Sodium aluminates solution filtrating process is an essential section in the alumina production process.To improve quality of products,automatic control is extensively used and plays an important role in the production chain.However,filtrating process is a complex production process which has a number of equipments to be controlled.The control system of filtrating involves not only logical control but also closed loop control and there is complex interaction between them.To reduce the field test cost and risk efficiently,based on a real control system and a virtual filtrating process object,this paper presents a simulation experiment platform for sodium aluminates solution filtrating process control.The presented platform makes it possible for research on the modeling,control strategy,optimization of the filtrating process,and provides a functional experiment platform for the research of advanced control technique,such as fault diagnosis,faulttolerant control of the filtrating process.

In this paper,an algorithm for gas pipeline leak detection based on Hilbert-Huang transform is studied.The characteristics of the leaked acoustic signal in frequency domain can be analyzed by computing and comparing the before-and-after fluctuations of the energy of Hilbert marginal spectrum.The method judges whether the leak occurs or not by comparing the Hilbert spectrums of current signal with normal signal in the sensitive frequency band,and then locates leak according to the time difference of the leak acoustics arriving in the upstream and downstream ends.Off-line tests based on real history data of industrial pipeline proved the validity of the method.

In refinery,hydrogen,which is a precious and clean energy resource,is also the by-product and significant material source of petroleum refining and petrochemical hydrogenation process.To reduce cost and save energy for petrochemical industry,the hydrogen system should be operated under the optimal scheme to meet the varying demand from process.In this paper,the impact factors of hydrogen system operational scheduling are elaborated.A mixed integer nonlinear programming(MINLP)model that involves all the impact factors is established to optimize the multi-period scheduling of hydrogen system and solved with Lingo software.A case study based on the data from certain refinery factory is showed to demonstrate the effectiveness and feasibility of the scheduling methodology,as well as this method proposed plays an important role in guiding the scheduling of hydrogen system in refinery.

Aiming at the unrealistic assumptions of oracles and the difficult determination of control parameter in fault detection for alumina evaporation process,a modified cost-sensitive active learning(MCAL)method is proposed.The formal description of MCAL is given and unrealistic assumptions of oracles are relaxed.MCAL hybridizes particle swarm optimization(PSO)and cost-sensitive active learning(CAL)to improve the classification accuracy with multi-oracles and reduce the cost of labeling instances sampled.This optimization mechanism employs the continuous-valued PSO to optimize the control parameter to maximize the value of information of instance and minimize the cost of oracle.MACL is applied on the benchmark of alumina evaporation process for fault detection.Experimental results show that MCAL correctly selects the control parameter,obtains low misclassification cost,reduces labeling costs and increases the rate of fault detection.

Considering the imprecision of mechanism model and the weak generalization of single model,a support vector machines(SVM)multi-model modeling algorithm based on affinity propagation clustering is presented for arachidonic acid(ARA)fermentation process.Firstly,ARA training samples are clustered into several classes by affinity propagation clustering algorithm.Then,the sub-models are trained by SVM according to corresponding sub-class samples.The testing samples are firstly assigned to appropriate sub-classes by similarity measurement,and then the predicted outputs are estimated by corresponding sub-models.The modeling experiments of ARA process indicate that,compared with the mechanism model and other single models,the proposed algorithm has superior regression accuracy and good generalization ability.

There are some issues caused by the network’s intervention,such as random time-varying delay,data packet dropout and out of order,which are widespread in data transmission.All these are not conductive to system control in networked control systems.Based on the thought of predictive control,an asNMPC method is applied to compensate the above disadvantages in nonlinear networked control systems.When the network transmission links exist in both feed-forward and feedback channel simultaneously,the given method can reach good control effections.The asNMPC method for nonlinear networked control system shows strong robustness comparing with DMC,especially in the case of model mismatch.

The usual direct optimization approaches can hardly obtain a “good” numerical solution for the dynamic optimization(DO)problem with control switching structure if the chosen control discretization does not properly reflect the switching structure.In this paper,by reformulating the DO problem as a multi-stage optimization problem,a second exploitation strategy is proposed for solving this problem.The potential control structure can be revealed from the first solution generated by the usual optimization approaches.Subsequently,the DO problem is partitioned as several stages,with each stage corresponding to a particular control arc.A control vector parameterization approach is applied to convert the multi-stage DO problem to a static nonlinear programming(NLP)problem.The control profiles and stage lengths act as decision variables.Based on the Pontryagin maximal principle,a multi-stage adjoint system is constructed to calculate the gradients required by the NLP solvers.Two examples are studied to demonstrate the effectiveness of this strategy.

A nonlinear robust continuum regression(NLRCR)modeling method combing radial basis function(RBF)and robust continuum regression(RCR)is presented to solve the problem that RCR algorithm hard to build the nonlinear system modeling of high predicting precision.Firstly,this approach uses RBF to carry out mapping transformation from nonlinear sample data to high dimension eigen space.Then,RCR algorithm is used for building linear regression modeling in high dimension eigen space.Simulation experiment is performed to show the effectiveness of this method.And it is utilized to develop a soft sensor model for the gasoline dry point in delayed coking,and highly precise prediction results are obtained compared with RCR and RBF-PLS methods.

A fuzzy optimization control strategy based on the property of distributed parameters is presented.Both the error of furnace temperature and its changing rates at different segments of the reheating furnace are used as the input of fuzzy logic,so that a group of 2-dimension fuzzy controller is designed with respect to every furnace temperature adjustment point.According to the features of two operative statuses,which are rising temperature segment and maintaining temperature segment,the fuzzy control rules are designed respectively.Based on the variable universe theory,an adaptive mutation differential evolution algorithm,in which the dynamic fitness function is constructed and the mutation probability is adaptively regulated according to the evolved generations,is applied to optimize the fuzzy controllers.The running results show that,the method proposed in this paper is good at controlling the temperature of titanium slab furnace.

In order to analyze nonlinear and dynamic characteristics of industrial process,a new method combining dynamic maximum variance unfolding and one-class support vector machine(DMVU-OCSVM)was presented.Manifold learning method DMVU was applied to obtain nonlinear and dynamic manifold features.OCSVM was used to build statistic model based on manifold information and nonlinear monitoring statistic was constructed to detect fault online.The simulation results on continuous stirred tank reactor system showed that the proposed method could detect process fault more effectively.

For the actual distillation process with side reactor, a dynamic simulation system including control system is developed based on reasonable simplifying and assumptions.Some typical fault disturbances which induced the dynamic imbalance between the separation capability and reaction capacity, such as the disturbance of unreasonable feed flow ratios for reactor, the disturbance of descended the reflux flow, the disturbance of insufficient steam heating for column bottom are introduced reasonably in the dynamic simulation system.The dynamic characteristics under the condition of above fault disturbances are simulated and researched deeply, then the virtual fault data is obtained.The problem which the fault information cann’t be gained by experimental research in the actual distillation process equipment is solved in this paper.

A multi-period,multi-product planning optimization under uncertainty is discussed.Based on the discrete-time modeling method,a mixed integer linear programming(MILP)model,in which the nonlinear part is converted to linear problem using fuzzy possibility method is proposed.Meanwhile,the usefulness of MPC as a tactical decision policy is integrated to the model.The effectiveness of the proposed model is illustrated through a real world refinery supply chain.

When a soft sensor model is constructed for a complicated production process,a single support vector machine(SVM)model or a compositional SVM model based on conventional clustering methods sometimes cannot track mutant signal well or obtain a satisfactory generalization.An improved linear discriminant analysis(LDA)algorithm is proposed in this paper so as to solve the problem.The feature vectors are obtained by combining boundary analysis with LDA between the categories.The original sample data are transformed in terms of the feature vectors,and sub-models based on SVM are respectively constructed by transformed data.And then the compositional parameters for sub-models are designed according to the sum of the effectual characteristic values in the every feature vector.Finally,a compositional SVM model is constructed.The simulation results show that the composition model can reduce the information interference among the different data categories and improve the inferential accuracy of the model.

Grinding and classification processes(GCP)are the key unit operations in metallurgical concentration plants.The product particle size directly affects the final product’s ore grade and metal recovery rate.GCP is essentially a multi-input-multi-output(MIMO)system characterized by strong disturbances,dead time and reverse response.Disturbance observer(DOB)based model predictive control(DOB-MPC)is proposed to handle the external and internal disturbances.The simulation results demonstrate that the proposed methods have better disturbance rejection properties than the MPC method,whether in rejecting external disturbances or in rejecting internal disturbances,such as model mismatches and coupling effects.

The fuzzy C means clustering is used to divide the swarms adaptively,and a fuzzy C means multi-swarms competitive PSO(FCMCPSO)algorithm is proposed.According to the scale of the swarms to select different optimal strategies,the swarm of large scale uses the standard particle swarm algorithm to optimize,and the swarm of small scale randomly searches in the optimal solution neighborhood,increasing the probability of jumping out of the local optimization.Within every clustering,the adaptive value of every clustering swarm by competitive learning is respectively found and arranged the order of the different adaptive value,and then the swarm of small adaptive value integrates with the neighboring swarm of large adaptive value,ensuring the particle swarms to search towards the optimal solution by the competition in the swarms.The validity was tested by the benchmark functions to improve the global search capability.At last,the proposed algorithm was used to optimize the operational conditions of high density polyethylene(HDPE)equipment in order to decrease the consumption of ethylene.

One multi-variables constructive model predictive control strategy is developed for continuous-time Hammerstein systems with constraints on the state,input and intermediate variable.Firstly,based on the special structure of Hammerstein systems,one control Lyapunov function is constructed for the linear subpart via the Raccati equation.This leads to a class of adjustable control law(i.e.stable control class)of the linear subpart.Secondly,a finite receding horizon optimal control problem is defined for the constrained Hammerstein system,which is computed on-line to derive the actual predictive control actions.The sufficient condition for asymptotically stability of the closed-loop system is obtained by the Lyapunov’s theorem.Finally,the simulation example of the grade transition control of industrial polypropylene plants is used to illustrate the effectiveness of the results obtained here.

On-line detection of gasoline octane number is very important for the production of high quality gasoline,and is essential for the reconciliation process.Compared with the deficiency of the traditional detection methods,a new octane number detection method based on Raman technology was proposed.The system consisted of on-line detection,data pre-processing,and calculate model of gasoline octane number.The results showed that the system detected the octane number of gasoline accurately,and displayed the Raman spectra obtained by the detector.

A new control structure of the simplified externally heat-integrated double distillation columns(S-EHIDDiC)is proposed,where three heat exchangers are used to approximate external heat integration between the rectifying/stripping sections of the high-pressure(HP)/ low-pressure(LP)distillation columns.The heat duties of the top/bottom heat exchangers are used to control the top/bottom product qualities of the HP/LP distillation columns.Manipulating the heat duties of the heat exchangers serves to change the reflux/reboiled flows of the HP/LP distillation columns,and it is,in principle,similar to the LV control of conventional distillation columns.Some additional decision variables exist in the S-EHIDDiC(i.e.,the feed splitting ratio,the pressure of the HP distillation column,and the heat duty of the intermediate heat exchanger)and can be utilized to minimize utility consumption.The proposed control scheme is evaluated and its feasibility is demonstrated.Improvement in thermodynamic efficiency can also be achieved with the adjustment of those additional decision variables.

One of the critical open issues in the nonlinear model predictive control(NMPC)scheme is the computational burden associated with the solution of the optimization problem,since at every sampling time a nonlinear dynamic optimization problem must be solved in real-time.To alleviate the aforementioned problem,a new NMPC algorithm using the global orthogonal collocation method is proposed.Higher order interpolation polynomial is used to simultaneously discrete state variables and control variables over the optimization horizon and the original continuous dynamic optimization is transcribed to a nonlinear programming problem(NLP).The NLP problem has a fixed structure with certain computational advantages and can be solved by an appropriate numerical optimization algorithm.Taking full advantage of the features of the global orthogonal collocation,the proposed algorithm provides the potential to reduce the scale of NLP and thereby reduce computational burden efficiently,even it works with a long optimization horizon.The effectiveness of the proposed algorithm is demonstrated by its application to a continuous polymerization process.It is found the algorithm achieves a smooth transition for large-magnitude setpoint changes and behaves well in the presence of disturbances.

Heat-setting process is one of the major energy-consuming unit in dyeing industry.Firstly,based on stereotypes of the physical structure and process principles,model of energy consumption and the key factors of the process is derived using Newton〖DK〗’s heat balance and heat exchange formula.Then,in the Matlab/Simulink software,the following model is constructed.Finally,after the conversion of the energy model into the optimization problem of minimum energy consumption,particle swarm optimization algorithm is introduced to solve them.Simulation of industrial data shows that the proposed energy consumption model is consistent with on-site process and particle swarm optimization results can provide a reference and guidance for the operation.

The test criteria of ASTM D2887 and ASTM D86 for distillation range is one of the important research area in gas chromatography.For correlation modeling the generalization and robustness is also the major target to evaluate.In this paper a modeling method based on the weighted mean-square deviation and variance optimization is proposed.This method is used in Engler boiling range distribution measurement which usually calculated by ASTM means,to calibrate the coefficients of the original model by the value errors between the model prediction and laboratory analysis.Theory study，mathematical calculation and technology application verified the robustness and generalization property of this method.The application results indicated that the correlation modeling method of the chromatography distillation range and Engler boiling range can be used to the application of simulation distillation and a reliable analysis means for quickly measuring the oil distillation range.

The advanced process control systems for the SC-1 type ethylene cracking heaters at Lanzhou Petrochemical Company 460KTA Ethylene Plant were designed,including the average coil outlet temperature controllers,the pass outlet temperature balance controllers,the total throughout controllers. The software and hardware structure of the control systems,the switching logic between advanced control and DCS regular control,the DCS operation interface for advanced control were introduced. The control steadiness and control accuracy for cracking heaters are greatly improved by using the advanced process control systems,and remarkable economic benefit is obtained.

This paper presents a new approach to construct controlled variables(CVs)for chemical processes.The necessary conditions of optimality(NCOs)are selected as CVs,thus the process is maintained at the optimum operation point under the actions of feedback controllers.However,some NCOs are online immeasurable,so the nonlinear function relationships between measured variables and unmeasured NCOs are found via neural network,and select the functions as CVs.An exothermic reactor example illustrates that the proposed method is able to achieve approximate optimal control.

Due to complexity and uncertainty of condenser fault diagnosis,a new method for fault diagnosis of condenser vacuum based on fuzzy rough set and case-based reasoning was presented.By analyzing the characteristics of condenser vacuum fault diagnosis,a fault tree was built.Weight membership was used to retrieval radical nodes of fault tree,and leaf nodes were retrieved by nearest neighbor algorithm.Considering many fault features and their relativity,fuzzy rough set was used in feature reduction and weight allocation to extract main features,reduce non-linear relationship between features and avoid human’s subjectivity influence on weight allocation.Simulating of turbo condenser fault data proved that the method is effective.

Load change problem synchronized with pipe network pressure increasing of three thermo-coupled columns in a cryogenic air separation unit was discussed.Due to the thermo-coupled features among the distillation columns,the simulation of the load change operation of a cryogenic air separation unit is difficult.Analytical results reveal that slight change of the operation variable can cause dramatic change of some other variables,which results in the difficulties of convergence due to inaccurate gradient approximation.The homotopy-based backtracking method(HBM)can efficiently deal with the load change simulation of the thermo-coupled system.The nitrogen block phenomena can also be simulated with this method.

As wavelet neural networks(WNN)algorithm usually has low convergence rate and easily falls into local minimum,an improved wavelet neural networks(IWNN)is proposed to modify the parameters.Firstly,the traditional Sigmoid function is replaced by hyperbolic tangent function for output layer.Secondly,the learning step is selected by adding momentum to the weight adjustment to improve learning efficiency.The proposed IWNN method is used to build soft sensing of light diesel pour point from fluidized catalytic unit(FCCU)main fractionator.Compared with the models of BP and WNN,the results obtained by the IWNN approach showed a better prediction accuracy and generalization capability.Moreover this modeling can be used to guide production efficiently.

Fermentation processes optimization involves multiple and conflicting objectives and the features of these processes contain many complexities.In this paper,a design of a Pareto-based distributed Q-learning(PDQL)optimization strategy was presented to solve Pareto optimal flow rate trajectories for the lysine fed-batch fermentation process.Q-learning algorithm and Pareto sorting method were combined to generate the nondominated solution set and to make it approximate the actual Pareto front.The strategy described the relation of multi-objectives with the help of rewards strategy.For enhancing searching capability,multiple randomly initialized groups of agents were used.The result of PDQL optimization was compared to PSO with the aggregated function method to test its performance.

As for low forecast accuracy of cobalt ion concentration by least square support vector machine(LS-SVM)method in the purification process of zinc hydrometallurgy，a multi-model fusion soft sensor modeling method based on combination LS-SVM with ARMA model was introduced in purification of zinc.Firstly the series of cobalt ion concentration was decomposed by wavelet transform,the decomposed sub-sequences were reconstructed by phase space reconstruction.Each sub-sequence was modeled by LS-SVM method in phase space,then the output of each model was integrated by wavelet reconstruction.Secondly correction was made for error of LS-SVM modeling output in ARMA model.Finally the output of two models were integrated,the integration value was the estimated value of cobalt ion concentration.The method was applied in prediction of cobalt ion concentration in the entrance of purification process of zinc hydrometallurgy. The results showed that this method had better generalization performance and high prediction accuracy than LS-SVM method,which showed good potential for application.

Varying load of the industrial process impacts the characteristics of nonlinear dynamic system.The study focuses on the nonlinear dynamic processes with varying load,and the models of the processes which can be presented as ordinary differential equations or semi-explicit Heisenberg differential-Algebraic equations.The nonlinear model predictive control(NMPC)algorithm based on the rigorous model is applied here.A two-tier structure of steady-state and dynamic optimization is presented,and simultaneous approach is adopted to solve optimal problem.And then the stability of the approach is analyzed.Simulations on the cascade continuous stirred tank reactors show the validity of the NMPC.

Initially,the techniques to construct unilateral and bilateral nonlinear membership functions involved in fuzzy programming are introduced.Motivated by readily solving nonlinear fuzzy programming problems,this paper subsequently presents novel approaches to convert the nonlinear membership functions to piecewise linear ones.Therein,Gaussian integral functions are employed to calculate the discrete points,which are then offered to decision makers for benchmark options to build the piecewise linear membership functions.Finally,the formulations of the corresponding fuzzy programming are presented,along with a numerical example studied to demonstrate the effectiveness of the contribution.

For the determination of individual dye concentrations in the mixture,especially in the case that the concentration of the mixture is too high and the absorbance system is to be nonlinear because of chemical changes such as hydrolysis and the interaction among the dyes,a novel determination method based on particle filter is proposed and a nonlinear absorbance model is developed.The feasibility and validity of this method can be verified by real examples.The numerical results for the determination of individual dye concentrations in the mixed reactive dye liquors by particle filter are compared to the results of the determination method based on extended Kalman filter,showing that the particle filter can produce better filtering effect than extended Kalman filter and the mean squared error(MSE)between its determinated results and actual values is less than 0.015.This demonstrates that the nonlinear absorbance model developed in this paper is accurate and the determination method based on particle filter is effective and feasible.

Melt index prediction of polypropylene is important but very difficult.A method based on PCA-RBF neural network which is optimized by a new ant colony system is proposed.Principal component analysis(PCA)is used to map high-dimension initial data to new low-dimension data,and then the corrections of the input variables are eliminated and the most relevant process features are selected.Radical basis function(RBF)neural network is used to characterize the nonlinearity.Finally the new ACS which works for continuous optimization problems is employed to optimize the weights of the RBF neural network.According to the research on the data from a real plant,it shows that the model works well and provides promising accuracy and reliability.

A model predictive control strategy based on neural network is presented for a continuous stirred tank reactor(CSTR).A segmentation method was adopted to identify Hammerstein-Wiener model coefficient by least squares support vector machines and then to construct a nonlinear predictive controller which was by a linear optimal component and radial basis function neural networks in series.A nonlinear predictive control algorithm based on least support vector machines Hammerstein-Wiener model was realized by using BP neural network to train predictive input sequences and to solve nonlinear predictive control rules by Quasi-Newton method.The simulation results of CSTR illustrate that this approach is effective tracking and controlling product concentration.

Using a single model to describe a complex nonlinear object,it usually suffers from low accuracy and poor generalization.A multiple model soft sensor approach is presented based on local reconstruction and fusion manifold clustering.In order to restraining the impacts of outliers to clustering results,the data set is split into several small disjoint sub-clusters.By reconstructing linear manifold level based on every sub-cluster respectively,the sub-clusters which are closer and in the same manifold level are merged.Meanwhile,support vector machine is used to construct regression model in terms of each sub-class and a soft-sensor composed model based on the multiple sub-models is obtained finally.The proposed algorithm is used in a soft sensor modeling for the Bisphenol-A productive process,and the result of simulation shows the effectiveness of the algorithm.

This paper gives a research on IB(inverse Boosting)algorithm and proposes an improved version of IB called IB+.Both IB and IB+ algorithm will enhance the weight of samples which have been classified correctly during the training process.The most difference between IB and IB+ is the method to update the weight of training samples in each iteration.For IB algorithm,the weight of training samples will be updated according to an inverse error vector which was decided by the performance of the last trained single net.However the IB+ algorithm adopts a mesosphere ensemble net instead of a single net to determine the inverse error vector thus a more suitable sample distribution will be achieved.Further experiment results show that the performance of ensemble net which was developed using an inverse error vector to create new sample distribution will be decided by the performance of base single net not the degree of correlation.

The measurement and control of mass flow rate of gas/solid two-phase flows have not been effectively solved for a long time in the industry.In this paper,a novel helical shaped surface plates capacitance transducer is designed for industrial point of view.It is based on the electrical capacitance tomography(ECT).The influence of the opening angle of electrode,the pipeline permittivity and pipeline thickness on sensitivity distribution and homogeneous error are analyzed by the method of finite element simulation.Through the simulation analysis,the range of model parameters can be fixed.Furthermore,it can provide the theoretical basis for development of sensitivity distribution,optimization of sensor and the performance of transducer.

A triangular reaction network lumping three xylene isomers into a species is proposed for hydroisomerization of C8-aromatics and the corresponding kinetic model for an industrial reactor is also developed.The kinetic parameters are estimated by the traditional optimization method based on several sets of balanced plant data.Meanwhile,the model is simulated by large number of plant data at different operating conditions and a good agreement is obtained between the estimated and observed values.When compared with other models,the good performance of the proposed model is demonstrated in detail in terms of the indexes such as relative error,relative mean square error and parameter estimation time.

A novel coal chemical industry about decomposing coal with plasma is introduced,and three-dimensional visualization monitoring system which collects production real-time data and monitors the real process about product production is described.And also three components of this monitoring system,including three-dimensional modeling environment,process integrated simulation system,visualization monitoring platform are introduced.According to the specific chemical process,the principle and realization about each component in detail is explained.With the visualization and interaction about production process data and application,this monitoring system can provide effective supports for process simulation and visualization monitoring about the novel coal chemical industry decomposing coal with plasma.

In order to handle the problem of fault detection for industrial process,an improved multiscale principal component analysis(MSPCA)is proposed.Firstly,considering the nonstationary and random nature of data in the process industry which contains different noises inevitably,an improved wavelet threshold denoising method which combines multiple wavelet transform with a new threshold function based on the characteristics of wavelet analysis is proposed.The data collected from the industry condition are processed by means of the improved wavelet threshold denoising method.Using wavelets,the individual variable is decomposed into approximations and details at different scales.Contributions from each scale are collected in separate matrices,and a PCA model is then constructed to extract correlation at each scale.According to the simulation of propylene polymerization,and comparing the improved MSPCA with traditional MSPCA,it shows that the improved MSPCA has enhanced the accuracy of process monitoring and fault diagnosis.

For simulating the accelerated cooling process and its control system,a hardware-in-loop platform is established in this paper.The platform is designed including a pilot product line,the Level 1 and Level 2 control system.The control algorithms of presetting set points,feed-forward compensation control and adaptive feed-back control are realized in this platform,which makes the platform’s control system have the same functions as the real product line owned.One can analyze the in-field problems and study how to improve the control accuracy through this platform in the first step,and then apply the results to the real product line.Therefore,the technical innovation risks are decreased dramatically.

The coupled atmospheric reaction-vacuum distillation technology,that is a distillation column connected with side reactors,is employed in the production of benzyl chloride.An independent reaction amount is introduced in the reactor model to simulate the process.Moreover,a systematic design approach based on the Powell method is developed to seek the optimum configuration parameters.The effects of the design parameters,such as the number of stages in the distillation column,the number of stages between adjacent reactors and the number of reactors,on the process performance are well investigated.Results demonstrate that with the given vapor boilup rate,insufficient separation leads to a poor system performance,and excessive stages or reactors will not be benefit for the improvement of the reaction capability.The optimum match between reaction operation and separation operation can be achieved with the optimum design parameters obtained through the process design approach.

Particle swarm optimization(PSO)which has the general purpose optimization method received much attention in past years.In many studies,PSO has been successful in a variety of optimization problems.But the speed of convergence of standard PSO algorithm on high dimensional search space is unacceptable in practice.The concept of particle health was proposed,and gives an algorithm for particle health calculation in this paper.A new variation of PSO model proposed based on particle health（HPSO） can effectively reduce the probability of local optimum and enhance convergence speed especially for high dimensional search spaces.The proposed were tested by a variety of high-dimensional benchmark functions,and compared with standard PSO algorithm and decreasing inertia weight variation(WPSO). It was found that the application of these modifications resulted in significant gain in speed and efficiency.

Air separation is an essential industry for national economic development,its energy consumption has become the bottleneck of the development.Internal thermal coupled air separation column(ITCASC)changed the structure of traditional air separation column,which can achieve good energy efficiency and is a research front of air separation energy saving control.This paper presents an ITCASC hybrid method with the PCA-CGA-RBF statistical model of liquid phase composition,pressure and equilibrium temperature to instead of the traditional mechanisms of equilibrium stage modeling method,the proposed method can improve the efficiency of solving the model significantly.The results show that the solution time decreases from 31.06 s to 11.18 s,64.01% reduction,while the accuracy remains basically unchanged,which paves the way of further optimization and control researches.

For some chemical processes that are difficult to reach a steady state due to the existence of batch process or material recycle,as well as some nonlinear processes that difficultly operate nearby their desired set-points or reference trajectory because of the presence of disturbance,the sub-optimal or invalid optimal solutions will be derived if the normal steady state optimization scheme is used.In this paper,a dynamic real-time optimization strategy used for solving the optimum of the significant variables in chemical processes dynamically is proposed.Such a scheme is implemented in a hierarchical control structure,namely,the real-time optimization,which locates the top level in such hierarchy,is in charge of solving the optimum of some critical variables based on the dynamic model of a process,such optimization scheme is a kind of dynamic optimization other than steady state one,meanwhile,such scheme considers both dynamics of a process and material,prices of product in recent market,so the optimal economic profit will be arrived.The feasibility and efficiency of the proposed method is illustrated by a case study.

A dual model RBF(radial basis function)neural network was proposed in this paper.One is used for self-learning,which learns one time a day.The other is used for on-line correcting,which is the running model currently.Both the self-learning model and the on-line correcting model are corrected six times every day and should track the current conditions of the system quickly.At the same time,the accuracy of the two models should be compared.If the accuracy of the on-line correcting model is less than the one of the self-learning model,the latter becomes the new currently running model instead of the old one. Otherwise,the currently model is maintained.To solve the problem of neural network large prediction errors,a network algorithm analysis is given and the influence factors of the network prediction accuracy are found.At last,an improved algorithm of RBF neural network modeling is proposed,which combines K-means clustering method with the recursive descent algorithm.Simulation and practical application proved the effectiveness of the improved method.

In order to deal with the nonlinear and complexity of Aromatics isomerization process,multi-model algorithm with affinity propagation(AP)cluster and least squares support vector machines is proposed to make up the shortcoming of single model.Firstly,the data of isomerization process are clustered into several groups with AP algorithm.Then a sub-model is constructed for each group with LS-SVM. By comparing Euclidean distances between the test sample and all cluster centers,the group which the test sample belongs to is determined.Finally,the test sample is input into the corresponding sub-model to predict the output.Experimental results show higher accuracy of the proposed algorithm，compared with single model and k-means clustering based neural networks.

A hybrid algorithm based on extremal optimization(EO)with adaptive lévy mutation and information fashion algorithm(EOIFA)was proposed in this paper.It applied the idea of combination mechanism of global and local search.In the process of the global search,IFA is an evolutionary algorithm based on the difference in group that can quickly approach an approximate optimal solution.EO has powerful local search capability.During the local search,it helps IFA out of local maximum points by selecting the approximate solution of the worst element to proceed adaptive lévy mutation.The EOIFA is applied to train artificial neural network(NN)to construct a practical soft-sensor of removal efficiencies for supercritical water oxidation.The obtained results indicate that the new method proposed by this paper is feasible and effective.

By adopting Newton’s Cooling Law,heat transfers between fluid and surfaces of metal tube(heat transfer oil and internal metal tube surface,cold air and external surface of metal tube)are illustrated respectively.Then based on Fourier Experiment Law,heat transfer inside inner metal tube is introduced,and boundary condition is presented.At last,dynamic relationship between hot air temperature,flux and temperature of heat transfer oil,flux of air is summarized.Simulation and experiment show that dynamic and static characters between model and real plant are coincident.Static data show that mean and standard deviation are all less than 5%,which means model is exact enough for practice application.Moreover,cause of error is analyzed.

A modeling method by radial basis function(RBF)network based on enhanced global K’-means algorithm(EGK’M)was presented.An EGK’M algorithm was proposed to determine the hidden layer structure of RBF network,including the number of hidden layer nodes,the position of each center and the width of basis function.KPCA algorithm was used to extract non-linear feature information and to achieve secondary selection of auxiliary variable.The obtained model was compared with the model based on principle components analysis with EGK’M-RBF and the model based on KPCA with RBF network based on K’-means algorithm.Experiment results demonstrate that the model proposed in this paper gives better predictive ability,smaller absolute error and mean square error.

The existed evaluation of energy efficiency of ethylene equipments does not consider the different raw materials, process technology and equipment sizes, and ignores the relative factors of energy efficiency index, so it cannot analyse the saving energy choice among energy consumption factors.The paper puts forward the energy efficiency analysis method for ethylene equipments according to technologies, scales and data distribution of energy consumption including fuel, power, steam and water inner ethylene process boundary.Furthermore, energy efficiency virtual benchmarking method is proposed based on dependent function analytic hierarchy process model (DFAHP), with which it can manipulate multi-factors and energy efficiency index together.It outweighs the traditional evaluation method such as the mean method and optimal index method.The applications of an ethylene equipment, even the whole ethylene industry for energy efficiency analysis are tested.The validity of proposed method is verified, which is more helpful to find the saving energy chance and quantified energy targets.