Arc sensor, one of the most popular and effective sensors for automatic seam tracking of arc welding, is systematically studied. A mathematic model of the static and the dynamic properties of the sensor derived on the basis of control theory and experimental method are described. The model extends our understanding of arc sensor from qualitative level to quantitative level. This is important for its design and further development.

Real-time measurement of temperature field can be achieved by colorimetric method using ICCD as sen-sor. However, due to limited dynamic response range of the ICCD, the measurable range of temperature is very small. In order to expand the temperature measuring range, a device was developed which can take several two-color images with different shutoff times subsequently within very short time. First, the principle of real-time measurement of tem-perature field with two-color thermal images is introduced. Then theoretical analysis on this method, including deter-mination of various parameters, such as shutoff time, number of temperature range, wavelength of the colors, etc. is described. Experiments were conducted, which proved that theoretical analysis coincides with the experimental re-sults.

The real-time measurement of welding temperature field by colorimetric method is described, and with the data acquired from it closed-loop control system of the parameters of temperature field is developed and tested. Ex-perimental results prove that it has high measurement speed (time of a field within 0.5s) and good dynamic response quality. Weld penetration can be controlled satisfactorily under unstable welding condition.

Due to violent disturbance of the welding arc, signal processing is the key problem of ap-plication of the sensor. By means of the new technique the arc sensor can recognize not only V-groove but also lap joint and butt joint. The sensor has good recognition ability even for welding process with very large current disturbance, e.g. pulsed arc and short circuit welding process, etc. The proposed technique is developed on the basis of modern digital filtering theory and mathematic transformation of the signals from time domain into frequency domain.