Considering the cluster effect of industrial park, it is expectable that could improve energy efficiency of the entire park by recovering the waste heat in each single plant via inter-plant medium fluids which absorb heat in some plants and release heat into other plants. Obviously, the selection of heat exchange medium and the setting of medium heat recovery loop will highly affect the optimal design of the waste heat recovery system and the energy-saving effect. Thus, other than single medium integration strategy, this work uses both hot water and thermal oil as mediums to implement the inter-plant waste heat integration. A heat exchanger network (HEN) superstructure coupling the matches between medium fluids and inner-plant process streams and the allocation of medium streams across plants is proposed. Accordingly, a mixed-integer non-linear programming (MINLP) model is formulated for network optimization aiming at the target of minimum total annual cost. At last, an example involving three plants is studied in three synthesis cases (single medium-single loop, single medium-dual loops, and dual mediums-dual loops). The effectiveness of the proposed method was verified by comparison.

For a quaternary reaction separation system having the most unfavorable relative volatility order (i.e., the reactant is the lightest and heaviest component, the product is the intermediate component), an external is introduced between the top and bottom of the reactive distillation column. The seperation for an hypothetic ideal reversible exothermic reaction was taken as an illustrative example to study the problem of the decentralized control system designs. The results show that the reactive distillation columns with a top-bottom external recycle has better closed-loop control performance and stronger controll ability in the face of large disturbances as compared to the conventianal reactive distillation columns due to the fact that the introducing of external recycle enhance the system reaction rate. In addition, the processes have more operating valuable (i.e., the external recycle flow rate) compared to conventianal reactive distillation columns; therefore, one can utilize the extra operating valuable to control the product purity. The unique control system can further enhance process closed-loop control performance.

The ethylene cracking furnace equipped with the bottom burner and the side wall burner is more and more widely used. Different combustion modes affect the turbulent flow state in the furnace. Considering the turbulent flow in the cracking furnace and the gas injection, the combustion and heat transfer are strong. The nonlinear coupling effect, for this purpose, explores the influence of different turbulence models in the cracking furnace/reactor coupling simulation is critical for the precise design and optimization of the cracking furnace. In this paper, a coupling simulation for a 100000 t industrial ethylene cracking furnace was carried out for different turbulence models. The turbulent flow model established by the standard k-ε model, RNG k-ε and Realizable k-ε model was evaluated by CFD numerical simulation. The simulation results of the three turbulence models are compared with the industrial data. The distribution of velocity, temperature and turbulence capacity in the cracking furnace is analyzed. The results show that the Realizable k-ε model is superior to the other two models in flame stability and reaction efficiency. And based on the Realizable k-ε turbulence equation, the calculation results of the heat flux and the outer wall temperature distribution of the furnace tube are closer to the actual working conditions.

There are many safety threats in the chemical industry park, such as dangerous goods storage tanks and transport vehicles. The danger situation in the park need to be sensed in real time and potential safety threats must be discovered and eliminated in time. The traditional method relies on a single data source such as real-time monitoring of dangerous goods storage tanks for hazard identification, which is difficult to meet the current needs of the chemical park for safety status assessment. From the point of view of big data analysis, this paper integrates the data of dangerous goods storage tank sensors, dangerous goods transportation(DGT) and geographic information in the chemical park. Based on the characteristics of Gaussian diffusion of dangerous goods leakage, a multi-source heterogeneous data fusion method is proposed. The danger situation identification method realizes the dangerous situation awareness of the park and displays in real time the potential dangerous areas in the entire chemical park. Combined with the actual data of a chemical park, the effectiveness of the proposed method is verified.

Knowledge generation is the basis of industrial knowledge automation. Process industry knowledge is usually generated by directed tasks, such as optimization scheduling, optimization operation, fault diagnosis, etc. The solution generation requires not only understanding the operation mechanism and production data, but also relying on domain expert experience. Such forms of knowledge representation are difficult to unify, poorly ported, and inconvenient to share and reuse. Aiming at the problem of resource scheduling optimization of ethylene cracking furnace group, the P-graph method is used to construct the superstructure model of the solution, which is designed to represent the knowledge of P-graph ontology and database mapping into knowledge rules, and automatically generate RDF (resource description framework) represents the solution knowledge and builds a knowledge repository. Finally, the actual production data of the ethylene production plant was used to verify the feasibility and practicability of the proposed method.

The chemical production processes are increasingly complex and the traditional extreme learning machine (ELM) cannot effectively model the chemical processes data. To tackle this problem, a novel PCE-RELM model based on principal components extraction (PCE) is proposed. Through principal component analysis of the ELM hidden layer, the principal component features of the data are extracted, the linear correlation between variables is removed, and the research problem is simplified. The influence of the number of hidden layer nodes on the accuracy of the model can be reduced, the number of hidden layer nodes in the ELM can be quickly and randomly selected, and the ELM model becomes robust. To verify the effectiveness of the proposed method, the PCE-RELM model was applied to modeling the purified terephthalic acid (PTA) production process. The simulation results show that, compared with the traditional ELM, the PCE-RELM model has the advantages of simple design, good robustness and high accuracy, which can guide the chemical process control and analysis.

Industrial production equipment usually sets sensor alarm thresholds for alarms, but it is difficult to capture time series abnormalities below the alarm thresholds. The traditional statistics based detection method has great challenges in these time series anomaly detection. In this paper, an approach to the anomaly detection of process units is proposed by using the long short term memory (LSTM) time series reconstruction. At first, an LSTM network is introduced to vectorize the time series of sensor data, and another LSTM network is utilized to reconstruct the time series in reverse sequence. Then, the errors between the reconstructed values and the actual values are used to estimate the anomaly probability by the maximum likelihood estimation method. Eventually, anormaly detection is achieved by learning the abnormal alarm thresholds. Simulation resutls on the ECG standard testing data, energy data and sensor data of the dangerous goods tank have shown the effectiveness of the proposed method on data with different lengths.

The generality of the CFD-PBM coupled model was studied by comparing the simulation results with experimental data under different operating pressures and liquid properties. The results show that the CFD-PBM coupled model with the modified pressure factor obtained from the internal-flow bubble breakup model can well predict the influence trend of pressure on the hydrodynamics of bubble column. The gas holdup increases significantly with increasing pressure. In addition, the simulation results for different liquid viscosity and surface tension are consistent with the experimental results. With increasing liquid viscosity, the bubble breakup rate decreases, which leads to a wider bubble size distribution, lower drag correction factor and decreased gas holdup. As the surface tension decreases, the bubble breakup rate increases, which results in smaller bubbles and higher gas holdup. The CFD-PBM coupling model has good versatility because it considers the effects of pressure, liquid viscosity and surface tension on bubble coalescence, fracture and gas-liquid interaction.

The mathematical model of the acetylene hydrogenation reactor established by traditional single modeling method does not meet the needs of industrial practical applications in predictive performance. This paper proposes a mechanism and neural network nesting modeling method, which fully utilizes the mechanism model. It makes full use of mass and energy balance information in mechanism model to reduce the degree of constraint violation of the neural network model, which can describe the process characteristics of industrial reactor well. The optimization problem which targets the operational profits as the objective function is studied basing on the hybrid model. The main decision variables include several key parameters, such as the reactor feed hydrogen-alkyne ratio, the feed temperature, and the two-stage reactor operating cycle and many more. For the long-term operation of the reactor, processing capacity of the reactor will decrease due to the decreased catalyst activity, and an improved optimizing strategy is proposed by adjusting the hydrogen-alkyne ratio as well as the reaction temperature simultaneously. The sequence method is used to discretize the operating cycle of the reactor. The two-stage difference algorithm is improved by introducing differential mutation strategy and potential solution alternative strategy. Then the optimization problem is solved by combining the incremental coding method with the improved two-stage difference algorithm. And the results confirm the effectiveness. Furthermore the optimal operating cycle and operating strategy of the reactor are given.

To improve the energy integration and equipment sharing level of LNG, a conceptual design of natural gas liquefaction system based on large-scale AP-X^TM liquefaction process and integrated gas subcooling technology (GSP) integrated natural gas condensate (NGL) recovery process was proposed. The performance of the multi-stream heat exchanger, the unit power consumption and the recovery rate of ethane were considered as the three of basic characteristics to evaluate process performance. The simulation and analysis results show that the unit power consumption of proposed process is reduced to 0.45 kW·h·(kg LNG)^-1 which is reduced by 6% compared with conventional independent process. Furthermore, recovery rate of ethane is 93% which prove that NGL s efficient separation is achieved. The thermodynamic analysis, exergy analysis and economic analysis prove that the proposed configuration has high thermodynamics performance and economic value. This study can provide guidance for natural gas engineering research and retrofitted design of natural gas liquefaction technology.

In addition to concentration of contaminant, stream properties such as toxicity, pH, chemical oxygen demand are also important characters indicating water quality. Such situation would make the design of water-using process just according to contaminant concentration fail to meet the increasingly stringent production and environmental requirements, and prompt it to be necessary to consider the simultaneous integration of stream properties in water network synthesis. In this paper, a superstructure involving property interceptors in semi-continuous operation is established for the integration of batch water network with concerning environmental constraints and the minimum total annual cost target. The property interceptors can operate at various treating rates in different time intervals, and a series of tanks are placed before and after the interceptors to satisfy the batch operation of process sources and sinks, in assistance of which the sources in the tanks in front of interceptors are allowed to be reused to sinks directly without treated by interceptors. The calculation results show that the proposed method can effectively reduce the total annual cost and reduce the number of interceptors, which verifies the effectiveness and superiority of the proposed method.

A key step in the design of polymer molecules is to obtain repeating unit structures that meet a variety of properties. As a new developing method in chemical engineering, computer-aided molecular design (CAMD) technique can generate the repeat unit structures which satisfy the constraints using group contribution method, molecular dynamics (MD) technique can be used to simulate computer experiments to acquire systems properties at the micro level. This paper establishes a general CAMD-MD method to design polymers. First, repeat unit structures are identified based on group contribution method. Second, the weight of properties is determined respectively by using the analytic hierarchy process and properties of candidate structures are simulated based on molecular dynamics method. The CAMD-MD method is finally applied to the actual rubber structure, and the properties such as cohesive energy density, density, glass transition temperature and thermal conductivity are simulated to verify the feasibility of the method.

Reaction solvents have been widely used in liquid-liquid homogeneous organic synthesis. They have significant impacts on reaction rates and selectivity, which have contributed to the development of new process route for green synthesis. A computer-aided molecular design (CAMD) reaction solvent design method based on Dragon descriptor and SMILES (simplified molecular-input line-entry system) coding is proposed. First, a reaction kinetic model was constructed to make quantitative predictions for reaction rate constants k by the decision tree-genetic algorithm (DT-GA). Then, through SMILES code techniques and Dragon software, computer-aided molecular design (CAMD) method was integrated with the DT-GA to establish a mixed integer nonlinear programming (MINLP) model consists of objective functions and constraint equations. Afterwards, a decomposition-based algorithm was employed to solve this MINLP optimization problem, which achieves the objective of reaction solvent design. Finally, an example of Diels-Alder reaction was adapted to demonstrate the feasibility and effectiveness of this method.

A hybrid fault-tolerant controller with infinitely adjustable time-domain parameters is designed to ensure fault-tolerant control performance. Firstly a multi-phase state space model by acquiring input and output data is established, and further the state space model is into transformed an extended state space model containing state variables and output tracking errors, and the switched system model is used to represent it, so as to design the controller in an infinite horizon. Then, to obtain the minimum running time, the dwell time method depending on the Lyapunov function is proposed for different stages. Finally, taking the injection molding process as an example, the system simulation is carried out. The simulation shows that the proposed method is feasible and effective.

The ethylene cracking furnace model based on BP and RBF ignores the reaction mechanism of cracking furnace and has the disadvantage of large prediction error. Therefore, a two-way long-term time-memory network (BLSTM) model based on the reaction mechanism of ethylene cracking furnace to forecast key parameters such as ethylene yield is proposed. To solve the problem that BLSTM modeling lacks available data, a BLSTM model using cross iteration (CIBLSTM) is provided. The CIBLSTM model uses a forward-reverse cross-iteration method to gradually approximate the true value of the lacks available data, and then a prediction model is established for the ethylene cracking furnace. To verify the validity of the proposed CIBLSTM model, nine industrial actual raw materials and analytical data were selected for simulation test. The simulation results verify the validity and practicability of the proposed CIBLSTM model. The proposed method can also be applied to other complex chemical processes modeling.

There are differences in equipment and technology between different cracking devices in the ethylene industry. There is also a difference in the yield and energy consumption of ethylene products from cracking devices of different furnace types or processes in each ethylene feedstock plant. With the commissioning and starting of the new ethylene plant, it is necessary to simultaneously operate a large number of differential cracking devices, thereby providing space for the optimization of ethylene cracking raw materials to achieve improved material efficiency and lower energy consumption. This paper proposes a modeling and optimization method based on P-graph for the scheduling of raw materials and energy optimization among such plants(SGBP). This algorithm extracts the structural information of the process by P-graph itself, and preserves the suboptimal solution set while accelerating the solution. Afterwards, using a certain ethylene plant as an example, the proposed method was applied to achieve the modeling and optimization of raw material scheduling. The advantages of the proposed method has been verified via comparing with MINLP and one kind of intelligent optimization algorithm. It can provide simultaneously more abundant optimal solution and suboptimal solution. The optimal result of the proposed method is equivalent to the optimization effect of MINLP. The overall energy consumption after optimization is significantly reduced, and a variety of alternative operation options can be provided for the production plan personnel to choose flexible raw material deployment plans.

Because some raw material consumption in complex chemical production process is difficult to measure directly, a soft sensor method based on the deep learning is proposed. Based on a period of the historical data, the proposed method extracts multi-scale information from the historical data using stationary wavelet decomposition. Then the observable data at every point of time are combined to get a complete dataset which is divided into the training dataset and the testing dataset. Moreover, the soft sensor model is trained and obtained by using the depth learning algorithm based on the attention mechanism. Finally, the proposed method is applied to the soft measurement of acetic acid consumption in a terephthalic acid (PTA) production unit. Compared with the extreme learning machine (ELM), multi-layer perceptron (MLP) and common long short-term memory (LSTM) method, the result analysis shows that the validity and the applicability of the proposed model is verified. Meanwhile, the consumption of acetic acid in the PTA production plant is predicted and analyzed to improve the production capacity and reduce the energy consumption.

The petroleum drilling project is a high-risk and costly system project. To effectively scan for potential problems of drilling, reduce non-productive time and lower related risks, this paper proposes an improved intelligent warning method based on moving window sparse principal component analysis (MWSPCA) integrating case-based reasoning (CBR) (MWSPCA-CBR). First, the MWSPCA is used to analyze the real-time data in the drilling process, and the time of occurrence of the anomaly is quickly located. Then the abnormal data is analyzed by using the CBR method to give possible exception types, and the associated handling methods are provided for monitoring experts. Finally, the proposed method is applied to intelligent warn abnormal problems of the petroleum drilling, the experimental results verify the feasibility and effectiveness of the proposed method and provide new ideas for reducing risks and costs during the petroleum drilling process.

In the modern process industry process, DCS collects and stores a large amount of operational temporal data. If valuable operational experience and operational information can be extracted, the performance of the operating system can be greatly improved. However, operational experience is vague and cannot be quantified by value. Therefore, the operational data with time series is symbolized so that the operational experience is represented in a block form. And we propose a hierarchical clustering algorithm based on Levenshtein distance for time series. By clustering of historical operational data in the time series of variables, a variety of similar operating modes are obtained, and the process variables corresponding to the type of operation mode perform performance analysis to obtain and preserve the operational experience required in the actual work process, thereby guiding the process operation of production. In order to verify the proposed method, it is applied to the continuous multi component distillation operation process. The results show the effectiveness of the proposed method.

Complex industrial process has long processes, many system levels and a wide range of potential fault distribution space, which is a hotspot in the current fault diagnosis field. Firstly, the current mainstream fault diagnosis methods in process monitoring are classified and summarized. Secondly, this study adopts the combination of quantitative and qualitative, which is based on data and knowledge. A system-level holographic fault diagnosis framework for complex industrial process is proposed, which provides a complete set of techniques and solutions for process monitoring in complex industrial plant-wide process. Compared with current fault diagnosis methods, the framework not only includes fault detection and fault identification, but also includes root cause diagnosis, fault propagation path identification, quantitative fault diagnosis and evaluation. It is quite a practical method for process systems, which can effectively reduce or avoid the fault occurrence, guarantee the quality of the product, and improve the production efficiency and safety of enterprises. Finally, the development trend of fault diagnosis technology and problems to be solved are prospected.

The use of decommissioned power batteries for hybrid power systems（HPS） can effectively reduce investment costs, while operational optimization for decommissioned battery energy storage systems can reduce the operating costs of hybrid power systems and increase the operational benefits of hybrid power systems. In this work, an operation optimization model of HPS featuring multi-groups of the retired batteries with different initial capacities is developed to minimize the total annual cost (TAC) by considering the capacity degradation characteristics of the retired batteries. The application of the proposed model was verified through a case study of a HPS with a photovoltaic device and a retired batteries energy storage system. Results show that the charge/discharge sequence and frequency of the retired batteries from different groups are distinct due to the difference in their initial states. In addition, the optimization of HPS operation can not only alleviate the capacity degradation of the retired batteries but also reach a lower TAC than that obtained by using a fixed-ratio operation scheme.

The actual chemical process reaction system usually consists of several interconnected cascaded reactor. To establish its mechanism model and estimate its dynamic parameters, it is necessary to repeatedly solve large-scale nonlinear differential equations. The calculation is very expensive. Since this parameter optimization process involves solving large sets of differential equations, it is very time consuming. According to the problem that intelligent optimization algorithm always needs enormous computing resource while a satisfied solution is acceptable for an industrial application, an improved state transfer algorithm (STA) is proposed to estimate the kinetics parameters of the cascaded plug flow reactors model. This method uses an opposite operator to initialize the start status of STA, and a tolerable error threshold to break the optimization process. This improvement is helpful since much computing time is saved while the global search capability and fast convergence capability of standard STA are retained. Simulation study, whose target is optimization the kinetics parameters of the cascaded plug flow reactors of an industrial hydrocracking process, verified the effectiveness and superiority of this proposed method.

This article presents a model-based optimization of the hybrid desalination system of reverse osmosis and pressure retarded osmosis with open-loop and closed-loop configurations. First, a constrained nonlinear optimization model for both configurations is formulated to minimize the normalized specific energy consumption (NSEC). In this optimization model, a set of dimensionless parameters in relation to membrane area, operating and design variables are used to build characteristic equations and simply the model formulation. To compare fairly, the pre- and post-treatment energy consumptions are considered as calculating NSEC of the open-loop hybrid configuration. By solving this model, the underlying impacts of the total dimensionless membrane area and water recovery ratio on NSECs, the membrane allocation, and the applied pressure are systematically explored. The results show that when the water recovery rate is at a normal level (≤0.55) and the total system area is sufficient (≥0.9), the closed-loop structure has a clear advantage over the open-loop structure in terms of energy saving and reduction of pretreatment costs.

The research object of this paper is acetic acid dehydration tower in PTA plant. The dynamic model was based on the dynamic mathematical model of the balance level and supplemented with the set dynamic parameters and the steady state model which was derived by using Aspen Plus. Next, the temperature of the sensitive stage was controlled by the reflux flow, and the column kettle reboiler duty was proportional to the feed flow F ₁. Aspen Dynamics was used to simulate the control strategy CS1, which has the dynamic response analysis of the key index parameters in the acetic acid dehydration tower after disturbance of the feed flow. To ensure the concentration of acetic acid at the bottom of the tower is more stable when the feed flow is disturbed, the heat load of the bottom tank reboiler is used to control the acetic acid concentration in the tower, and the control strategy CS2 is designed. Two kinds of different control strategies were analyzed and compared to their dynamic responses to obtain better control strategies under the same target conditions, which provided direction and guidance for the actual production and control strategy design.

The computational fluid dynamics phase field method was used to simulate the deposition process of single shear thinning non-Newtonian fluid droplets on a random rough surface. The analysis revealed the influence of random rough surface morphology on the movement state and equilibrium state of droplets. It was shown that, even on a smooth surface,under the same operating conditions,a shear-thinning droplet can demonstrate quite different impact behavior as compared with a Newtonian droplet. The shear-thinning property offers a much larger spreading ratio,and shorter time to reach equilibrium. The initial spreading phase is followed by a recoiling to equilibrium phase for the shear-thinning droplet,while the Newtonian droplet has a second spreading phase after the recoiling phase. On randomly rough surfaces,the maximum spreading ratio increases with the increase of either root-mean-square roughness(Rr)or Wenzel roughness parameter(Wr). With the same value of Wr,increasing Rr can lead to the decrease of the final spreading ratio,and slight decreases of equilibrium contact area and contact angle. With the same value of Rr,increasing Wr offers a faster deposition to reach an equilibrium state,and a linearly increased contact area.

Industrial fresh water desalination systems normally involve membrane(e.g. ultrafiltration, reverse osmosis etc.), ion exchangers and storage tanks. Note that ion exchange and membrane separation processes can be operated semi-continuously and they can be considered as batch partitioning regeneration units with a single inlet and two outlet streams. However, there is limited research on the synthesis of batch water networks involving freshwater desalination. This work established flow rate balance and contaminant mass balance model for batch partitioning regeneration units and tanks in the desalinated water system. According to the different demand for desalinated water under different condition, the GAMS software platform is used to solve the mathematical model. The needed capacity of each storage tank and the fresh water usage in each condition are determined via solving the model. The capacity of the storage tank can be designed according to the maximum capacity requirement of each storage tank. Simplified industrial case studies validate the validity of the proposed model.

For nonlinear system with nonlinear correlation of input signals, a partial approximate least absolute deviation based on radial basis function is proposed. In this paper, the observation data matrix is first extended by columns, and the output of the RBF network is used as an extension of the observed data matrix. Then, the extended observation data matrix and the output matrix are linearly regressed by using the partial approximate least absolute deviation. An approximate least absolute deviation objective function is established by introducing a deterministic function to replace the absolute value under certain situations. The proposed method can overcome the disadvantage of large square residual of least square criterion when the identification data is disturbed by the impulse noise which obeys symmetrical alpha stable (SαS) distribution. By adopting principal component analysis to eliminate the nonlinear correlation among the elements of data vector of nonlinear systems, the unique solution of model parameters can be easily acquired by the proposed method. The simulation experiments show that the proposed algorithm can directly recognize the nonlinear system with nonlinear correlation of input signals and suppress the influence of impulse noise.

Due to the urbanization, industrialization and rapid growth of vehicles, haze weather has become a serious environmental problem that needs to be controlled urgently in many Chinese megacities. Vehicle exhaust emissions from large amounts of petroleum fuel consumption may be a key factor in causing urban ash pollution. In this work, the fault tree analysis (FTA) method is investigated and employed for the risk assessment and causation mechanism of urban haze related to vehicle emissions in Hangzhou. After identifying all important risk factors, a haze fault tree system of “haze weather–excess emission of vehicle exhausts” is established by using the deductive FTA method. Based on the structure, probability and critical importance degree analysis, the contribution and effect of basic risk factors to the top event “haze weather-excess emission of vehicle exhausts” in Hangzhou is also carried out. The analysis results showed that “excess vehicles”, “severe traffic jam”, “high pollution vehicle’s using” and “supervision defect” were the most important risk factors for causing excess emission of vehicle exhausts in Hangzhou. This study may provide a concise and effective method for environmental risk assessment and management of oil consuming by vehicles related to urban haze in China.

Crude propylene oxide (PO) produced from propylene epoxidation contains contaminations such as acetaldehyde, methanol, methyl formate, water and so on. It is difficult to remove them by using simple distillation because the relative volatilities of these components to PO are close to 1. Meanwhile, the PO is easy to react with water and 1,2-propanediol (PG) that will affect the extractive efficiency is then produced. Therefore, a process flowsheet combined extractive distillation and liquid-liquid extraction is proposed here to effectively remove these contaminations and improve PO product yield, in which the PO hydrolysis reaction (non-catalytic reactive distillation) and the processes of water washing for PO recover and side-drawing for the removal of PG are considered. The whole process is then simulated by using process simulation software Aspen Plus, in which the NRTL thermodynamic model is employed to character the thermodynamic properties. The effects of major design parameters of the separation process, such as the solvent ratio, the theoretical stage number of extractive distillation column, the location of feed, and the temperature of solvent, are investigated. The results show that the process is reasonable, reliable, and economical superior to existing processes, which can guide industrial process design and operation optimization.

To solve the control problems of nonlinear process systems, nonlinear model-predictive control algorithms are studied. pH neutralization process is a typical nonlinear process in chemical process systems. In view of the characteristic of pH neutralization process, the entire model of pH neutralization process system and the inverse model of static nonlinear block are established by neural networks. Then two novel nonlinear predictive control strategies are studied based on model-predictive control and Hammerstein model. The neural networks model predictive control (NNMPC), which is a global solution strategy for nonlinear predictive control systems and nonlinear Hammerstein model predictive control (NLHMPC), which is a strategy based on two steppes separation control are developed and simulated by MATLAB. Control simulation results show that the NNMPC and NLHMPC control strategies have better performances on set-point tracking and anti-interference control response than PID control. They can give effective control performance to nonlinear processes.

It is difficult to achieve real-time accurate measurement for effluent biochemical oxygen demand (BOD). To solve this problem, a soft-measurement method based on mutual information and a self-organizing RBF neural network is proposed for BOD prediction in this paper. First, a method based on mutual information is employed to extract feature variables, and these variables are used as inputs to the soft-measurement model. Second, a self-organizing radial basis function (RBF) neural network based on error-correction method and sensitivity analysis is designed, and the improved Levenberg-Marquardt (LM) algorithm is used to train parameters of the neural network to shorten its training time. Finally, the soft-measurement model is applied to UCI public datasets and the real wastewater treatment process. The results show that the soft-measurement model has a more compact structure and relatively short training time, and improves the prediction accuracy, which realizes a fast and accurate prediction for BOD.

Dioxin (DXN) emitted from the municipal solid waste incineration (MSWI) process is a persistent pollutant of the “century poison”. DXN is one of the highly toxic and persistent pollution. The principal model of DXN emission is difficult to obtained duo to the complex multi-stage and multi-temperature phase’s physical chemical characteristics. In practical, DXN emission concentration is off-line measured with month or quarter period by quantified national laboratory with long lag time delay. Aiming at these problems, a new DXN emission concentration soft measuring method based on selective ensemble (SEN) kernel learning algorithm is proposed. At first, candidate kernel parameters and regularization parameters are given based on prior knowledge. Then, candidate sub-sub-models based on these super parameters are constructed. Thirdly, coupled optimization and weighting algorithms are used to build SEN-sub-models. Finally, these SEN-sub-models are selective combined as final SEN model by using optimization and weighting algorithms again. Simulation results based on the concrete compression strength and incineration process DXN data validate effectiveness of the proposed approach.

There is a clear restriction on the safe distance of hazardous chemicals stacking, but there is no effective monitoring and early warning means for monitoring. A safe distance monitoring system presented for hazardous chemicals storage stack based on indoor location technology, calculated the safe distance of stacking in warehouse by acquiring data from each node on the stack, software restore the storage state of 3D visualized hazardous chemicals. The results show that the monitoring and early warning of stacking safety distance can be effectively carried out.

To solve the problems of low accuracy and large computation in dynamic non-linear detection process of complex industrial system, a fault detection method of improved dynamic kernel principal component analysis (IDKPCA) is proposed. First, the undistinguishable degree is used to eliminate the variables with low correlation degree or no correlation degree, so as to the amount of data is reduced, then the augmented matrix is constructed for the new data after screening by extending the observed value, the nonlinear spatial correlation characteristics of variable data is extracted by KPCA, finally monitoring statistics T ² and SPE are used to diagnose system failure and identify fault variables. Simulation experiment shows that this method can effectively monitor and diagnose the fault of wind turbine, and compared with KPCA method, the improved dynamic kernel principal component analysis method is more sensitive to minor faults.

Due to the complex and changeable working conditions in chemical production process, a single soft sensor model cannot meet the requirements of the system for estimation accuracy. A new method of multi-manifold modeling is proposed in this paper based on a modified expanding search clustering algorithm. This algorithm uses the distance between manifolds instead of the Euclidean distance, adaptively determines the neighborhood radius, and introduces the local density to determine the center of clustering. The features of sub-manifold obtained after clustering are extracted by kernel isometric mapping method in manifold leaning respectively, and develop sub-models based on Gaussian process regression. The method was applied to the soft measurement modeling of a bisphenol A production device. The simulation results verify the effectiveness of the method.

Consider the problem of the dissolved gas influences the measurement accuracy of the chemical oxygen demand (COD) based on the method of ozone combined with ultraviolet radiation (O₃/UV), we proposed estimation models of dissolved gas quantity for COD determination. In terms of the parameters collected and experimental results from the COD standard solution digestion with different concentrations, the estimation models of dissolved oxygen and carbon dioxide based on PLC-LSSVMs are established respectively. The outputs of the estimation models are used as compensation for the COD detection model. The results show that the estimation models have higher estimation accuracy than the model established by PLS or LSSVMs alone. The relative errors of the COD measured by O₃/UV method after the estimated dissolved gas amount is compensated and the national standard method is both less than 5%. This study is of great significance for improving the accuracy of COD detection by O₃/UV method.

Traditional monitoring methods only use sensor observation information to perform process fault monitoring, while ignoring other valid information contained in the original data. Aiming to this problem, a PCA fault monitoring algorithm based on multi-block information extraction is proposed. Firstly, two kinds of information of the cumulative error and the change rate of process variables are defined, so that new feature information can be extracted from the data. The process is divided into three sub-blocks based on each feature, and each sub-block is processed by the PCA method. Modeling and monitoring are carried out, and monitoring results are integrated by Bayesian method. Finally, a fault diagnosis method with weighted contribution graph is proposed to find the source variable which causes the fault. The validity and feasibility of the proposed method are demonstrated by numerical examples and the application of Tennessee-Eastman (TE) process monitoring.

It is difficult to solve the multi-objective optimization problem of improving efficiency, reducing loss and reducing emissions for the catalytic cracking reaction regeneration process. The improved multi-objective quantum-based particle swarm optimization-crowding entropy sorting (MQPSO-CES) is used to solve the problem. A multi-objective optimization model is established to maximize the light oil absorption rate and synchronously minimize the coke yield and sulfide emissions. Particularly, crowding entropy sorting is used to update the archive, which accurately estimates the distribution of the non-dominated solutions. Afterwards, an adaptive factor is introduced to self-adaptively and dynamically adjust the construction of the attractor, which can balance the convergence and diversity of the proposed algorithm. In addition, with the application of a piecewise Gauss mutation operator, the precision of the local search can be enhanced. Finally, the multi-objective model is resolved with the novel algorithm. The results indicate that the improved algorithm can outperform other algorithms with convergent and well-distributed approximate Pareto fronts when dealing with ZDT3-4 and DTLZ1-2 benchmark problems. In addition, the proposed algorithm can obtain 76.22% of light oil absorption rate, 5.72% of coke yield and 626 mg/m³ of sulfide emissions in the reaction and generation process, illustrate its superiority compared with other algorithms.

According to the complexity of fault mechanism, the lack of prior knowledge, and the low diagnosis precision of traditional shallow layer neural network for the fault diagnosis of petrochemical air compressor, a kind of petrochemical air compressor fault diagnosis method is put forward based on the deep belief network because of its advantage in feature extraction and nonlinear data processing. By using state monitoring data of the air compressor, the method realizes the unsupervised characteristics learning and supervised fine-tuning of training network, constructs the deep network model of the air compressor fault, thus achieving the effective intelligent diagnosis for fault types of the air compressor. The effectiveness of the method is compared with the traditional fault diagnosis method. The results show that the diagnostic accuracy of the method is better than the traditional fault diagnosis method and the stability is better.

In view of the problem that the load (filling rate and ball ratio) of a ball mill is difficult to be judged by experience during the grinding process, a method of mill load identification based on the multi-scale fractal box dimension of modified ensemble empirical mode decomposition (MEEMD) and extreme learning machine (ELM) is proposed. Firstly, the MEEMD algorithm is used to decompose the grind signals in different load states to get intrinsic mode components. Then, the correlation coefficient method is used to reconstruct the sensitive modal components to get the signal after noise reduction. By analyzing the multi-scale fractal box dimension of the reconstructed signal. The results show that there are obvious differences in the multi-scale fractal box dimensions of the under load, normal load and overloading state, and it can be well divided into different load states of the mill. The multi-scale fractal box dimension of regrinding signal is used as the input of extreme learning machine (ELM), and the load state of mill is output. The load identification model of mill is established. The effectiveness of the method is verified by grinding experiments. The recognition rate is as high as 94.8%, and the model can accurately identify the mill load status.

In the momentum of reducing CO₂ emission of coal to ethylene glycol process, a novel carbon dioxide utilized coal to ethylene glycol (CUCtEG) process is proposed, simulated, and optimized. The novel process is assisted with coke oven gas to enhance resource and energy efficiencies as well as reduce CO₂ emission by integrating with dry/steam-mixed methane reforming technologies. Based on the rigorous steady-state simulation of the process, key operational parameters of the novel process are investigated and optimized. The optimal feed ratio of coke oven gas to coal and the split ratio of that for steam methane reforming reaction are 0.68 and 0.74. Compared with the traditional process, the CO₂ emission of the new process is reduced by 94.05%, and the exergy efficiency is increased by 15.17%.

It is known that due to the change of light condition on rare earth extraction industry, makes soft-sensing model for rare earth element content based on the color characteristics of the rare earth solution have larger error. First, the Grey Edge algorithm with parameter optimization is used to correct the image of rare earth solution under different illumination conditions to the standard illumination. Then, the first moment of H, S and I components in the HSI color space of the Pr/Nd solution image is used as a model. The weighted least squares support vector machine (WLSSVM) is used to model the component content. Finally, the simulation experiments of rare earth solution images under different illumination conditions are carried out. The simulation results show that the images of rare earth solutions under different illumination conditions can meet the high-accuracy and rapid requirements of element component content detection in rare earth extraction process.