Machine monitoring method

• US 4,989,159 A
• Filed: 02/15/1989
• Issued: 01/29/1991
• Est. Priority Date: 10/13/1988
• Status: Expired due to Fees

• Current Assignee: Jan Liszka-Hackzell, Ludwik Liszka
• Original Assignee: Jan Liszka-Hackzell, Ludwik Liszka
• Inventors: Liszka-Hackzell, Jan, Liszka, Ludwik
• Primary Examiner(s): Lall, Parshotam S.
• Assistant Examiner(s): ZANELLI, MICHAEL J
• Application Number:
  US07/311,490
• Time in Patent Office: 713 Days
• Field of Search: 364/505, 364/506, 364/507, 364/508, 364/554, 364/581, 364/550, 364/551.01, 364/485, 73/658, 73/659, 73/660, 382/30
• US Class Current: 702/56
• CPC Class Codes: G01H 1/003 of rotating machines G01H1/10 takes precedence
• CPC Class Codes: G07C 3/00 Registering or indicating the condition or the working of machines or other apparatus, other than vehicles

Abstract

Method of monitoring the operational state of a machine, particularly one having two mutually co-acting rotating machine parts operating at different revolutionary rates. Sensed vibration spectra are compared with theoretically calculated spectra and the peaks in the respective spectra are mutually matched. Each match or assignation is alloted an adjustment weight, which is specific for a given machine part and which increases with the amplitude of the peak and decreases with the frequency distance between the peaks. The adjustment weights are summed into weights which are attributable to the respective machine part. The part weights form weight vectors, and a reference class is formed on the basis of a plurality of weight vectors obtained during normal operation. Each new weight vector is compared with the reference class, and it is then determined whether the difference exceeds a predetermined statistical spread value.

Claims

1. A method of continuously monitoring the operational state of a machine, particularly a complicated machine with at least two rotating parts, working at different revolutionary rates and in mutual co-action, sensed vibration states being processed by analysis of frequency spectra while utilizing sampling and pattern recognition techniques, abnormal operational conditions being detected by calculating the probability of a sensed vibration state differing significantly from normal operational states, which are represented by a reference class calculated on the basis of previously sensed vibration states during normal operation of the machine, said method comprising the steps of:

(a) calculating expected peaks in a vibration spectrum theoretically for each vibration-generating machine part or partial system in the machine and the occurring revolutionary rates;

(b) selecting a plurality of peaks in the respective vibration spectrum during continuous sensing of the actual vibration states;

(c) matching at least one of said expected peaks to each selected actual peak, and assigning to each such match an adjustment weight, which increases with the amplitude of the actual peak and decreases with the frequency distance between the actual and its matched calculated peak;

(d) summing all adjusting weights associated with a given machine part or partial system to form a part weight;

(e) forming a weight vector corresponding to the machine in its entirety, the components of this vector comprising said part weights; and

,

(f) forming said reference class on the basis of a plurality of weight vectors, each new weight vector being compared with said reference class, whereupon it is determined whether the difference exceeds a predetermined statistical spread value.

2. A method as claimed in claim 1, wherein said adjustment weight is substantially proportional to a height of the actual peak above a background level and substantially inversely proportional to said frequency distance.

3. A method as claimed in claim 1, wherein separate measurements are made within different frequency ranges, the part weights originating from these different frequency ranges being mutually combined while taking into account the resolution in the respective frequency range.

4. A method as claimed in claim 3, wherein part weights originating from the frequency ranges with relatively high resolution are given a greater share in the respective combined weight than part weights originating from frequency ranges with relatively low resolution.

5. A method as claimed in claim 4, wherein the part weights originating from different frequency ranges are combined by summing after multiplication of each part weight by a factor 1/B, where B is the bandwidth corresponding to the resolution in the respective frequency range.

6. A method as claimed in any one of claims 1 to 5, wherein in detecting an abnormal operational state, a fault diagnosis is made by determined what part weights in the weight vector in question substantially contribute to the statistical deviation, these part weights or partial vectors being attributable to individual machine parts or partial systems in the machine.