In this paper, a new detection and classification system of faulty bearings is presented. This system is based on artificial intelligent techniques and vibration signals in the frequency domain produced by the faulty bearings. The system consists of several neuro-fuzzy systems in cascade, along with measurement equipment for the vibration spectral data. These neuro-fuzzy systems have been used as bi-classifiers. That is, each neuro-fuzzy system is specialized in the classification between two different types of rolling bearing status. A careful selection process for rules has been included in the learning algorithm. Moreover, the demodulated vibration signal has been used as input to the neuro-fuzzy systems based on the Sugeno approach. Several trials were carried out, taking into account the vibration spectral data collected by the measurement equipment for each bearing. Different results with three types of faulty bearings using the proposed approach are shown, where satisfactory results have been achieved.
Position measurement in machine tools is usually carried out through rotary or linear encoders. Quite often, it is considered that part dimensional errors are only due to direct strain, thermal or vibration e ects in the machine tool structure. There is, however, another important source of error: the encoder itself. This paper is intended to recognize, classify and determine the value of the di erent kinds of error in optical linear encoders: strain and displacement errors, thermal errors and errors caused by vibrations. The error identi®cation will also permit a better understanding of some of the limitations that appear in the use of these devices.
Optical encoders are sensors based on grating interference patterns. Tolerances inherent to the manufacturing process can induce errors in the position accuracy as the measurement signals stand apart from the ideal conditions. In case the encoder is working under vibrations, the oscillating movement of the scanning head is registered by the encoder system as a displacement, introducing an error into the counter to be added up to graduation, system and installation errors. Behavior improvement can be based on different techniques trying to compensate the error from measurement signals processing. In this work a new “ad hoc” methodology is presented to compensate the error of the encoder when is working under the influence of vibration. The methodology is based on fitting techniques to the Lissajous figure of the deteriorated measurement signals and the use of a look up table, giving as a result a compensation procedure in which a higher accuracy of the sensor is obtained.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.