This paper describes a smart eddy-current sensor for locating and identifying metal tags used to recognize buried pipes. We first describe the tags used and the technology of our locator, which is based on the induction-balance principle. We also describe an essential measurement-distance system controlling the sampling of data. An originality of the detector is the use of two flat coils carved onto an epoxy-resin support to generate the best possible electromagnetic field. The tags are made up of metal pieces of varying sizes separated by spaces of varying sizes. Intelligent pattern-recognition methods and their combination by the theory of evidence, which provide good and reliable recognition, are briefly presented. Finally, the system's performance for various depths of burial is analysed and ways to improve these results are presented.

La partie sensible des capteurs de proximité magnétiques est souvent constituée d'une bobine enroulée autour d'un noyau. Nous avons réalisé une bobine dont la géométrie est différente. Elle permet par son incorporation dans un oscillateur, la réalisation d'un capteur de proximité peu coûteux. Nous présentons ici les caractéristiques physiques et électriques de plusieurs échantillons. Cette étude constitue une première approche d'optimisation du capteur et de son électronique associée. Nous montrons que l'on peut détecter correctement un bord de tôle ainsi que la présence de trous dans une tôle

No abstract

A smart sensor is a device whose response takes into account and corrects for environment variables. This study shows that such a device can be produced. The major influential variables studied were temperature and humidity. The sensor was a sensitive coil inserted in an LC oscillator circuit. Oscillator operation was improved to obtain sufficient stability. Data were acquired on a testbench. Although the measurements were distorted by many phenomena (noise, dilation and so on) it was quite possible to design a displacement measurement smart sensor, based on several theoretical models. They predict a frequency response versus distance and temperature. The experimental results were in good agreement with the models.