Due to too much constraints and variables in process scheduling model based on uniform discrete-time representation, batch short-term scheduling based on continuous-time representation has received an intensive concern and study.In this paper a brief introduction of process scheduling was given, a short-term scheduling MILP model based continuous-time representation of printing and dyeing industry was presented after studying on short-term scheduling and production recipe of printing and dyeing.Then MILP model was translated into a solvable model described by ILOP OPL language, this model together with data of two cases from a printing and dyeing corporation in Zhejiang Province were solved by ILOG CPLEX, the results were illustrated in Gantt charts.The result shows effectiveness of MILP model of printing and dyeing process and production resources in shop floor are optimized.

It is the most effective resource use practices for slurry and slime to be used as fuel for power generation.The wet slurry pipeline transportation is an important part for the process of power generation, and there are design and operational optimization issues when pipeline transportation system was used .In order to achieve system optimization and energy conservation, the slurry wet pipeline transportation system model and the boiler heat utilization model were established based on rheological properties of slurry transportation, slurry heat and fluidized bed boiler combustion characteristics.Then an integrated optimization objective function based on power consumption cost and slurry consumption cost was established.The whole optimal problem was optimized by nonlinear optimization algorithm of SQP.Finally, with the consideration of parameters such as slurry price, wet slurry pipeline transports distance, power consumption, and other parameters which can impact the optimization objective, analysis and optimization calculations were carried out to obtain the detail results.The research results and present findings are of important significance to the operational optimization and energy conservation of fluidized bed boiler.

Considering that it is difficult to analyze the convergence of iterative learning optimal control for quality control of batch processes, a novel iterative learning control based on data-driven neural fuzzy model for product quality control in batch process is proposed in this paper, which results in the convergence of the product quality and control trajectory in batch axes.Moreover, the rigorous proof is given.Lastly, to verify the efficiency of the proposed algorithm, it was applied to a benchmark batch process.The simulation results show that the proposed method is better and can be applied to practical processes, thus it provides a new way for the control of batch processes.

Dozens of advantages have been reported on using support vector data description (SVDD) in the fields of non-Gaussian process monitoring and fault diagnosis.However, during the SVDD model construction offline, usually the whole training data set is used.Due to the tremendous size of the data set, the computation burden for modeling is quite heavy, which leads to the difficulty in updating model online.Therefore, this paper proposes a fast SVDD algorithm on the basis of the feature samples.In this new algorithm, the feature samples are used in stead of the whole training data set for modeling in order to significantly reduce the computation complexity.Concurrently, PCA is replaced by the local tangent space alignment (LTSA) to extract the underlying manifold structure of the process data set, since the traditional dimension reduction methods, such as PCA, have poor capability to handle nonlinearity.Next SVDD is applied on the manifold.At last, corresponding statistical indices are used for fault detection purpose.The proposed method has been tested on the Tennessee Eastman (TE) process, while the simulation results show the efficiency of it.

Problem in this paper is based on the Saint-Venant equations of the transfer function model.It was applied to urban drainage systems combined sewage overflow control to minimize spillover rate.Using the existing drainage system facilities，through flexible traffic control and optimization，the purpose of the smallest overflow ratio is achieved.Firstly，the general transfer function model used in drain pipes is derived based on Saint-Venant equations.Then，for the middle section of two flanking the node into the drains is further analysed，based on the above model calculation. Finally，MIKEURBAN software is applied to simulate，and a comparative analysis of the results between the model and the software is achieved.

Based on the Gelormino sewer LTI discrete model, the paper deduced a feasible generalized data model with the practical requirements, and proposed system identification method:(1)utilizing data mining technology to achieve clustering model structure,(2)applying correlation analysis to determine model structure, and(3)using least squares method and the fading memory recursive least squares method to achieve model parameter identification and on-line identification.The actual calculation result shows that the data model established by the above method can simulate and predict the inflow, the sewage and the water level changes, it can be used to guide the operation and management of urban sewage pumping system.

Parameter estimation of chemical processes can be presented as a tough optimization problem which can be solved with optimization methods.As a robust global searching method, genetic algorithm (GA) has been frequently applied in this area.However, the genetic algorithm has some shortcomings, such as weak local search ability and tends to premature.Furthermore, the encoding of GA cannot reflect the genetic information of biological organism.To overcome the deficiencies of GA, a new DNA genetic algorithm is proposed for the parameter estimation of chemical engineering processes.The proposed method uses nucleotide bases to present the individual. The novel crossover and mutation operators inspired by DNA molecular are designed.The novel crossover operators include permutation crossover operator and translocation crossover operator, and the novel mutation consists of anticodon mutation and maximum-minimum mutation.The solutions with two typical test functions show that the proposed method outperforms the other two methods in the searching speed, searching precision and the success rate.Finally, this method is applied to estimate the parameters of heavy oil thermal cracking model.The results of eight cross validation shows that the proposed algorithm possesses small self-check relative error, prediction relative error, and the standard deviation of relative error.Compared with the other two models, the model established by the proposed DNA genetic algorithm has smaller modeling error.

In petrochemical enterprises, material balance results with different spatial and time scales are used for tasks of different levels.In this paper, a rectification method and a material balance strategy based on a multilevel material model are proposed for multiscale material data.First, the engineering background and analysis of traditional methods are reviewed.Then the multiscale method is formulated and rectification steps from yield accounting scale to scheduling scale and equipment scale are discussed.Additional constraints with data from other scales contribute to redundancy of measurements and consistence of data used for different tasks.Generally, measurements on higher levels are more accurate in plants.So material balance based on rectified data is carried out from high levels to lower levels.This method contributes to the accuracy of reconciliation results and reduces difficulty with solving large scale problem.A simulation example which proves the effectiveness of the method is presented.

The estimation of kinetic parameters is an important topic for chemical process model application.Different optimization algorithms are used to estimate parameters for the eight lumps model of FCC（fluid catalytic cracking）process.It is shown that the particle swarm optimization（PSO）algorithm is simple and can be easily implemented.The PSO algorithm also exhibits a good global optimization performance that avoids the dependence on initial parameters.Furthermore a hybrid particle swarm optimization（HPSO）algorithm combined with Levenberg-Marquardt algorithm is proposed to improve the effect of parameter estimation.By use of real industrial data，the simulation results show that model prediction accuracy is ensured by HPSO method.

Dynamic model is the basis of dynamic optimization in chemical process.In this paper a dynamic model for esterification section of poly(ethylene-terephthalate) （PET） was developed using segment method.Different from other researches, the model considered the interaction between esterification reactor and the column.Dynamic responses of process operating condition were analyzed with basic control strategies.Simulation results show that properties of products such as carboxyl end concentration, degree of polymerization response quickly to step change of feed molar ratio, reactor temperature, pressure and liquid level.Also these step changes significantly affected the changes in gas flow of ethylene glycol and water.Control system plays a significant role in the stability of PET esterification process.

A new multi-swarm competitive particle swarm optimization(MSCPSO)algorithm is proposed, which avoids to immersing the local optimization and improves the global searching performance.Then radical basis functions neural network (RBFNN) is trained by proposed MSCPSO to model cracking productions of furnace for online predictions.At the same time, the intelligent optimal control method for cracking depth is studied integrated by MSCPSO-RBFNN.The optimal function is maximized the sum of ethylene and propylene yields.And then cracking depth which is satisfied to the optimal function is input the depth’s controller, which is linked into the advanced process control system of coil out temperature (COT), so the depth’s controlling is realized optimally.The applications are showed that the yields of ethylene and propylene are increased, and the depth’s control is more stable than before.The proposed optimal control method has good adaptability, stability and reliability.

For the problem of coupling in a class of multiple input multiple output（MIMO）systems，this paper proposes a dynamic nonlinear decoupling control method based on active disturbance rejection control（ADRC）.This method does not rely on the accurate model of the system.According to the partly known and unknown coupling matrix of the controller respectively，this method takes the model perturbation，the external disturbance and the dynamic coupling including the interaction of the input variables as a total disturbance to each channel based on the local static decoupling.By introducing virtual control and state variables，the extended state observer（ESO）is designed to estimate the total disturbance and then it is fed back to the controller to compensate the disturbances.And then，the non-linear SISO ADRC is designed to ensure the stability of the closed-loop system for each decoupling sub-object.Finally，the simulation results of a distillation column model control show that the designed controller not only has good decoupling performance，but also ensures good robustness and adaptability in the condition of modeling uncertainty and external disturbance.

Accurate and reliable prediction of polypropylene melt index is crucial in the control of propylene polymerization process, as well as in prompting the profit of the product.An optimization algorithm, the PSO_SA, based on particle swarm optimization (PSO) and simulated annealing (SA) is proposed, which uses the strength of PSO and SA to make up with the weaknesses of each other, and then the optimization capability and performance are improved.The PSO_SA is used to optimize the structure of the RBF neural network which is employed to predict the melt index of polypropylene.A research on the optimized model is carried out based on the data from a real plant and the model achieves a prediction accuracy of 0.75% in MRE.The result shows that the proposed approach has great prediction accuracy and reliability.

Dynamic systems with a mix of continuous and discrete components called hybrid systems frequently arise in chemical engineering applications.Information structure of those systems consists of discrete events and continuous variables.Since many of these applications are safety critical，it is important to use reliable methods to simulate hybrid systems.In this paper，a formal design method based on duration calculi to study modeling and design of such systems with integrated qualitative/quantitative performance is proposed.It is shown that duration calculi represents a very powerful tool in the analysis and design of simple hybrid control systems.The duration calculi can be used to capture and define interval temporal logic requirements for hybrid systems，to define qualitative/quantitative behavior and semantics of hybrid control systems.It also offers the formal description of the behavior of hybrid systems，the formal description of optimizing of hybrid systems and formal methods，and the formal description of design steps based on duration calculi.Finally，an example shows that the design method is effective.

Coal blending optimization is of great significance for the optimal operation of Texaco coal gasification process.A process model from a management and decision-making perspective is constructed to solve the problem of Texaco coal blending optimization.The model takes into account mixed-coal indicators, inventory costs, market prices, operating costs and consumptions of stockpiling and transit.The model is calculated with particle swarm optimization with prior crossover differential evolution(PSOPDE), which can avoid prematurity more easily than basic PSO and DE, and has superior features in solution accuracy and efficiency. The simulation results of a coal blending optimal process of a fertilizer plant validate the feasibility of the model and algorithms.

An optimal control model is presented for determining the best injection strategies of polymer flooding in enhanced oil recovery.The performance criterion of the optimal control problem（OCP）is the profits gained from oil recovered over a given time，which subjected to the nonlinear partial differential equations of porous media flow，integral inequality constraints and boundary constraints of control variables.The adjoint problem of the OCP and the gradient of the objective functional are derived by using the necessary conditions of optimal control for a 2-D distributed parameter system.A gradient based method is given for solving the OCP numerically and the results of a study case illustrate the effectiveness of the proposed method.

Based on active set SQP quasi-sequential approach，a quasi-sequential approach based on primal-dual interior-point method was presented.Quasi-sequential approach included two layers called simulation layer and optimization layer.In the simulation layer orthogonal collocation method was used to discretize both state variables and control variables，and the discretized DAE system was solved at each NLP iteration to eliminate equality constraints and state variables，so that the optimization problem is reduced to a smaller NLP problem only with inequality constraints and control variables.Recent studies showed that interior-point method took advantage of active set method in large-scale optimization problems，thus in the optimization layer a primal-dual interior-point method is employed.FORTRAN is used to code quasi-sequential approach and a distillation optimal control problem is optimized to demonstrate the efficiency.Results show that this approach has the capacity to solve large-scale dynamic optimization problems

According to a type of flow shop scheduling problem with multi-customer participating, this paper researches that customers establish coalition by cooperation, and total costs of customers can be saved by rearrangement of processing tasks.Cooperation games of general flow shop scheduling problem is restricted, a type of flow shop scheduling problem on processing time associated with workstage is presented, processing time of all processing tasks on same workstage is equal, corresponding cooperation games are balanced and have a nonempty core.Starting from cooperative game theory, a model of flow shop scheduling on processing time associated with workstage based on cooperative games is build, processing tasks are rearranged according to optimizing linear cost of customers.After getting optimum scheduling sequence, a kind of cost allocation method based on weighted marginal cost of predecessors and followers is put forward, it is also proved that this method gives a core allocation of cooperative game on this type of flow shop scheduling.At last, this type of flow shop scheduling model based on cooperative games and cost allocation method proposed in this paper are tested and verified.

A consensus-based decentralized reliability prediction method based on particle filtering is proposed for a nonlinear dynamic system with hidden degradation process.The degradation process is hidden means that the degradation cannot be measured directly.In this method，the original system is divided into several interconnected subsystems with overlapping decompositions，and then using the particle filtering algorithm at each subsystem，the states and parameters of the original system are estimated by each subsystem with the information of local measurements and the communications between subsystems，subsequently reliability of original system can also be predicted.Simultaneously，the method makes consensus-based estimation and prediction results for all subsystems with consensus strategy.Finally，from simulation experiments of a five-tank vessels model the degradation of system reliability can be predicted well，as well as the estimation and prediction results of all subsystems can be consensus.

It is important to analyze the fault propagation mechanism of large scale automatic system which is comprised of many tightly connected subsystems.Most existed fault propagation analysis methods are bothered by “knowledge explosion” problem，which is an obstacle for the application of these methods.A knowledge-guided data driven fault propagation analysis method is proposed in this paper. Firstly，by using correlation analysis，it is proved that correlation between the designated component（DC）of faults occurred in input and output system can tell some fault propagation information.Secondly，the fault propagation relation matrix is determined by comparing the correlation between input DC and output DC of typical faulty data. The main criterion is to set the corresponding element in the fault propagation relation matrix to be 1 when the correlation coefficient between input DC and output DC is larger.Thirdly，according to the sampling data of input DC and output DC of the case when a common fault is occurred，a DC regress model is established to predict the fault imperil level for the output system.Finally，simulation study shows its efficiency for knowledge-guided data driven fault propagation analysis method.

Since the traditional tolerant control methods can not guarantee the stability of the nonlinear systems with concurrent actuator and sensor faults，the robust tolerant control possessing integrity for a class of time-delay uncertain fuzzy systems is discussed in this paper.The uncertain nonlinear systems are described by the T-S fuzzy model，and the standard and normalized fault matrices are defined in advance，then based on the system structure transformation by Newton-Leibniz formula，the delay dependent sufficient conditions on the existence of the robust tolerant controllers are derived by linear matrix inequalities，so that the closed-loop systems are stable even when the actuator and（or）the sensor faults occur，and the generalized robust performance also can be satisfied to restrain the influence of the disturbances，the initial and delayed states.Finally，the simulation results indicate the feasibility and the validity of the proposed method.

In the processing of state，actuator and sensor fault detection and isolation，the last is the most difficult.In the under-measurement system and the sensor redundancy system，this problem becomes more difficult.Firstly，three typical mathematical models of sensor fault containing lock-in-place，gain time-varying and bias time-varying are established.Secondly，the output equation to converted sensor fault into state disturbance of the dynamic system is reported.The problem of converted result is processed with least-square solutions and the basic solutions in the under-measurement system.For the sensor redundancy system，two parallel residual generators are constructed using twice system state.Then，the residual generator which can well be used to fault detection and isolation is designed.Finally，the effectiveness of this method is verified through example simulation.

The on-line crude oil blend scheduling is important to support the implementation of advanced control and optimize the refining plan,which is so complicated that some new efficient algorithm is required.In practice, there may be one kind of primal crude oil blended with one of the others in sequence, or some two kinds of crude oil be selected sequentially to be blended, say no primal crude oil.For the case of no primal crude oil, it is hard to match the crude oils because the match is determined by the sequence and the blending ratio together.A novel description and its construction algorithm are presented, then the problem is solved by the optimization scheme based on order for its double-layer structure.By the simulation on the practical data of crude oil distillation, it is shown that the algorithm based on order can improve the computational efficiency significantly.

Aiming at the features of having multi-disturbance and real time scheduling, framework of molten iron dynamic scheduling system based on two-stage case-based reasoning（CBR） is proposed according to the indices of molten iron oriented iron-steel correspondence, such as time, mass, composition, temperature and so on.In terms of characteristic of various disturbance of production site, two-stage case library is classified.Key technology of molten iron dynamic scheduling system including case representation, case retrieving, case adaptation and case maintenance is discussed.And the satisfaction evaluation indices of dynamic scheduling are computed.The experimental results using a lot of practical production data show molten iron dynamic scheduling system has velocity on reasoning, higher stability and validity on results.The technical methods of two-stage case-based reasoning have wide application prospects.

A high-precision on-line method of predicting melt index of propylene polymerization based on principal component analysis (PCA) and adaptive particle swarm optimization (APSO) is proposed to overcome the high correlation characteristics and high nonlinear characteristics in the propylene polymerization process.APSO is employed to get better search efficiency and higher precision than classical particle swarm optimization (PSO), and PCA is applied to reduce the complexity of the statistical model.A new method of optimizing both structure and parameters of radial basis function (RBF) network is also proposed.The validity of these methods is demonstrated through practical data in real factory, and research result shows higher precision and shorter computing time than before.

The estimation of kerosene dry point is important for quality control in process industry.Since the crude oil changes with varying compositions, there does not exist any effective model to cover different inherent characteristics.In this paper,adaptive hinging hyperplane (AHH), a novel piecewise linear method is introduced to build soft sensor model.A dry point estimation simulating experiment shows the good result.And the test on practical data reveals that AHH method has a better performance than the existing models based on classification information.It is expected that the method can be extended to other chemical process estimation with varying situations.

Considering the performance of a predictive model is heavily depended on its training samples, a new on-line adaptive modeling approach based on incremental support vector regression (SVR) is presented.When a new sample arrives, it is firstly checked by the Karush-Kuhn-Tucker(KKT) condition of established model, only those which contain sufficient new information can be introduced into the training sample set.In this way, the model generalization ability will be maintained while the update frequency can be reduced.If the new sample cannot be described by the established model and, therefore, has a large prediction error, the model must be updated and the useless sample should be deleted from the model, to adapt the process characteristics.In this case, the useless sample, while not the oldest one, is selectively deleted from the model based on the similarity between samples.The proposed method is illustrated through the application to an industrial polypropylene unit to predict its melt index.The results show that, compared with other methods, the proposed method exhibits good generalization ability while the update frequency significantly lower, and therefore trace the process characteristics effectively.

It is very important for ethylene pyrolysis process to obtain product yields on line.To address the problem with few valid sampling data, soft-sensor models of several kinds of product yields were developed based on support vector regression (SVR).Particle swam optimization (PSO) algorithm was used to determine the proper parameters of SVR model, and model efficiency and performance were then improved.SVR based product yield models got high accuracy and good trend tracking performance on the real industrial data.

Cultural algorithm is a new computational framework.The framework depicts cultural evolution as a process of dual inheritance from both population space and belief space.During the evolution process of cultural algorithm, the effective implicit information is abstracted and utilized.This paper presents an integrated intelligent optimization (IIO) to solve large dimension non-linear optimization problems.Genetic algorithm (GA) and particle swarm optimization (PSO) are used in the framework of cultural algorithm.Integrated with neural networks, an integrated intelligent optimization neural network model is proposed.Then IIO is applied to optimize parameters of neural network (NN) in soft-sensor modelling of inlet ammonia content of ammonia synthetic tower.The results show that the model based IIO-NN has more precision and better performance than the model based on BP-NN, GA-NN and PSO-NN.

Currently the evaluation of energy consumption of ethylene equipments does not consider the difference of raw materials, process technology and equipment size among ethylene units, and the statistical method is not uniform, so energy consumption indices are not unfair for different ethylene equipments.This paper proposes hierarchical linear optimal fusion algorithm to extract energy indices according to technologies, scales and data distribution of energy consumption including fuel, power, steam and water.The proposed algorithm has no strict limitation for the length and distribution of data, and can remove abnormal consumption data.Quality of data fusion can be ensured by optimal weights which are calculated by covariance.The validity of algorithm has been verified by actual application, which is obtained monthly and annual energy consumption indices from both various type plants and the ethylene industry.Such indices are helpful for decision-maker to identify the quantitative energy consumption targets and major factors, and the proposed method can also be used to other process units for evaluating energy consumption.

Based on earlier experiments, this paper compares characteristics of coal combustion and pollutant formation under O₂/CO₂ atmosphere with conventional air by numerical simulation.It focuses particularly on the combustion fundaments， such as flame temperature, heat transfer, and NO_x formations.The results showed that coal combustion under O2/CO2 atmosphere excelled combustion under conventional air.And the obtained data are necessary for reasonably organizing flow field, controlling flame temperature and adopting appropriate O₂ concentration, which help to design new oxy-fuel combustor.

Distillation consumes one third of the total energy in industry.It has great potential for energy saving.Distillation energy-saving technology in internal thermally coupled distillation is the most promising, which can save more than 40% energy compared with traditional distillation process, but it has not been widely used.The bottleneck that prevents the process from being commercialized is the operational difficulties due to the nonlinearity, complex dynamics and interactive nature of the process.Based on an accurate second-order model, an internal model control (IMC) system for ITCDIC is presented and performance evaluation shows its effectiveness.It is more stable and reliable than PID.Also, compared with IMC reported, it has a wider range of operating field and more robust.

A general mixture signal model (MSM) together with support vector data description (SVDD) are proposed to address the monitoring of non-Gaussian processes.Mixture signal model involves Gaussian, non-Gaussian and measurements noises.Methods to extract and monitor the corresponding mixture signals are presented.A general SVDD kernel function parameterization and optimization approach is proposed to monitor the non-Gaussian signal sources.Industrial application demonstrate that the general proposed kernel function is capable of characterizing the non-Gaussian behaviors encapsulated in process data and detect abnormal events promptly.

Raw material conveying and proportioning are important link in ethyl acetate production.In these parts，high-performance motor drive mechanisms are needed which are used to provide high-precise speed and stable toque output，especially in the low frequency situation.Meanwhile it is required that the drive mechanisms should also provide parameters like current，torque and other analog as well as digital signal to distributed control system（DCS）or other production control systems.It is hard to meet the requirements by using common inverter.Accordingly，a new motor diver system was designed.The nonlinear，strong coupling characteristics were founded by analyzing the math model of AC induction motor.Because it was difficult to control well by using common method，Clarke’s and Park’s transform was used for decoupling，a field oriented control algorithm was introduced，a current-speed double closed-loop scheme was proposed,and software was written based on digital signal processor(DSP),and finally the inverter was developed.The device was used in the ethyl acetate production process.From the results of the present production，it is shown that the device has high control accuracy，dynamic response over wide frequency and stable operation.The general performance of that is better than normal V/F，and it has a certain value for popularization.

An adaptive control method based on adaptive-network-based fuzzy inference system（ANFIS）and multiple models is proposed for a class of uncertain discrete-time nonlinear systems with unstable zero-dynamics.The method is composed of a linear robust adaptive controller，a nonlinear adaptive controller using ANFIS and a switching mechanism.The linear controller ensures the boundedness of the input and output signals，and the nonlinear controller improves performance of the system.By using the switching scheme，both improved performance and stability are achieved simultaneously.When the ANFIS is used as a compensator for the unmodelled dynamics，firstly，a continuous，monotonic and invertible one-to-one mapping is used in this paper to transfer the non-compact definition domain of the unmodelled dynamics into a bounded closed set.As a result，the universal approximation property of ANFIS can be guaranteed.Furthermore，the ANFIS can successfully tackle the relatively low convergence rate of neural network and avoid the possibility that the network becomes trapped in local minima，thereby improving the control effect.A lemma is established and the analysis of stability and convergence of the closed-loop system are proven theoretically.Last，by comparing the simulation results，the effectiveness of the proposed method is illustrated.

Due to the complexity and high non-linearity of microbial fermentation process, it is difficult to model the process of fermentation precisely.Although there is a predictive error between traditional mechanistic model and testing result, it reflects the process of mechanism.Neural network modeling method is classified as “black box” method: the process of modeling uses no priori knowledge, and gets a good result, so single soft sensor modeling method can not have the advantages of the others.This paper uses RBF neural network and mechanistic model to model and constitutes a “cinder box” hybrid model, which is a good modeling method.

The accuracy of grey Verhulst model greatly depends on background value and model structure parameter. In this paper, an improved-grey Verhulst model is proposed.Firstly, the cause of grey Verhulst model’s inaccuracy is analyzed.Secondly, the background value is rebuilt by using original data sequence.Thirdly, model structure parameter is obtained by optimization based on the principle about information overlap of grey system.Finally, the improved-grey Verhulst modeling is applied for building the model of national production of crude oil and results show its effectiveness.

Due to heavy calculation burden for solving Diophantine’s equation and inverse matrix，the traditional GPC （generalized predictive control）algorithm is unsuitable in some areas where both small system overshoot and fast track speed are required.In this paper，an improved generalized predictive control algorithm for fast restraining overshoot is proposed.Firstly，the calculation of inverse matrix is not necessary by softening coefficient matrix to obtain smoothing current input.Secondly，the next input increment is estimated by simple GPC.Finally，the current input is combined by next input increment in order to overcome potential overshoot.The simulation results show the effectiveness of the algorithm.

Sontag-type constructive predictive control scheme is an efficient algorithm to reduce the on-line computational load of constrained nonlinear predictive control and to decouple the stability of closed-loop systems from the optimality of performance indexes.By employing inverse optimized control theory，the issues of inverse optimality and robustness of the controller are discussed for continuous-time，constrained nonlinear systems.Then the sufficient condition of robustness of the controller is established under the condition that the scheme is nominal stability.This will help to understand the control algorithm and provide the theoretical basis for its application.Finally，a simulation example is used to demonstrate the effectiveness of the result in the paper.

Effective controlling of the endpoint steel temperature and contents of carbon, sulphur, etc.is one of the main tasks of BOF steelmaking process.A multivariable neural network model was established in this paper.The input data were pretreated and standardized.Rolling optimal control method was used to increase the accuracy of the model.Simulation and experiment comparisons show that the model is validated and has high hit rate.

It is difficult to establish the precise mathematical model of fermentation temperature，for the complex reaction mechanism of the beer fermentation process.Fermentation temperature plant is a large time-delay system，some typical algorithms for this kind of systems can not act a good effort as the precise model for this process can not be got.It will be more difficult to guarantee the effort of the controller while the interference in the fermentation process is unpredictable.Aiming at these problems，a new predictive control strategy which can quickly suppress disturbances has been provided by this paper.The strategy is based on the idea of single step control of dynamic matrix control（DMC），and combines the advantages of both time optimal control and DMC.It is proved by the simulation result that this strategy acts a good effect on the fermentation temperature control and will quickly suppress the interference.

To reduce the computational complexity of the dynamic optimization section in model predictive control algorithm，a new method based on linear programming instead of quadratic programming was proposed.For both single-input single-output and multi-input multi-output model predictive control，the linear programming problem to describe the dynamic optimization is constructed by treating control increment，output increments and some deviation variables as optimization variables，inducing equality or inequality linear constraints and choosing a linear objective function.Moreover，soft constraints can be considered as a part of the linear programming problem to improve the performance indicators of dynamic process and to achieve the purpose of smooth control.Finally，a simulation example illustrates the effectiveness of the presented approach.

Shortcomings of the power reaching law was discussed.An improved discrete reaching law is deduced，which makes up the deficiency in the discrete system based on single power reaching law，and a new discrete variable structure controller is obtained.The variable structure control system designed by using this new controller can gradually approach to zero.By analysing chattering of the discrete reaching law，it is found that the system chattering is decreased and fast reaching speed is kept.Simulation and analysis results both show that the effectiveness and feasibility of the proposed method，which can make the system keep continuity and eliminate system chattering effectively，and ensure the system asymptotic stability.

A control strategy，in which PID controller is adopted in inner-loop and self-tuning fuzzy predictive functional control algorithm is used in outer-loop，is proposed for cascade time-delay system.The internal loop and the main plant constitute generalized control plant which is controlled by the sum of inputs produced respectively by predictive functional controller and self-tuning fuzzy controller.The strategy is applied in main steam temperature of thermal plant.Simulation results show that the proposed strategy has better control performances compared with other control algorithm.

In the iron ore sintering process，to optimize two key parameters，the burning through points and the bunker-level，an multi-objective optimal control method is presented based on fuzzy satisfactory degree.First，an intelligent optimal control model of the burning through points is established by using fuzzy control，predictive control and switching control strategy.Next，an expert control model is designed for the bunker-lever.Finally，the satisfactory-degree function of the system is designed to obtain satisfactory solutions，which transforms the multi-objective optimization problems into the single-objective optimization problems and eventually results in multi-objective control of the sintering process.Results of simulation and actual runs show that the proposed method is feasible and effective.

An intelligent cooperative decoupling controller（ICDC）is presented and applied to the liquid-level-concentration control in the water bath stretching slot of the filament production.It consists of a control center，several control and decoupling units and their corresponding output units.The control center coordinates each control and decoupling unit which takes effect independently.The system information is exchanged among them to make decoupling.The control signals are integrated by the output unit and then sent to the plant.Simulation results demonstrate that the proposed ICDC can rapidly response to the variation of control variables，completely eliminate the coupling influence and make smooth regulation without overshoot，which has better performance than that of the conventional schemes.

To formulate the fuzzy programming with piecewise linear membership function，a novel modeling methodology is proposed.By means of compatible configuration of binary variables，the relationship between the sub-membership functions is explicitly characterized，which helps the fuzzy optimization efficiently prevent the constraint-free problem happened in the general modeling methodologies，achieving satisfied optimal solutions.Additionally，compared with the conventional modeling methodologies，fewer binary items are required to model the membership function in some scenarios，possibly reducing the complexity of the calculations.In this paper，two methods consistent with this approach are introduced in detail，along with a numerical example employed to demonstrate the benefits of the contribution.

Aiming at the control problems of air-cushioned headbox in papermaking process，such as the nonlinearity and serious coupling relationship between total pressure and liquid level，a double variable PID decoupling controller based on BP neural network is proposed to release the influence between the two variables.The controller changes the parameters：proportion（k_p），integral（k_i），and differential coefficient（k_d）of PID itself on line with neural network’s capability of non-linear description，in order to find the best PID parameters，which can release the influence between total pressure and liquid level of the air-cushioned headbox.The results of simulation on air-cushioned headbox’s linear and non-linear model indicate that the controller has the characteristics of simple realization，quick dynamic behavior，high control effect and better practical value.

A multiple constrained generalized predictive control algorithm for cascade industrial systems is presented by considering fully of the amplitude constraints of inputs，the amplitude constraints of input increments，the amplitude constraints of internal variables，the amplitude constraints of outputs and the constraints of output errors.It uses internal variables for prediction and feedback.The simulation results show that the control algorithm is robust and works effectively for multiple constrained cascade industrial systems.