Stefania Tronci

Monitoring Pollutant Emissions in a 4.8 MW power Plant through Neural Network

Abstract The prime aim of this paper is to address the relevant issues associated to the development of neural-based software sensors for monitoring the pollutant emissions coming out from combustion chambers. The objective is to prove the potential of software sensors as alternative monitoring systems to conventional analytical equipment. The preliminary results refer to a 4.8 MW power pilot plant operating at the Enel Santa Gilla Research Center in Cagliari (Italy).

Reconstruction of chaotic time series by neural models: a case study

Abstract This work analyses the problems related to the reconstruction of a dynamical system, which exhibits chaotic behaviour, from time series associated with a single observable of the system itself, by using feedforward neural network model. The starting network architecture is obtained setting the number of input neurons according to the Takens’ theorem, and then is imporved by slightly increasing the number of inputs. The choice of the number of the hidden neurons is based on the results obtained testing different net structures. The effectiveness of the method is demonstrated by applying it to the Brusselator system (Phys. Lett. 91 (1982) 263).

CATALYTIC CONVERTER DESIGN FOR MINIMISATION OF COLD-START EMISSIONS

Abstract The axial catalyst distribution in a monolithic converter that minimises cold-start pollutant emissions is investigated numerically, under the constraint of fixed total catalyst surface area. Various warm-up mechanisms can be present during the transient period. The catalyst distribution affects greatly which mechanisms prevail. For the optimal distribution, a large amount of catalyst is required in the upstream section of the monolith This ensures that the hot spot is kept at the monolith inlet throughout the warm-up period, and hence heat transfer by convection dominates. A result of practical significance is that the evolution and the steady-state value of the temperature of the exhaust gas stream at the monolith inlet do not affect significantly the form of The optimal distribution. Even though the local catalyst surface area of the optimal distribution in the downstream section of the converter is reduced as compared to the uniform distribution, steady-state performance is not adversely affe...

A stochastic approach for the prediction of PSD in crystallization processes: Analytical solution for the asymptotic behavior and parameter estimation

Abstract Recently, a novel stochastic formulation based on the Fokker–Planck equation (FPE) for the description of anti-solvent mediated crystal growth process was proposed. Here, we further expand these results by analyzing the asymptotic (end of the batch) solution of the FPE for the CSD. In this regard, the analytical solution of the stationary FPE is exploited for predicting the end of the batch CSD as function of the model parameters. Furthermore, the availability of such analytical solution is used to simplify and diminish the computational burden of the parameter estimation problem. Two alternative approaches for parameter estimation are discussed based on the use of the analytical solution of the FPE and of the dynamic of the logistic equation (deterministic component of the FPE approach). Validations against experimental data for the NaCl–water–ethanol anti-solvent crystallization system are presented.

Development of a 4-Measurable States Activated Sludge Process Model Deduced from the ASM1

Abstract The Activated Sludge Model No.1 (ASMI) allows the prediction of organic matter degradation, nitrification and denitrification in the activated sludge bioreactors using thirteen states variables. This paper deals with modal reduction procedure applied to the ASM1, that followed by physical considerations about the process allows the development of a 4-state variables model which is still able to describe the nonlinear behavior of bioreactors in the Activated Sludge Process (ASP). Furthermore, coupling the reduced bioreactor model with a simplified model for the secondary settler (as the Takacs model), the overall ASP behavior is deduced.

Optimal Catalyst Distribution in Catalytic Plate Reactors

Steam reforming of methane taking place in a Catalytic Plate Reactor (CPR) having as heat source catalytic methane combustion is studied. The CPR thermal behaviour is maintained in an operating range that provides maximum conversions and at the same time is not detrimental to the catalyst. A non-uniform catalyst distribution on either channel, provides a degree of flexibility to control the reactor thermal behaviour. Optimisation is employed in order to obtain optimal non-uniform catalyst distribution having as objective the minimisation of the difference between the temperature along the reactor and its inlet value. In this way, the reactor operates close to isothermal conditions avoiding formation of hot or cold spots and high conversions are achieved. Two types of catalyst distributions are presented. First, a non-uniform distribution along the catalytic plate which can maintain the temperature along the reactor within +/- 2.5 K from the inlet value. A two-step catalyst distribution, which can be easier to implement in practice, is further considered. This type of catalyst distribution maintains the reactor temperature in a wider range, +/-19 K if it is used in the endothermic channel, and +/-38 K if it is used in the exothermic one.

Conversion of glucose and sorbitol in the presence of Ru/C and Pt/C catalysts

The conversion of glucose and sorbitol in the presence of Ru and Pt catalysts supported on carbon was carried out at different pressure and temperature conditions, using a batch and a semi-batch reactor. Attempts were made to improve the selectivity of glycols and alcohols (ethanol), introducing a promoter and inhibiter of the hydrogenolysis in the reactant mixture. On the basis of these results, which confirm the higher activity of Ru with respect to Pt, and the important role of an inhibitor like sulphur, the mechanism driving these reactions and the promising thermocatalytic conditions are clearer.

STATE ESTIMATION IN A CATALYTIC REACTOR VIA A CONSTRUCTIVE APPROACH

Abstract In this work the problem of designing a reduced-order observer to infer the concentration in a catalytic reactor with temperature measurements and a simplified model is addressed. The process motion has a poor observability property, meaning that a carefully tuned high-order EKF technique is required to design an observer. Here, a constructive low-high gain estimation scheme is considered to achieve the same task with a low-order nonlinear observer. An error dynamics analysis yields a convergence criterion coupled with a simple tuning guideline. The proposed reduced-order geometric-type scheme is tested with experimental data and compared with its full-order EKF.

Global-nonlinear stochastic estimation of exothermic reactors with temperature measurement

Abstract In this work, the problem of estimating the concentration of exothermic rectors with temperature measurement is addressed. The problem is treated within a global nonlinear stochastic framework, according to the Fokker Plank-based Kushner filtering theory. The on-line solution of the associated two-dimensional partial differential equation driven by the temperature measurements yields the evolution of the conditioned concentration-temperature probability density function (PDF), with considerable more information than the one provided by standard EKF based on a local-nonlinear approach. A catalytic reactor with deterministic multistability, experimental data, and previously addressed with EKF is addressed as case example, yielding: (i) the on-line evolution of the (possibly bimodal) conditioned concentration probability density function, and (ii) mean uncertainty values which are better than the ones drawn with EKF.

Accurate and efficient solution of distributed dynamical system models

Abstract Efficient and accurate methods for numerical simulation of time evolution of distributed dynamical systems are presented and discussed. As case example, stochastic models based on the Fokker-Planck equation are considered, describing the evolution of the probability density function (PDF) of the state variables. Unlike most of the methods presented so far, the proposed methods fulfill the normalization and positivity conditions for the PDF even for very small values of the model diffusivities. Hence, the methods appear to be suitable for simulating chemical distributed process systems.