Empirical Modeling Techniques

In every case in which a kinetic model is selected to represent adequately a reaction, the rate surface predicted by the model must be compared to the surface observed in the data. In the methods discussed in Section II, only one section through the entire rate surface was examined for example, the dependence of initial rate on total pressure could be investigated when in fact the total rate surface constituted the dependence of rate on several component partial pressures and temperature. The misleading results obtain- 

In this section, methods are described for obtaining a quantitative mathematical representation of the entire reaction-rate surface. In many cases these models will be entirely empirical, bearing no direct relationship to the underlying physical phenomena generating the data. An excellent empirical representation of the data will be obtained, however, since the data are statistically sound. In other cases, these empirical models will describe the characteristic shape of the kinetic surface and thus will provide suggestions about the nature of the reaction mechanism. For example, the empirical model may require a given reaction order or a maximum in the rate surface, each of which can eliminate broad classes of reaction mechanisms. 

The importance of the parameter estimates becomes apparent from the data analysis. Suppose a nonlinear reaction-rate equation contains two independent variables and a set of unknown parameters  

The rate surface can be approximately represented, in the region where the data are taken, by a Taylor expansion 

the coefficients b represent first and second partial derivatives of the rate expression f(xx, x2 K) or functions thereof. 

---

However, there are prices to pay for the advantages above. Most empirical modeling techniques need to be fed large amounts of good data. Furthermore, empirical models can only be safely applied to conditions that were represented in the data used to build the model (i.e., extrapolation of such models is very dangerous). In addition, the availability of multiple response variables for building a model results in the temptation to overfit models, in order to obtain artificially optimistic results. Finally, multivariate models are usually much more difficult to explain to others, especially those not well versed in math and statistics. 

These different definitions of interatomic distances are very important when accurate parameters are required, and sometimes the effects of ignoring the corrections can be significant even when lower precision is needed. But what we should remember is that rotational spectroscopy is one of only two general experimental methods for the determination of gas-phase structures, and that such structures are the basis of empirical modeling techniques and provide the data against which aU computational methods can be evaluated. 

The second part of the book deals with empirical modeling. Various empirical modeling techniques are used that are all data based. Some techrtiques enable the user to develop linear models with other techniques non-linear models can be developed. It is good practice to always start with the most simple linear model and proceed to more complicated methods only if required. 

Molecular dipole moments are often used as descriptors in QPSR models. They are calculated reliably by most quantum mechanical techniques, not least because they are part of the parameterization data for semi-empirical MO techniques. Higher multipole moments are especially easily available from semi-empirical calculations using the natural atomic orbital-point charge (NAO-PC) technique [40], but can also be calculated rehably using ab-initio or DFT methods. They have been used for some QSPR models. 

At times, it is possible to build an empirical mathematical model of a process in the form of equations involving all the key variables that enter into the optimisation problem. Such an empirical model may be made from operating plant data or from the case study results of a simulator, in which case the resultant model would be a model of a model. Practically all of the optimisation techniques described can then be appHed to this empirical model. 

Implementation Issues A critical factor in the successful application of any model-based technique is the availability of a suitaole dynamic model. In typical MPC applications, an empirical model is identified from data acquired during extensive plant tests. The experiments generally consist of a series of bump tests in the manipulated variables. Typically, the manipulated variables are adjusted one at a time and the plant tests require a period of one to three weeks. The step or impulse response coefficients are then calculated using linear-regression techniques such as least-sqiiares methods. However, details concerning the procedures utihzed in the plant tests and subsequent model identification are considered to be proprietary information. The scaling and conditioning of plant data for use in model identification and control calculations can be key factors in the success of the apphcation. 

In the past three decades, industrial polymerization research and development aimed at controlling average polymer properties such as molecular weight averages, melt flow index and copolymer composition. These properties were modeled using either first principle models or empirical models represented by differential equations or statistical model equations. However, recent advances in polymerization chemistry, polymerization catalysis, polymer characterization techniques, and computational tools are making the molecular level design and control of polymer microstructure a reality. 

Peterson used the skill score to evaluate the performance of his empirical statistical model based on orthogonal functions. The skill score equals 1.0 when all calculated and observed concentrations agree, but 0 when the number of correctly predicted results equals that expected by chance occurrences. The statistical technique had a skill score of 0.304. An 89-day, 40-station set of the data was used to check a Gaussian diffusion model, and this technique gave the diffusion model a skill score of only 0.15. (Recall that the statistical empirical model was used for 24-h averaged sulfur dioxide concentrations at 40 sites in St. Louis for the winder of 1964-1965.)... 

Most of the models in this book are empirical models that provide an approximate description of the true behavior of a system. However, the techniques presented for use with empirical models are applicable to many mechanistic models as well. 

Such applications of NN as a predictive method make the artificial neural networks another technique of data treatment, comparable to parametric empirical modeling by, for example, numerical regression methods [e.g., 10,11] briefly mentioned in section 16.1. The main advantage of NN is that the network needs not be programmed because it learns from sets of experimental data, which results in the possibility of representing even the most complex implicit functions, and also in better modeling without prescribing a functional form of the actual relationship. Another field of... 

In the next sections, we will present a series of empirical model based case studies that illustrate different planning techniques commonly used in practice by many refineries. 

As with diatomic molecules, the performance of semi-empirical models in dealing with frequencies in one-heavy-atom hydrides is very poor. These techniques are not to be trusted for this purpose. 

For continuous process systems, empirical models are used most often for control system development and implementation. Model predictive control strategies often make use of linear input-output models, developed through empirical identification steps conducted on the actual plant. Linear input-output models are obtained from a fit to input-output data from this plant. For batch processes such as autoclave curing, however, the time-dependent nature of these processes—and the extreme state variations that occur during them—prevent use of these models. Hence, one must use a nonlinear process model, obtained through a nonlinear regression technique for fitting data from many batch runs. 

J. G. Magallanes, P. Smichowski and J. Marrero, Optimisation and empirical modeling of HG-ICP-AES analytical technique through artificial neural networks, J. Chem. Inf. Comput. Sci., 41(3), 2001, 824-829. 

Collision-induced dipoles manifest themselves mainly in collision-induced spectra, in the spectra and the properties of van der Waals molecules, and in certain virial dielectric properties. Dipole moments of a number of van der Waals complexes have been measured directly by molecular beam deflection and other techniques. Empirical models of induced dipole moments have been obtained from such measurements that are consistent with spectral moments, spectral line shapes, virial coefficients, etc. We will briefly review the methods and results obtained. 

As shown in this review, test equipment integrated with several diagnostic techniques is preferred for a deeper insight into the mechanisms that cause performance losses and spatial non-uniform distribution. As a consequence, more information, which is simultaneously obtained with these diagnostic tools, will strongly support development of empirical models or validate theoretical models predicting performance as a function of operating conditions and fuel cell characteristic properties. 

The -1" level represents the lower limit, while the "+1" level represents the upper limit of each variable. A statistical modeling technique was used to obtain an empirical model able to reproduce the experimental data. 

---