Anhydrous ethanol is one of the biofuels produced today and it is a subset of renewable energy. It is considered to be an excellent alternative clean-burning fuel to gasoline. Anhydrous ethanol is ...commercially produced by either catalytic hydration of ethylene or fermentation of biomass. Any biological material that has sugar, starch or cellulose can be used as biomass for producing anhydrous ethanol. Since ethanol–water solution forms a minimum-boiling azeotrope of composition of 89.4mol% ethanol and 10.6mol% water at 78.2°C and standard atmospheric pressure, the dilute ethanol–water solutions produced by fermentation process can be continuously rectified to give at best solutions containing 89.4mol% ethanol at standard atmospheric pressure. Therefore, special process for removal of the remaining water is required for manufacture of anhydrous ethanol. Various processes for producing anhydrous ethanol have been used/suggested. These include: (i) chemical dehydration process, (ii) dehydration by vacuum distillation process, (iii) azeotropic distillation process, (iv) extractive distillation processes, (v) membrane processes, (vi) adsorption processes and (vii) diffusion distillation process. These processes of manufacturing anhydrous ethanol have been improved continuously due to the increasingly strict requirements for quantity and quality of this product. The literature available on these processes is reviewed. These processes are also compared on the basis of energy requirements.
This perspective paper features the process intensification (PI) application for advanced distillation-based processes. Starting with the historical background of generic PI, we subsequently narrow ...down the discussion to extractive distillation (ED), reactive distillation (RD), and hybrid reactive-extractive distillation (RED). We categorize the existing PI techniques onto internal and external intensification, where the former does not involve altering the distillation configuration while the latter does. Instead of deliberating the technical aspects, we explicitly highlight the contribution of PI applied to ED, RD, and RED towards societal impact covering energy, economic, environmental, control, and safety perspectives. The future perspectives of PI are discussed in the last section, covering the development of hybrid PI technologies, exploring the energy efficiency of different PI configurations, prioritizing PI beyond energy by considering some other sustainability aspects, and linking PI with the ever-increasing Industry 4.0 applications.
•Historical application of PI of conventional distillation.•Extension of PI to advanced distillation processes covering ED, RD, RED.•Categorization of internal and external PI techniques.•Contribution of PI to sustainability and societal impact.•Future perspectives of PI applied to ED, RD, and RED.
Online Subclass Knowledge Distillation Tzelepi, Maria; Passalis, Nikolaos; Tefas, Anastasios
Expert systems with applications,
11/2021, Volume:
181
Journal Article
Peer reviewed
•A novel distillation method aiming to reveal the subclass similarities is proposed.•The OSKD method derives the soft labels from the model itself, in an online manner.•The OSKD method is ...model-agnostic.•The experiments validate the effectiveness of the OSKD method.
Knowledge Distillation has been established as a highly promising approach for training compact and faster models by transferring knowledge from more heavyweight and powerful models, so as to satisfy the computation and storage requirements of deploying state-of-the-art deep neural models on embedded systems. However, conventional knowledge distillation requires multiple stages of training rendering it a computationally and memory demanding procedure. In this paper, a novel single-stage self knowledge distillation method is proposed, namely Online Subclass Knowledge Distillation (OSKD), that aims at revealing the similarities inside classes, improving the performance of any deep neural model in an online manner. Hence, as opposed to existing online distillation methods, we are able to acquire further knowledge from the model itself, without building multiple identical models or using multiple models to teach each other, rendering the OSKD approach more effective. The experimental evaluation on five datasets indicates that the proposed method enhances the classification performance, while comparison results against existing online distillation methods validate the superiority of the proposed method.
The current state-of-the-art commercial styrene distillation schemes, featured by conventional distillation columns to purify styrene, are introduced with energy, exergy and economic analyses. ...Amongst all the procedures the separation of ethylbenzene/styrene, the critical close-boiling system, accounts for ∼65% of the total energy requirement. To improve the energetic efficiency, double-effect distillation (DED) and heat pump distillation (HPD) are suggested as competitive improvements on conventional distillation schemes (CDSs), which give birth to advanced distillation schemes (ADSs). In addition, sensitivity analysis is carried out to determine the optimal operational parameters of columns in styrene distillation process. Taking the CDSs as benchmark processes, the ADSs with DED and HPD can lower operating costs by up to 30% and 40%, respectively. The synergistic effect makes retrofit proposals' payback period very attractive, through considerably energy costs reduction and uttermost equipment reuse. In the view of total annualized cost (TAC), the ADSs can cut a corner of ∼35–40% from the CDSs. Specifically, the ADS using HPD slightly outperforms its DED counterpart in TAC comparison. Despite energetic or monetary advantage, the ADSs also show their environmental drawback of higher exergy losses than the CDSs.
•Energy efficient distillation technologies are applied on EB/SM column.•Innovative retrofit scenarios with attractive payback periods are proposed.•Significant reduction in energy consumption as well as total annualized cost.•Optimization of advanced distillation schemes through sensitivity analysis.
Multi‐feed, multi‐product distillation columns are ubiquitous in multicomponent distillation systems. The minimum reflux ratio of a distillation column is directly related to its energy consumption ...and capital cost. Thus, it is a key parameter for distillation systems design, operation, and comparison. In this series, we present the first accurate shortcut based algorithmic method to determine the minimum reflux condition for any general multi‐feed, multi‐product (MFMP) distillation column separating any ideal multicomponent mixture. The classic McCabe‐Thiele or Underwood method is a special case of this general approach. Compared with existing techniques, this method does not involve any rigorous tray‐by‐tray calculation, nor does it require guessing of key components. In this first part of the series, we present the mathematical model for a general MFMP column, derive constraints for feasible separation and minimum reflux condition, discuss their geometric interpretations, and present an illustrative example to demonstrate the effectiveness of our approach.
Drone object detection with real-time deployment is a research hotspot. On the one hand, the performance of tiny object detection is closely related to the ground detection capability of the drone ...platform. Existing methods are keen on designing complex networks to enhance the accuracy of tiny objects, which significantly increases computational costs. On the other hand, the limited drone hardware resources urgently require lightweight models for deployment. To address the dilemma of balancing the detection accuracy and computational efficiency, we propose a Regenerated-Decoupled-Responsive knowledge distillation (RDR-KD) framework specifically for drone scenes. First, we design the Regenerated Distillation and the Decoupled Distillation to fully transfer the tiny object feature information from the teacher model to the student model. Meanwhile, we devise the logit-based Responsive Distillation based on focal loss and EIoU to alleviate class imbalance. Finally, we conduct extensive experiments on the VisDrone2019 dataset. The experimental results demonstrate that the proposed RDR-KD framework improves AP and APS of the student model by 3.3% and 2.9% respectively, which outperforms other state-of-the-art distillation frameworks.
The upgrading of bio-oil with the existing distillation technologies is mainly hindered by polymerisation due to the reactive nature of bio-oil when it is heated up during distillation. We propose ...that the distillation of bio-oil at high pressure can significantly reduce the extent of polymerisation. This study aims to investigate the roles of process parameters, especially pressure, in the high-pressure reactive distillation of bio-oil produced from the pyrolysis of mallee woody biomass. The bio-oil distillation fraction yields and properties were evaluated to demonstrate the advantages of the distillation at elevated pressures over distillation at atmospheric pressure. Our results indicate that high-pressure distillation can achieve high distillate yields with reduced polymerisation because high pressure can retain water and other light components in the liquid phase to reduce the extent of polymerisation. Besides reducing the polymerisation, high pressure could also intensify the separation during the flash separation step. The distillate yield was around 90% when the distillation was carried out at 200 °C at a pressure higher than 20 barg.
•High-pressure distillation of bio-oil can reduce the polymerisation extent.•High-pressure distillation of bio-oil can achieve high distillate yields.•High pressure can retain water and other light components in the liquid phase.•Distillate yield is ~90% when distilling bio-oil at 200 °C and pressure ≥20 barg.
•A review of literature studies on extractive distillation is provided.•Both continuous and batch extractive distillation processes are considered.•Separation of azeotropic mixtures and low relative ...volatility mixtures are shown.•The choice of the entrainer is discussed along with the process feasibility.•Process operation and control are surveyed.
Extractive distillation processes enable the separation of non-ideal mixtures, including minimum or maximum boiling azeotropes and low relative volatility mixtures. Unlike azeotropic distillation, the entrainer fed at another location than the main mixture induces an extractive section within the column. A general feasibility criterion shows that intermediate and light entrainers and heterogeneous entrainers are suitable along common heavy entrainers. Entrainer selection rules rely upon selectivity ratios and residue curve map (rcm) topology including univolatility curves. For each type of entrainer, we define extractive separation classes that summarize feasibility regions, achievable products and entrainer – feed flow rate ratio limits. Case studies are listed as Supplementary materials. Depending on the separation class, a direct or an indirect split column configuration will allow to obtain a distillate product or a bottom product, which is usually a saddle point of rcm. Batch and continuous process operations differ mainly by the feasible ranges for the entrainer – feed flow rate ratio and reflux ratio. The batch process is feasible under total reflux and can orient the still path by changing the reflux policy. Optimisation of the extractive process must systematically consider the extractive column along with the entrainer regeneration column that requires energy and may limit the product purity in the extractive column through recycle. For the sake of reducing the energy cost and the total cost, pressure change can be beneficial as it affects volatility, or new process structures can be devised, namely heat integrated extractive distillation, extractive divided wall column or processes with preconcentrator.
Medical image segmentation is crucial for enhancing diagnostic accuracy through pixel labeling. State-of-the-art networks, despite their performance, have high computational demands, limiting ...real-time use on constrained devices. Lightweight networks face challenges in balancing detail processing with precision. Vision Transformer models, while promising, also have computational concerns. This study presents a novel method that merges Vision Transformer strengths with a unique knowledge distillation technique. A pivotal element of our approach is the Token Importance Ranking Distillation, which facilitates the meticulous transfer of top-k token importance rankings between a complex teacher model and a simplified student model, guided by a specialized ranking loss function. This method is essential for optimizing the student model to effectively emulate the teacher model’s ability to encapsulate vital semantic and spatial information. Additionally, we introduce an innovative methodology in structural texture knowledge, utilizing a Contourlet Decomposition Module (CDM), which enriches the models with nuanced texture representation, crucial for extracting directional features and capturing intricate global and local contexts in medical imaging. Complementing this, we deploy a unique multi-stage distillation strategy, the Space Channel Cascade Fusion (SCCF), to refine both spatial and channel information concurrently, mitigating redundancy and enhancing representational effectiveness in feature maps. Experimental results demonstrate the effectiveness of our approach in elevating the performance of student models while diminishing computational demands, thereby enabling efficient, real-time medical image segmentation on resource-constrained devices.