نظارت بر وضعیت خودکار بدون سنسور برای کنترل نگهداری پیشگویانه ابزار ماشین آلات

Modern manufacturing systems are characterized by numerous interacting machine tools each with sophisticated maintenance. In order to be competitive, it is possible to reduce the system downtime by applying sensorless automated condition monitoring (SACM). This paper presents newly developed and tested SACM-algorithms based only on signals which are available in position controlled drives such as position, speed and motor current. The algorithm is based on comparing current characteristic parameters with those which were taken when the machine was new.

The operational availability of machine tools is an essential prerequisite for the profitability of the manufacturing industry. Maintenance is driven by increasing cost effectiveness. It is state-of-the-art to use accessory measurement systems to analyse bearing failure or shaft unbalance conditions. This paper presents experimental results and specific SACM algorithms for signal analyses on the wear of feed drives. A common characteristic of the components examined is the rolling contact. Rolling contacts show wear in a typical sequence. When used in the Hertzian contact stress range a micro-damage is caused in the metallic structure. Due to many load cycles and roll-overs pitting or peeling occurs eventually. If the surfaces of the rolling pairing are damaged, failure, caused by wear, is due within a short period of time. Surface damages can be caused by different loads or environmental influences. Loads and environmental influences, such as overload, corrosion and dirt, often cannot be fully taken into account for the design and can partly change during operation. The change of the surface or the geometry of the counter parts is ultimately the cause of failure. Generally, damages release energy, which is converted into vibrations. It is therefore possible to record and describe damages by examining the signal energy and signal power [1]. The damage types have different effects. They cause measurable physical phenomena, the signal characteristics of which permit a classification into damage types [2]. 1.1. Periodical damages Periodical damages produce signals, which are caused by periodical impulse-like excitation. The cycle period depends on the relative speed of the components. Hence a characteristic frequency is generated, which can be clearly assigned to the components involved. It can be observed that the impulse sequences show modulations [3]. Periodical damages cause impulses with a speed-dependent frequency. This frequency excites the resonance frequencies of the components. Increasing wear causes a rise of the amplitudes of the excitation frequency and the resonance frequencies. At the same time the resonance frequencies shift down. The shift is caused by non-periodical damages, e.g. loss of stiffness and change in friction [4]. 1.2. Non-periodical damages Non-periodical damages do not appear in a cyclic manner. They can occur spontaneously in time and place, discretely, in irregular sequence or for an indefinite period or length. Their signal aspects do not show any specific, speed-proportional frequencies. This damage type is, for example, caused by a change in friction or stiffness of a component [5]. For the modelling of mechanical components non-periodical damages are described by linear and non-linear variables. Non-periodical damages can be recognized by means of physical parameter estimation procedures or state observers [6]. 1.3. Characteristic wear parameter Spectra of damage-free components, which are caused by constant excitation, are mainly determined by mechanical imperfections, such as pitch errors and unbalances. If wear is to be described by a characteristic parameter, three effects need to be taken into account: 1. Damages produce impulse-like excitations 2. Resonance frequencies move due to increasing wear 3. Interferences of excitation and resonance can occur due to shift Measurements are taken at constant speed so that vibrations are excited uniformly and transient excitation is avoided. Therefore it should be possible to interpret the position and or the speed signal to detect wear.

The measurement and evaluation of positioning signals allows the computation of characteristic parameters which enable us to quantify wear. It was shown that the accuracy of the positioning is crucial to initiate maintenance. The proposed frequency and distance domain characteristic parameters help to quantify the wear of feeddrives and even allow to localize and to assign damages to specific parts. The described methods may be implemented by using the sensors which are already included in the machine i.e. linear and rotary encoders. The interpretation of the signals is embedded in the resources of the CNC-control. Future effort will be aiming towards the interpretation of the presented characteristic parameters and the application of the algorithms on different types of machine tools. With additional effort the algorithms should also be applicable to coupled multi axis kinematics.