery.