Sequential Iterative Algorithm Research Articles

Efficient Inverse Fractional Neural Network-Based Simultaneous Schemes for Nonlinear Engineering Applications

Finding all the roots of a nonlinear equation is an important and difficult task that arises naturally in numerous scientific and engineering applications. Sequential iterative algorithms frequently use a deflating strategy to compute all the roots of the nonlinear equation, as rounding errors have the potential to produce inaccurate results. On the other hand, simultaneous iterative parallel techniques require an accurate initial estimation of the roots to converge effectively. In this paper, we propose a new class of global neural network-based root-finding algorithms for locating real and complex polynomial roots, which exploits the ability of machine learning techniques to learn from data and make accurate predictions. The approximations computed by the neural network are used to initialize two efficient fractional Caputo-inverse simultaneous algorithms of convergence orders ς+2 and 2ς+4, respectively. The results of our numerical experiments on selected engineering applications show that the new inverse parallel fractional schemes have the potential to outperform other state-of-the-art nonlinear root-finding methods in terms of both accuracy and elapsed solution time.

Pilot-scale validation and novel design of reactive distillation with HY@SiC foam catalytic packing for cyclohexyl acetate production

For reactive distillation (RD) process limited by the separation efficiency of gas–liquid mass transfer, if the physical size of the alternating operation in reaction and separation can be further reduced, it would greatly promote the efficiency of the RD process. For this purpose, a structured HY@SiC catalytic packing was developed in which the gas–liquid mass transfer separation process and the catalytic reaction occurred simultaneously on its surface. The HY@SiC catalytic packing was used in a pilot-scale column for experimental study on production of cyclohexyl acetate (CA) by RD to verify the performance of the catalytic packing for the first time. Furthermore, a RD model based on the HY@SiC catalytic packing was established based on the reaction kinetics of HY zeolite catalyst obtained in this work. Finally, a novel RD process with double recycle stream for CA production was proposed, which will be optimized design with TAC minimize as the optimization objective based on sequential iterative algorithm.

Numerical scheme for estimating all roots of non-linear equations with applications

<abstract><p>The roots of non-linear equations are a major challenge in many scientific and professional fields. This problem has been approached in a number of ways, including use of the sequential Newton's method and the traditional Weierstrass simultaneous iterative scheme. To approximate all of the roots of a given nonlinear equation, sequential iterative algorithms must use a deflation strategy because rounding errors can produce inaccurate results. This study aims to develop an efficient numerical simultaneous scheme for approximating all nonlinear equations' roots of convergence order 12. The numerical outcomes of the considered engineering problems show that, in terms of accuracy, validations, error, computational CPU time, and residual error, recently developed simultaneous methods perform better than existing methods in the literature.</p></abstract>

Integrated synthesis and control of heat exchanger networks with dynamic flexibility consideration

• Dynamic flexible aspect is taken into account in early integration stage. • A sequential iterative algorithm for the problem solution is devised. • Stream matches, heat transfer areas and bypass fractions are optimized. • Results indicate considerable cost (−14.95%) and control performances (+61.83%). • Major advantages found for case studies with multiple disturbances. Flexible synthesis and control which can be both used to reduce the influences of disturbances on heat exchanger networks are often considered as two separate issues assisting network development at different stages. A proper integration of these separate yet connected tasks carries the promise of achieving superior designs. Currently, the study of integration of heat exchanger networks is still towards a sequential procedure consisting of initial flexible synthesis based on steady-state economic calculations followed by a controller. To overcome this challenge, based on the stage-wise superstructure, an optimization framework in this work is presented to address dynamic flexible synthesis and advanced control simultaneously for maximizing performance in face of disturbances. The framework is based on a sequential iterative procedure that decomposes the overall problem into two stages. The first stage is performed by the dynamic flexibility analysis where the design variables are chosen. In the second stage, the optimization variables are adjusted during the discrete time intervals on the realizations of the control actions and the variation ranges of outlet stream temperatures. The sequential iterative is to map the temperature regulations to the network configuration retrofits. The application to two case studies indicates that the proposed framework returns solutions which are considerably better all in terms of dynamic flexibility, economics and control performance than those published in literature. Compared to previous literature studies, the optimized solutions feature a total annual cost reduction up to 14.95%, a decrease in control action up to 48.58% and an increase in control performance up to 61.83%. Moreover, the application to two case studies indicates that allows solving a real-world problem with up to 26 hot streams and 29 cold streams (leading to models with a size of 523 binary variables and 1655 equations).

A Band Subset Selection Approach Based on Sparse Self-Representation and Band Grouping for Hyperspectral Image Classification

Band subset selection (BSS) is one of the ways to implement band selection (BS) for a hyperspectral image (HSI). Different from conventional BS methods, which select bands one by one, BSS selects a band subset each time and preserves the best one from the collection of the band subsets. This paper proposes a BSS method, called band grouping-based sparse self-representation BSS (BG-SSRBSS), for hyperspectral image classification. It formulates BS as a sparse self-representation (SSR) problem in which the entire bands can be represented by a set of informatively complementary bands. The BG-SSRBSS consists of two steps. To tackle the issue of selecting redundant bands, it first applies band grouping (BG) techniques to pre-group the entire bands to form multiple band groups, and then performs band group subset selection (BGSS) to find the optimal band group subset. The corresponding representative bands are taken as the BS result. To efficiently find the nearly global optimal subset among all possible band group subsets, sequential and successive iterative search algorithms are adopted. Land cover classification experiments conducted on three real HSI datasets show that BG-SSRBSS can improve classification accuracy by 4–20% compared to the existing BSS methods and requires less computation time.

Mechanism analysis and process optimization of acetone–methanol azeotrope separation using 1-ethyl-3-methylimidazolium acetate based mixed extractants

The development of high-performance extractants and intensification technologies is crucial for saving energy in the separation of azeotropes by extractive distillation. An effective method for separating azeotropes by extractive distillation using an organic solvent–ionic liquid mixed extractant is proposed. The selectivity and solubility of 96 ionic liquids were calculated based on the conductor-like screening model for real solvents, and the best extractant, 1-ethyl-3-methylimidazolium acetate ([EMIM][AC]), was obtained. The interaction energy, bond length, and independent gradient model based on the Hirshfeld partition between different components and azeotropes were calculated by applying quantum chemistry to analyze the structure–activity relationship between molecules. The results showed that dimethyl sulfoxide has a stronger interaction with methanol than with other organic extractants, [EMIM][AC] + dimethyl sulfoxide is more likely to form hydrogen bonds with methanol, the O atom in dimethyl sulfoxide mainly forms hydrogen bonds with the H atom in methanol, and the H atom in [EMIM][AC] mainly forms hydrogen bonds with the O atom of methanol. In addition, a process for separating acetone–methanol by extractive distillation using a mixed extractant was designed. The minimum total annual cost (TAC) was considered as the objective function, and a sequential iterative optimization algorithm was used to optimize the separation process and obtain the optimal operating parameters. The results showed that the TAC of the extractive distillation process of [EMIM][AC] + dimethyl sulfoxide mixed extractant was 15.00% lower than that of [EMIM][AC] + N,N-dimethylformamide and 42.10% lower than that of [EMIM][AC] + N-methylpyrrolidone. In the extractive distillation process, the pollutant gas emissions of [EMIM][AC] + dimethyl sulfoxide decreased by 43.29% compared with those of [EMIM][AC] + N-methylpyrrolidone and 7.87% compared with those of [EMM][AC] + N,N-dimethylformamide. Finally, an energy-saving process of vapor-recompression extractive distillation was explored. The results showed that the TAC of vapor-recompression-assisted extractive distillation was 3.98% lower than that of [EMIM][AC] + dimethyl sulfoxide and that the pollutant gas emissions were 20.28% less than those of [EMIM][AC] + dimethyl sulfoxide. This study on the structure–activity relationship of mixed extractants and process optimization provides theoretical guidance for the separation of azeotropes in the industry.

Enhanced transesterification reactive distillation for producing isopropanol: From kinetics, pilot-scale experiments, and process design to sustainability evaluation

The demand for isopropyl alcohol (IPA) has increased significantly due to its excellent disinfection and sterilization effect. The production of IPA by isopropyl acetate (IPAC) with methanol (MeOH) transesterification is limited by chemical equilibrium, and there are three binary azeotropic mixtures in the system, which makes the production process more complex and energy consumption higher. Thus, the reaction distillation (RD) process is designed and developed for producing IPA more effectively and economically. The KRD001 as the commercial catalyst is utilized to build the kinetic reaction model, which supplemented the primary data of the chemical industry. The Pilot-scale RD experiments for producing IPA are explored to verify the feasibility of RD and the reliability of the model and to ensure the establishment of the follow-up process. The hybrid distillation process (RDC-HDC) of RD combining pressure-swing distillation (PSD) is designed and optimized based on the pressure sensitivity analysis and sequential iterative optimization algorithm on the minimum TAC under the 2,800 tons annual output of the factory. The thermal coupling processes (RDWC-HDC and RDWC-RDC) are explored to find energy-saving schemes. The RDWC-HDC process can reduce 20.46% in energy consumption, save 25.65% in total annual cost (TAC), and reduce annually the 1834.73 tons and 9557.13 $ in CO2 emissions, 55.20 tons and 20.59 $ in SO2 emissions and 27.60 tons, and 20.56 $ in NOx emissions in contrast with the RDC-HDC process, respectively. The exergy destruction elaborates that the RDWC-HDC process is positive for energy saving. The RDWC-HDC process is the most economical and attractive for producing IPA through the energy, economy, environment, and exergy (4E) analyses.

Extractive distillation of ethyl acetate-ethanol-water ternary system using deep eutectic solvent (DES): A techno economic and energy assessment

The extractive distillation (ED) process has been established as a commercially viable process that can be used to break the azeotropic systems. The present study investigates the performance and feasibility of conventional solvent (glycerol) and choline chloride-glycerol based deep eutectic solvent (DES) as entrainers for the separation of ethyl acetate-ethanol-water ternary mixture through ED. A comprehensive approach is proposed with phase-equilibrium measurements and process simulation through Aspen Plus for different entrainer to feed ratio (E/F). Optimization of the process is based on performance indicators such as Total Annual Cost, CO2 emissions, and Specific Energy Consumption. A sequential iterative algorithm was used to find the optimum E/F in distillation trains. The performance analysis has shown a reduction of 45.89% in global TAC with DES as an entrainer. The combined entrainer requirement for the dehydration was 268.3 kmol/h and 118.86 kmol/h with glycerol and DES respectively. The process sees a 58.21% total energy saving with a 15% and 44.5% reduction in Specific Energy Consumption and CO2 emissions respectively as compared to glycerol as an entrainer. The study provides a comparison and proves that DES is an effective entrainer to break the azeotropic point as a low-cost and eco-friendly alternative to conventional solvents.

Economic effect of an efficient and environmentally friendly extractive distillation/pervaporation process on the separation of ternary azeotropes with different compositions

Exploring the process of cost-effective and efficient technology to separate ternary aqueous azeotropic systems has become essential research in chemistry and materials science. Traditional distillation technology and promising membrane technology occupy an irreplaceable position in wastewater treatment. However, single separation technology for the separation of azeotropic systems has disadvantages. To better meet the industrial practical needs of different azeotropic compositions in each process of rapeseed oil production, the separation performance and economic cost of single extractive distillation and an extractive distillation coupled pervaporation process for ternary azeotropic systems with different feed compositions were studied. Environmentally friendly and effective solvents were selected as extractants in the extractive distillation process from two aspects: selectivity and biological toxicity. Sequential iterative optimization algorithms were used to optimize the process with the ratio of water to the total amount of cyclohexane and isopropanol ranging from 1:9 to 9:1, considering cost as the evaluation index. The results showed that the cost of the processes increases significantly with increasing water content in the feed composition. Because the flux and selectivity of the membrane module were considerably affected by temperature when the water content was less than 0.4, the integrated process was more economical; otherwise, the extractive distillation process was more economical when the water content was greater than 0.4.

Process design and optimization of the efficient production of butyl acrylate by reactive azeotropic distillation/pervaporation using different feed ratios

Present industrial development has a great demand for butyl acrylate, an important basic chemical. Efficient synthesis and separation processes are very important for the clean manufacturing of butyl acrylate. In this work, two processes for the preparation of butyl acrylate by reactive distillation coupled with azeotropic distillation (RD-AD) or pervaporation (RD-PV) were proposed. By changing the ratio of reactants, the economic performance of the two processes under different specifications of butyl acrylate was analyzed, and the environmental impact of the two processes in the production process were analyzed. First, the process design using different purity specifications was completed focusing on the reaction kinetics. Then, based on the sequential iterative algorithm, process optimization was carried out with the minimum total annual cost (TAC) as the objective function, and the process with optimal operating conditions was obtained. Finally, TAC and the global warming potential (GWP), acidification potential (AP), abiotic depletion potential (ADP) and human toxic potential (HTP) were used to assess the economic and environmental performance of the proposed processes. The results showed that the RD-AD process had better economic performance, whereas the RD-PV process had better environmental performance. According to the results, the parameters that influenced the TAC for products with different purities were obtained. The TAC of the RD-AD process was 41.39% lower than that of the RD-PV process, and GWP, AP, and ADP of the RD-PV process were respectively 12.67%, 13.68%, and 13.07% lower than those of the RD-AD process.

Process design and intensification for the clean separation of ternary multi-azeotropes system via special distillation coupled with reaction

A ternary mixture of diisopropyl ether, isopropanol and water exists in the production process of isopropanol. Ensuring the clean separation of this mixture is of great significance for the clean production of isopropanol from both economy and environment. However, there are three binary and one ternary minimum boiling point homogeneous azeotropes in this system, which complicate the clean production of isopropanol. In this study, with the aid of the hydration reaction of ethylene oxide, clean separation of a ternary mixture of diisopropyl ether–isopropanol–water was realized. Three separation processes—reaction–pressure swing distillation, reaction–extractive distillation and reaction–dividing wall extractive distillation—were proposed. These processes were optimized based on the sequential iterative algorithm, with the minimum total annual cost as the objective. Reaction-pressure swing distillation had the best economic performance, reaction-extractive distillation followed, and reaction-dividing wall extractive distillation performed the worst. The carbon dioxide, sulfur dioxide, and nitrogen oxide emissions of the three processes were calculated based on the optimization results. Reaction-pressure swing distillation had the lowest emissions, followed by reaction-dividing wall extractive distillation. However, reaction-extractive distillation had the highest emissions. The exergy analysis of distillation units and exchanger was performed, and the results showed that reaction-pressure swing distillation had the best exergy performance. Reaction-pressure swing distillation was better than reaction-extractive distillation and reaction-dividing wall extractive distillation in terms of economy, environment, and thermodynamic efficiency. This study provides an original idea for the clean separation of diisopropyl ether-isopropanol-water azeotropes.

Analysis and intensification of energy saving process for separation of azeotrope by ionic liquid extractive distillation based on molecular dynamics simulation

N-heptane and ethyl acetate are important solvents for chemical intermediates such as hexanitrohexaazaisowurtzitane. Based on molecular dynamics simulation, the interaction between mixed extractant and azeotrope was discussed through microscopic mechanism, and the optimal mixing ratio was determined. Based on the process, the influence of different ratio on the separation effect was further verified. It is found that the increase of the proportion of organic solvents can effectively promote the interaction between ionic liquids and esters. In addition, the separation processes of common distillation, extractive dividing wall column distillation and vapor recompression assisted extractive distillation were studied. Based on the sequential iterative algorithm, the process was optimized with the minimum annual total cost as the objective function, and the economy of vapor recompression assisted extractive distillation with mixed entrainers was the best. Compared with traditional organic solvent extractive distillation, the annual total cost can be reduced by 10.81%. According to the optimization results, the economic performance of carbon dioxide, nitrogen oxides and sulfur dioxide emissions was calculated. Through calculation, the technology of vapor recompression assisted extractive distillation with mixed entrainer is significantly better than other processes, and the gas emission can be reduced by 21.03%. In conclusion, the use of mixed entrainer in extractive distillation process can greatly improve the energy-saving effect than a single entrainer. This provides some guidance for industrial separation of other azeotropes.

Multi-objective optimization of a clean, high-efficiency synthesis process of methyl-ethyl-ketone oxime from ammoximation

The ammoximation of ketones is an important chemical reaction, as oximes have a wide range of industrial applications. In this study, the process of preparing methyl ethyl ketone oxime by ammoximation of methyl ethyl ketone was studied. The modeling of the reaction unit was based on the kinetic data, and the reaction conditions were analyzed. Based on the total annual cost, an economic evaluation of the process was carried out, and a sequential iterative optimization algorithm integrating the reaction conditions was used to complete the process optimization. The minimum annual total cost was 2.41 × 107 USD/yr. The exergy of methyl ethyl ketone ammoximation to methyl ethyl ketone oxime process was analyzed, and the unit with the largest exergy loss was determined. The exergy loss rate was 65% for the minimum annual total cost. Considering exergy and annual total cost as the two optimization objectives, the process was comprehensively optimized by a multi-objective optimization algorithm. The annual total cost after multi-objective optimization was 2.55 × 107 USD/yr, which is 5.8% higher than minimum total annual cost, and the exergy loss rate was 64%, which is 1.54% lower than maximum exergy-loss rate. Compared with single-objective optimization, multi-objective optimization has a better comprehensive performance.

Resource allocation method based on mobile edge computing in smart grid

Mobile edge computing(MEC) refers to offloading the collected data to the mobile edge server, which can calculate data efficiently, further improve the operating efficiency of the smart grid and reduce energy consumption. In the face of increasing data, the spectrum overhead and energy consumption of the network are becoming increasingly severe. Based on this, this paper proposes to use massive multiple input multiple output (MIMO) technology to improve the spectrum and energy efficiency of the system, and proposes a massive MIMO-MEC smart grid system framework. In addition, considering that it is easy to be eavesdropped in the process of offloading data, we propose to use physical layer security technology to ensure user information security. In order to minimize the energy consumption of applications such as smart meters, this paper proposes a sequential iterative optimization algorithm to jointly optimize the offloading rate and transmission power. The simulation results prove that as the number of antennas increases, the transmission rate of the system increases, which reduces the total energy consumption of the system.

A new two-scale computational model for hydromechanical coupling in jointed rocks

We develop a new computational model to describe hydro-mechanical coupling in fractured rocks composed of a linear poroelastic Biot medium and nonlinear elastic joints with constitutive response governed by the Barton–Bandis (BB) law. The model aims at capturing increase in stiffness induced by fracture closure during fluid withdrawal. The nonlinear hydro-mechanical formulation is constructed within the framework of the Discrete Fracture Model, with flow and geomechanical sub-systems coupled through a sequential iterative algorithm. The internal contact constraint arising from non-overlapping between opposite fracture faces is enforced through the weak fulfillment of the BB-law. Such a constraint is captured within the framework of the Augmented Lagrangian formulation, where the non-linear mechanical interaction is enforced adopting successive approximations of the Lagrange multiplier, interpreted as the contact pressure in the joint, supplemented by a penalty component associated with the rock stiffness. Furthermore, adopting a traditional flow based upscaling method, macroscopic permeabilities are numerically reconstructed with magnitude strongly dependent on the local stress state. Such a mechanical dependence of the homogenized properties is represented in a discrete manner through pseudo-coupling tables, with enormous potential to be explored within a preprocessing stage in reservoir simulators to compute multipliers relative to a chosen pre-stressed reference state, where input data is available. Numerical simulations are performed for some fracture arrangements illustrating the potential of the formulation proposed herein in bridging hydromechanical coupling at different scales in jointed rocks.

Energy-saving investigation of organic material recovery from wastewater via thermal coupling extractive distillation combined with heat pump based on thermoeconomic and environmental analysis

Cyclohexane and sec-butyl alcohol are widely used in the pharmaceutical and chemical industries. The separation of cyclohexane/sec-butyl alcohol/water azeotropic mixture is of great significance for mitigating environmental risk, recycling solvents, and maintaining developmental sustainability. Based on the COSMO-SAC model, the infinite dilution activity coefficients of the components were calculated, and the final solvent was further obtained by relative volatility and solvent power screening. Taking the total annual cost as the objective function, the extractive distillation process is optimized based on a sequential iterative optimization algorithm. To further reduce the process energy consumption, several energy-saving processes were explored. Energy-saving processes are compared in terms of total annual cost, CO2 emissions, and thermodynamic efficiency. It was found that the total annual cost and thermodynamic efficiency of the processes show an increasing trend, while CO2 emissions show a decreasing trend. The results of stage-exergy indicate that the processes with heat pump are relatively small. The heat pump combined with the thermal coupling extractive distillation process exhibits better thermoeconomic and environmental performance. It provides theoretical guidance for improving the separation and recovery of value organics and energy-saving optimization of processes, which conforms to the theme of sustainable and environmental development.

Modeling and optimizing a fare incentive strategy to manage queuing and crowding in mass transit systems

This paper solves the problem of optimizing a surcharge-reward scheme and analyzes equilibrium properties incorporating commuters’ departure time choice to relieve crowding and queuing congestion in mass transit systems. The surcharge-reward scheme incentivizes commuters to switch departure times from a pre-specified central period to shoulder periods. We formulate a bilevel model to design and optimize the surcharge-reward scheme. The upper-level problem minimizes the total equilibrium costs by determining the refundable surcharges, the rewards, and the corresponding central charging period. The lower-level problem determines the equilibrium of commuters’ departure times with respect to generalized travel costs. Equilibrium properties are analyzed and a sequential iterative solution algorithm is developed. We found that the existence of an optimal solution depends on the scheme design and there exists a lower bound on the surcharge to achieve the system optimum. Numerical studies are conducted on a commuting rail line in Copenhagen. The proposed algorithm converges efficiently, and the fare incentive scheme can simultaneously reduce the individual trip costs, total crowding costs, and total queuing time costs. The performance of the scheme increases with the rewards and surcharges up to a point and beyond which it stays unchanged.

Performance evaluation of different extractive distillation processes for separating ethanol/tert-butanol/water mixture

In this study, a few extractive distillation processes with ethylene glycol as solvent are proposed to separate high-purity ethanol and tert-butanol from industrial effluents, including direct ED, indirect ED and heat pump assisted indirect ED. All these processes are optimized with targeting minimum annual total costs by a sequential iteration algorithm. After that, a head-to-head comparison is completed regarding distributions of cost, energy and CO2 emission. The results showed that the direct ED is the most economical process reducing minimum annual total costs by 30.15 % than the heat-pump assisted indirect ED, while the latter is the most energy-efficient process saving energy consumption and CO2 emissions by 14.78 % and 21.79 % than the direct ED.

Energy-saving investigation of hybrid reactive distillation for n-butyl acetate production from two blending feedstocks

Several reactive hybrid distillation configurations are demonstrated for the n-butyl acetate production via transesterification with two blending reactive feeds, namely MM20 (70 wt% methanol, 30 wt% methyl acetate) and MM80 (35 mol% methanol, 65 mol% methyl acetate). Pervaporation (PV) with polyoctylmethylsiloxane membrane is used to replace a methanol recovery column because what’s obtained at the retentate side is high purity methanol rather than methyl acetate. The valid mathematical model for this PV process is developed in Aspen Custom Modeler. These configurations along with conventional reactive distillation (CRD) and reactive dividing-wall distillation (RDWC) are optimized by a sequential iterative algorithm based on the minimum total annual cost (TAC). The comparison with CRD shows that RDWC + PV is the optimal alternative of two hybrid configurations for MM80, saving 40.8% operating cost, 32.6% TAC, and 48.77% CO2 emissions. However, other two hybrid configurations for MM20 are unable to achieve the desired energy-saving.

Video Anomaly Detection with Sparse Coding Inspired Deep Neural Networks.

This paper presents an anomaly detection method that is based on a sparse coding inspired Deep Neural Networks (DNN). Specifically, in light of the success of sparse coding based anomaly detection, we propose a Temporally-coherent Sparse Coding (TSC), where a temporally-coherent term is used to preserve the similarity between two similar frames. The optimization of sparse coefficients in TSC with the Sequential Iterative Soft-Thresholding Algorithm (SIATA) is equivalent to a special stacked Recurrent Neural Networks (sRNN) architecture. Further, to reduce the computational cost in alternatively updating the dictionary and sparse coefficients in TSC optimization and to alleviate hyperparameters selection in TSC, we stack one more layer on top of the TSC-inspired sRNN to reconstruct the inputs, and arrive at an sRNN-AE. We further improve sRNN-AE in the following aspects: i) rather than using a predefined similarity measurement between two frames, we propose to learn a data-dependent similarity measurement between neighboring frames in sRNN-AE to make it more suitable for anomaly detection; ii) to reduce computational costs in the inference stage, we reduce the depth of the sRNN in sRNN-AE and, consequently, our framework achieves real-time anomaly detection; iii) to improve computational efficiency, we conduct temporal pooling over the appearance features of several consecutive frames for summarizing information temporally, then we feed appearance features and temporally summarized features into a separate sRNN-AE for more robust anomaly detection. To facilitate anomaly detection evaluation, we also build a large-scale anomaly detection dataset which is even larger than the summation of all existing datasets for anomaly detection in terms of both the volume of data and the diversity of scenes. Extensive experiments on both a toy dataset under controlled settings and real datasets demonstrate that our method significantly outperforms existing methods, which validates the effectiveness of our sRNN-AE method for anomaly detection. Codes and data have been released at https://github.com/StevenLiuWen/sRNN_TSC_Anomaly_Detection.