Optimal control of the drilling process requires that relevant physical models must be used while drilling. The reliability of these models must first be improved by comparing their results with actual downhole and surface data. Measurements of numerous drilling parameters and variables, sampled at high frequency, are then required. Trafor is an MWD system designed and operated by the Institut Français du Pétrole for research on drill-string dynamics. It measures 15 downhole and 5 surface mechanical parameters. Because Trafor is fitted with an electric link adapted to rotary drilling, it transmits the downhole measurements to the surface in real time with a very high sampling rate of 1800 measurements per second. Furthermore, all the signals are fully synchronized. Hence, Trafor allows cross analysis between any signals, regardless of whether they are from downhole or from the surface. New data recorded in a 1060 m deep well exhibit stick-slip behavior for various weights-ön-bit and rotary speeds. Downhole data analysis coupled with numerical simulations shows that, in this field case, stick-slip occurred as an unstable drilling process. The origin of the instability lies in the relationship between the torque at the bit and the rotary speed. From this relationship, we can explain why stick-slip is known to disappear whenever the rotary speed (rpm) is increased or weight-on-bit is decreased. A stability analysis is used to derive operating conditions that prevent stick-slip. Simulations of drilling with adjusted PID or with a downhole "anti stiçk-slip" tool are described. For both simulations, stick-slip can be cured.

Detecting drilling abnormal vibrations from the sole surface measurements is one of the major research goal today, but signal processing procedures are not efficient enough by themselves. This paper presents a new process that enhances the use of surface measurements with the help of a model representative of the dynamic phenomena of the whole drill string. We show how the model is transformed and reduced for real time application. One of the computed outputs of the reduced model is continuously fitted with the equivalent measured data to avoid its drifting away. Whenever a discrepancy is noticed between computed and measured data, the reduced model is automatically adapted. Therefore all the computed outputs of the reduced model and mainly the ones at the bit are close to the real movements of the bit which can be used for detecting abnormal vibrations of the drill string. Introduction Using surface measurements to detect drill string abnormal vibrations is not a recent idea. Whoever has been on a rig floor has noticed the vibrations of the Kelly cock or the noise generated in the drill stem. In 1971 Lutz et al. use an advance surface data acquisition and treatment system to produce a log called SNAP log. This log is built from surface measurements of the tension and vertical acceleration and can be read as a sonic. The same measurements were also used to detect bad functioning of the three cone bit. The idea of using surface measurements to detect down hole vibrations was again raised in the second half of the eighties. The first step was to engineer and build an advanced surface data acquisition system. One can notice the ADAM system or the DYNAFOR project. Subs were built to record surface measurements relative to the mechanical behavior of the drill string such as drill string tension and torque but also measurements relative to vibrations as vertical and rotational accelerations. These subs are connected under the power swivel and therefore measure directly the vibrations of the drill string. These subs are now commercially available. At the same time IFP has started a research program to understand the sources of abnormal vibrations in the drill string. One aspect of this program was the development of a data acquisition system named TRAFOR including a surface and a downhole measurements sub. An electric link between the two subs has been designed so that the system can be used during standard rotary drilling (Figure 1). This electric link allows synchronization between the downhole and surface data. TRAFOR provides a hill set of downhole and surface measurements and a high sampling rate equal to 360 data per second and per channel. The main characteristics of the TRAFOR system are listed in Table 1. The problem was raised for interpreting the data recorded with these enhance data acquisition systems. Many theoretical works were performed to build models that can be used to simulate the main drill string vibrations. Very few take into account the relationship between the drill string and the borehole wall or the bit and the formations. Clayer and al. show that it is necessary to take into account the rig for the top end of the drill string and the relationship between the bit and the formation for its bottom end. Both ends have in fact an important effect on the behavior of the drill string dynamic regime. Pavone et al. built a model simulating the dynamic torsion of the drill string. In this model the authors incorporate friction laws at the borehole walls and at the bit. These laws are derived from actual data recorded with the TRAFOR system. Through this model they show that the source of the stick slip phenomena is located at the bit level where reactive torque is decreasing when the rotation speed is increasing. P. 97