Method for predicting life span of rotary machine used in manufacturing apparatus and life predicting system

• US 6,885,972 B2
• Filed: 08/28/2002
• Issued: 04/26/2005
• Est. Priority Date: 08/31/2001
• Status: Expired due to Fees

• Current Assignee: Kabushiki Kaisha Toshiba (Toshiba Corporation)
• Original Assignee: Kabushiki Kaisha Toshiba (Toshiba Corporation)
• Inventors: Ishii, Ken, Furuhata, Takeo, Nakao, Takashi, Samata, Shuichi, Ushiku, Yukihiro
• Primary Examiner(s): ASSOUAD, PATRICK J
• Application Number:
  US10/229,005
• Publication Number:
  US 20030153997A1
• Time in Patent Office: 972 Days
• Field of Search: 702/34, 702/35, 702/36, 702/182, 702/183, 702/184, 702/185, 700/29, 700/30, 700/32, 700/108, 700/121, 700/123, 700/174, 700/175, 700/177, 700/266, 700/287
• US Class Current: 702/182
• CPC Class Codes: G05B 19/4065 Monitoring tool breakage, life or condition
• CPC Class Codes: G05B 2219/37252 Life of tool, service life, decay, wear estimation
• CPC Class Codes: Y02P 90/80 Management or planning

Abstract

A method for predicting life span of a rotary machine used in a manufacturing apparatus, includes: measuring rotary machine acceleration evaluation time series data with a sampling interval being less than a half the cycle of an analysis target frequency, a number of samplings being at least four times the analysis target frequency; generating evaluation diagnosis data based on variations in characteristics corresponding to the analysis target frequency by subjecting the evaluation time series data to frequency analysis; and determining the life span of the rotary machine using the evaluation diagnosis data.

Claims

1. A method for predicting a life span of a rotary machine used in a manufacturing apparatus, comprising:

measuring reference time series data of an acceleration of the rotary machine and evaluation time series data of the acceleration respectively, with a sampling interval of less than a half the reciprocal of an analysis target frequency, wherein the sampling interval is a time between samples within a single sampling sequence, a number of samples being at least four times a numerical value of the analysis target frequency, the rotary machine being a dry used in the manufacturing apparatus;

generating reference diagnosis data based on variations in a characteristic corresponding to the analysis target frequency by subjecting the reference time series data to a frequency analysis, the characteristic representing a peak acceleration value;

generating evaluation diagnosis data based on variations in a peak value by subjecting the evaluation time series data to the frequency analysis; and

forming a Mahalanobis space with the reference diagnosis data obtained from the reference time series data measured at a time that is earlier than the time where the evaluation time series data is measured by a predetermined length of time determined through an empirical rule under the same process conditions as the evaluation time series data, calculating a Mahalanobis distance of the evaluation diagnosis data based on the Mahalanobis space, and determining a time point where the Mahalanobis distance exceeds a threshold value as just before the end of the life span.

2. A method for predicting a life span of a rotary machine used in a manufacturing apparatus, comprisinq:

measuring reference time series data of an acceleration of the rotary machine and evaluation time series data of the acceleration respectively, with a sampling interval of less than a half the reciprocal of an analysis target frequency, wherein the sampling interval is a time between samples within a single sampling sequence, a number of samples being at least four times a numerical value of the analysis target frequency, the rotary machine being a dry pump used in the manufacturing apparatus;

generating reference diagnosis data based on variations in a characteristic corresponding to the analysis target frequency by subjecting the reference time series data to a frequency analysis, the characteristic representing a peak acceleration value;

generating evaluation diagnosis data based on variations in a peak value by subjecting the evaluation time series data to the frequency analysis; and

forming a Mahalanobis space with the reference diagnosis data obtained from the reference time series data measured at a time that is earlier than the time where the evaluation time series data is measured by a length of time determined through an empirical rule under the same process conditions as the evaluation time series data, calculating a Mahalanobis distance of the evaluation diagnosis data based on that Mahalanobis space, and determining a time point where that Mahalanobis distance exceeds a threshold value as just before the end of the life span.

3. A life predicting system, comprising:

a rotary machine;

an accelerometer for sampling and measuring reference time series data for an acceleration of the rotary machine and evaluation time series data of the acceleration respectively, with a sampling interval being less than half a reciprocal of an analysis target frequency, wherein the sampling interval is a time between samples within a single sampling sequence, a number of samples being at least four times a numerical value of the analysis target frequency,

a frequency analysis device for performing a frequency analysis on an output from the accelerometer;

a peak acceleration transition recording module for generating reference diagnosis data based on variations in a characteristic corresponding to the analysis target frequency from the reference time series data, generating evaluation diagnosis data based on variations in a peak value from the evaluation time series data, with results of the frequency analysis, and recording the reference diagnosis data and the evaluation diagnosis data; and

a life span determination unit for determining life span of the rotary machine using the reference diagnosis data and the evaluation diagnosis data, so as to form a Mahalanobis space with the reference diagnosis data obtained from the reference time series data measured at a time that is earlier than the time when the evaluation time series data is measured by a length of time determined through an empirical rule under the same process conditions as the evaluation time series data, calculate a Mahalanobis distance of the evaluation diagnosis data based on said Mahalanobis space, and determine a time point when said Mahalanobis distance exceeds a threshold value just before the end of the life span.

4. The system of claim 3, wherein the rotary machine is a dry pump used in semiconductor manufacturing apparatus.

5. The system of claim 4, wherein the analysis target frequency is fixed at an integral multiple of a reference frequency unique to the dry pump.

6. The system of claim 4, wherein the characteristic represents a peak acceleration value.

7. The system of claim 4, wherein the characteristic represents a frequency at the peak acceleration.

8. The system of claim 4, wherein the accelerometer is placed on a top surface of a gear box of the dry pump, on a side surface of the gear box, on a bottom surface of the gear box, on a top surface of a casing within 20 cm from the gear box, on a side surface of the casing, on a bottom surface of the casing.