Chemical research is nowadays more and more devoted toward the comprehension of chemical systems of high complexity, such as environmental or industrial ones. The usual experimental investigation of such systems requires a deep knowledge about their microscopic structure; expecially in order to estimate parameters as kinetic rate constants or stability ones. The information obtained from such studies is extremely useful expecially in industrial applications to establish optimal operation conditions. Even with the aid of sophisticated computer programs, based, for example, on general least-squares algorithms, the study of complex chemical systems in terms of detailed mechanism and kinetics is still a hard work because, to estimate values of parameters, it is necessary to know and solve the whole system of equations (differential ones in kinetics) governing the system . The main task of chemical kinetics is the estimation of mechanism, rate constants and the factors affecting them. This information is useful to predict optimal reaction conditions for example to reach high yield. Moreover, such a knowledge is useful for the development of more effective catalysts or new kinetic method of analysis. Detailed description of the system, based on knowledge of reaction mechanism and physico-chemical constants, is usually called “hard” modeling. This approach might suffer of several difficulties, due to the intrinsic complexity of the system and also of phenomena like possible undistinguishability or not unique identifiability of kinetic models . A new possibility is offered by so called “soft” modeling, which is able to find hidden relationships between a set of “input” and a set of “output” data with a limited or even no knowledge about the microscopic behaviour of the chemical system. The aim of this work is to study potential of neural networks (ANN) “soft” modeling in kinetics. The use of ANN will be studied for: 1. estimation of rate constants knowing the kinetic model; 2. estimation of analyte concentration determined by kinetic methods without any knowledge about the mechanism of the process; 3. application to some relevant industrial cases. Several different cases were studied as, for example, consecutive or cyclic reaction paths. For each case, both the network architecture and experimental design were evaluated and optimized; moreover the quality of the model was checked with verification data sets. The power of ANN in modeling complex systems was demonstrated and several different parameters can be estimated. In this context two different cases can be distinguished. In the first case it is necessary to have a limited knowledge about the mechanism of the reaction and we can estimate parameters for unknown systems under the condition that the mechanism is the same. In the second case it is possible to model the system and we can estimate outputs without any knowledge about the microscopic behaviour of the system. ANN represent simple, general and robust approach to model complex chemical kinetic systems and they are widely applicable even to problems where “hard” models are not available. Figure 1: A general structure of ANN architecture with INPUTS, OUTPUTS and 2 hidden layers. References:  Molga, E. J.; Van Woezik, B. A. A.; Westerterp, K. R., Neural networks for modelling of chemical reaction systems with complex kinetics: oxidation of 2-octanol with nitric acid. Chemical Engineering and Processing 2000, 39 (4), 323-334.  Vajda, S.; Rabitz, H., Identifiability and Distinguishability of General Reaction Systems. The Journal of Physical Chemistry 1994, 98 (20), 5265-5271.