Measuring and testing – Vibration – By mechanical waves
Reexamination Certificate
2000-08-14
2003-07-15
Kwok, Helen (Department: 2856)
Measuring and testing
Vibration
By mechanical waves
C073S597000, C073S646000, C073S660000, C702S056000
Reexamination Certificate
active
06591682
ABSTRACT:
BACKGROUND OF THE INVENTION
1. Field of the Invention
The invention relates to a device and a process for signal analysis. The signals can be of general form and can be present especially as one-dimensional, particularly time signals. Signals of this type in the area of engineering are derived for example from noise signals, as are caused by running machinery. Another example from the area of engineering is signals which represent a regular machine state, or which are characteristics of a brief or continuing defect. These defects can relate to the machinery itself, or a product produced by this machinery.
2. Description of Related Art
The numerous and varied signal analysis processes of type to which this invention is directed are known. If the ratio between a signal portion of interest and a noise portion is large enough, analysis is usually less difficult. But, it is of special interest and considerable difficulty to extract those signal portions which are strongly adulterated by noise. Likewise, it is of special interest to identify individual signals or those which occur only briefly or sporadically and possibly stochastically when a host of other signals is present at the same time.
For example, it is known that it is extremely difficult to identify bearing noise in a machine when the signal
oise ratio becomes less than a value of roughly 0.25 (compare British Journ. of NDT, Vol. 35 No. 10, p. 574 ft). The use of conventional Fourier analyses has almost no effect in this case.
Furthermore, a device for detecting or analyzing machinery damage of a type to which this invention is also directed, especially for detection of errors on roller bearings, is known from U.S. Pat. No. 3,842,663. This patent explains a standard problem which occurs when machinery or bearing damage is detected. This is based on the fact that the noise periodically emitted by bearing defects or other defective hardware parts is much smaller than regular noise which is produced by this machinery and may be present as solid-borne noise.
The approach presented in U.S. Pat. No. 3,842,663 is based on acquiring machinery noise, such as for example transmission or bearing noise, by means of a suitable detector and converting it into an electrical signal. In this case, the damaged bearing or machine parts produce pulse-like noise portions which preferably excite the natural resonances of this detector. Its electrical output signal is processed by a pre-filter and a demodulator. The resulting signal then corresponds essentially to a pulse series. In this pulse series, an individual pulse represents an individual noise caused by sudden motion components.
In periodically repeated pulses of this type, it is possible to determine an especially pronounced or emerging signal frequency with an output-connected spectral analyzer. This frequency, at a known base frequency of the machinery or certain rpm and characteristics of a bearing, allows a conclusion regarding a defect within the machine or bearing.
U.S. Pat. No. 5,679,900 describes how the aforementioned device can be used by using a switchable filter in a paper-making machine. This document also considers that the machinery defects can be characterized by pulse-like vibration portions compared to regular machinery noise, and that they can be emphasized by suitable vibration sensors and by suitable signal processing against the base noise of the machinery and mechanical assemblies.
But, the described devices and processes are subject to the disadvantage that efficient, and thus expensive, spectral analyzers must be provided for reliable representation of the results obtained or additional complex filter units must be connected on the input side.
In addition, the pulse like vibration portions tend to be characterized by significant amplitude, but of rather little energy content. This is because of their short duration. In terms of energy, such pulse-like vibration portions are likely to be “buried” in background noise and thus difficult to impossible to find.
SUMMARY OF THE INVENTION
The primary object of the invention is to improve the quality of analysis results and to significantly reduce the required effort for detecting machine damage, especially bearing damage, by cost-favorable electronic circuitry or data processing.
Another object of the invention is to make available a device and process for improved signal analysis which is especially suitable for detection of an early stage of incipient bearing or machinery damage.
It is also the object of the invention to devise a device and process by which improved, more effective analysis or detection is possible for recognition of those signal portions which are characterized by only a brief presence in time, i.e., transient signal effects. The process in accordance with the invention can, moreover, be used, for example, for detection of very noisy signals.
The invention is based on the finding that a sampled time signal usually present in digitized form can be interpreted as an ordered set of individual events with an inner correlation which is initially unknown. The invention is furthermore based on the finding that the results of conventional, known correlation analyses can be supplemented and improved by abandoning the practice of combining with one another of sections of these signals which are shifted relative to one another in time. Rather, in accordance with the invention, each individual event (according to the definition above) of a signal to be examined is combined at least once with one or more other individual events, which are separated from each other by a time difference. Then, the results of the individual combinations are sent for classification. In particular, it is of interest according to the invention to provide for classification purposes one such combination in the manner of a distribution which has been formed by summation. The x-axis of this distribution is preferably arranged by time classes. The individual time classes correspond to the time differences so that the results of the individual combinations can be fitted into the distribution using the underlying time differences. It is advantageous for combination of the events to perform multiplications each time. However, for this purpose, other binary logic operations can be used, for example, summation or determining a maximum.
According to one preferred approach of the invention, it is provided that each event be combined with all its predecessors or with those predecessors which are located in its immediate vicinity. According to another approach of the invention, it is provided that each event be combined with any other and the results of the combination likewise be sent for classification.
According to still another implementation of the invention, it is provided that events of predefined “quality” (i.e. having predefined characteristics) be preselected for constituting a “filtered” signal, consisting of such specific events which then are to be combined with any other, and the results of the combination be sent for classification in a comparable way.
In accordance with the invention, the classification results obtained by a distribution analysis can be evaluated either directly or can be sent for additional analysis. Here, it is especially useful to normalize those extreme values of a resulting frequency distribution by a mapping rule which can be interpreted to relate to an integral multiple of a time difference between two predetermined extreme values. One such extreme value, therefore, belongs to a time class of the distribution which is characterized by a smaller (difference) time value than the time classes of the integral multiple of the first extreme value.
In the direct evaluation of the resulting classification result (i.e. classification function), it can be enough to check the function values of individual time classes with reference to a predefined threshold value. Another direct evaluation involves, for example, checking the observed extreme values, especially maximum values, of a resulting distribution. According t
Kwok Helen
Nixon & Peabody LLP
Pruftechnik Dieter Busch AG
Safran David S.
Saint-Surin Jacques
LandOfFree
Device and process for signal analysis does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Device and process for signal analysis, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Device and process for signal analysis will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3077410