In cases where the model of plants is complex or varying in time, certain parameters need to be experimentally determined; doing these experiments we have a valuable tool for understanding the behavior of the plant. The behavior of a dynamic system can be understood throw the utilization of techniques as analytical model (from the basis principles of physics), with direct measure of many parameters of the plant with an identification of models from inputs and outputs. A procedure of modeling needs many interactions. Additional, acceptable control project needs the validation of the model. Examples of complex plants are iron ovens and helicopter rotors. Example of variant time plants is satellites that change their temperature when they are in orbit. A pendulum is an example of plant that can be modeling for basic principles. The most popular method is identifying of model plant using the experimental data, and then designing the control process. An engineer usually realizes certain numbers of experiments and uses them with many optimization techniques to build a model plant. The plant model is then used for designing model-based controllers.The methods that use experimental data can be divided on four categories:Indirect Control Project, Direct Control Project, Indirect Adaptive, and Direct Adaptive. The techniques are differentiated by operating offline and online and if the plant model is used directly or indirectly in the control of the project. When the project based on the plant model is used online is usually referred to as indirect adaptive control. The technique typically begins by assuming a nominal model for the plant. When new data are collected, the output is compared with the planned output for the plant model producing a nominal error. The gradient of the error with respect to the parameters of the plant is used to modify the parameters of the plant to improve the plant model. Periodically the control law is updated using the latest model obtained.One characteristic of the model-free techniques has a simple and integrated control law derivable directly from experimental data and performance specifications. However, the main attraction of model-free technique it is its implementation online, in these cases, we can divide the model-free technique in two parts:tuning model-free and modcl-frcc control. Techniques for tuning model-free arc characterized by directly determining the parameters of controlling data input and output through performance criteria desired for the closed loop system. These techniques can be explorer on determining parameters of the following controllers:PID (Proportional+Derivative+ Integral), GMV (Generalized Minimum Variance).The techniques of model-free control are characterized by the fact of using only the input and output data of the plant to be adapted. The control laws are derivable, generally, without the use of the traditional models(CAR (Controlled Auto Regressive)), CARMA (Controlled Auto Regressive Moving Average)). However, in some cases, it is common to obtain the control laws from simplified representations for the plant(PG (Pseudo-Gradient), PPD (Partial-Pseudo-Derivative)). The mechanism of adaptation uses identification techniques based on the LMS algorithm (Least Mean Squares) or techniques of computational intelligence such as neural networks, fuzzy systems.The close links between identification and control design may lead to an increase in project automation controllers. However, this conjecture can only be confirmed throw the experience of control engineers who are enabled for testing in both field based-control models as the model-free technique.Following the idea introduced above, the model-free controllers can be extended to other forms according to different dynamic linearization models. This thesis will present two types of model-free controllers with different control updating laws. A class of SISO discrete-time nonlinear systems can be transformed into a Partial Form Dynamic Linearization (PFDL) of a data model to build the PFDL based model-free control scheme, integrating control algorithm and PG estimation algorithm. This class of SISO discrete-time nonlinear systems can also be transformed into Full Form Dynamic Linearization (FFDL) data model.We evaluate the applicability of the controllers to tracking reference and disturbance rejection (regulation). To evaluate the characteristics of proposed controllers, nonlinear processes, including liquid level systems and heat exchanger are used.