Method for setting alarm levels for a machine

Abstract

A method for setting alarm levels for a machine provides defining at least one condition indicator reflecting the condition of the machine with respect to a defect to be monitored of the machine, the at least one condition indicator defined from machine kinematic data, recording measurements of process related parameters during a predetermined period during which the machine is operating normally, calculating at least one condition indicator value for the at least one condition indicator) using machine condition data, determining a graph of the at least one condition indicator value as a function of a first process related parameter chosen from the measured process related parameters, dividing the graph into operating classes, each operating class being representative of different operating conditions of the machine, calculating an alarm level value for each operating class, setting the determined alarm level value at the midpoint of each operating class.

Claims

1. Method for setting alarm levels for a machine, the method comprising: defining at least one condition indicator (CI) reflecting the condition of the machine with respect to a defect to be monitored of the machine, the at least one condition indicator being defined from machine kinematic data, recording measurements of machine condition data and of process related parameters during a predetermined period during which the machine is operating normally, calculating condition indicator values for the at least one condition indicator for each recorded machine condition data measurement, determining a graph representing the at least one condition indicator value, on a vertical axis, as a function of a first process related parameter chosen from the measured process related parameters, dividing the graph into operating classes along a horizontal axis, each operating class being representative of a different operating condition of the machine and having upper and lower boundaries for each operating class, calculating an alarm level value for each operating class based on the statistical characteristics of the condition indicator values for each operating class, and wherein setting the determined alarm level value in the graph at a horizontal axis midpoint of each operating class.

2. The method according to claim 1, further comprising connecting the alarm level values in the graph by linear interpolations.

3. The method according to claim 1, wherein the alarm level value of each operating class is calculated from a mean value and a standard deviation of the at least one condition indicator values in the considered operating class and a detection factor.

4. The method according to claim 1, wherein dividing the graph into operating classes comprises determining the lower and upper boundaries of each operating class such that the variations of the at least one condition indicator values in the considered operating class are smaller than an operating class definition threshold.

5. The method according to claim 1, wherein when the learning period is over, if M out of N values of the at least one condition indicator values are above an alarm level value, an alarm is triggered.

6. The method according to claim 1, wherein the process related parameters comprise a speed of the machine and/or a load applied on the machine and/or vibrations of the machine.

7. A system for setting alarm levels for a machine comprising: the system configured to define at least one condition indicator reflecting the condition of the machine with respect to a defect to be monitored of the machine, the at least one condition indicator being defined from the measurements of machine kinematic data, the system further configured to record, by one or more sensors, measurements of machine condition data and of process related parameters during a predetermined period during which the machine is operating normally, the system further configured to perform a first calculation to calculate condition indicator values for the at least one condition indicator for each recorded machine condition data measurement, the system further configured to determine a graph representing the at least one condition indicator value, on a vertical axis, as a function of a first process related parameter chosen from the measured process related parameters, the system further configured to divide the graph into operating classes along a horizontal axis, each operating class being representative of a different operating condition of the machine and having upper and lower boundaries for each operating class, the system further configured to perform a second calculation to calculate an alarm level value for each operating class based on the statistical characteristics of the condition indicator values for each operating class, the system further configured to set the determined alarm level value in the graph at a horizontal axis midpoint of each operating class.

8. The system according to claim 7, further comprising the system further configured to perform an interpolation to connect the alarm level values in the graph by linear interpolations.

Description

BRIEF DESCRIPTION OF THE DRAWINGS

(1) Other advantages and features of the invention will appear on examination of the detailed description of embodiments, in no way restrictive, and the appended drawings in which:

(2) FIG. 1 illustrates schematically an example of an embodiment of a machine according to the invention;

(3) FIG. 2 illustrates an embodiment of a method for setting alarm levels for the machine according to the invention; and

(4) FIG. 3 illustrates an example of a graph of a condition indicator.

DETAILED DESCRIPTION OF THE INVENTION

(5) Reference is made to FIG. 1 which represents an example of an embodiment of a machine 1 comprising sensors 2 and a condition monitoring system 3 connected to the sensors 2.

(6) The sensors 2 comprise at least one machine condition sensor 4 and at least one process related parameters sensor 5.

(7) The machine condition sensor 4 generates machine condition data of the machine 1, for example the power output of the machine 1.

(8) The machine condition sensor 4 comprises for example a power sensor.

(9) The process related parameters sensor 5 measure process related parameters of the machine 1.

(10) Process related parameters comprise for example operating speed, load or vibrations.

(11) The sensors 5 comprise for example speed sensors, load sensors and/or sensors configured to measure vibrations applied on the machine 1.

(12) The condition monitoring system 3 comprises defining means, recording means, first calculating means, determining means, dividing means, second calculating means, setting means and interpolation means.

(13) FIG. 2 represents an embodiment of a method for setting alarm levels for the machine 1.

(14) In a step 10, the defining means define at least one condition indicator CI reflecting the condition of the machine with respect to a defect of the machine 1 to be monitored.

(15) The condition indicator is defined from machine kinematic data.

(16) The condition indicator CI comprises for example a gear mesh frequency of a gear of the machine 1 equal to a gear main shaft speed times the number of teeth of a gear bearing, the number of teeth of the gear bearing being a kinematic data of the machine 1.

(17) In step 11, during a learning period in which the machine 1 is operating normally, measurements of the process related parameters are recorded by the recording means for a predetermined period.

(18) During the learning period, the machine 1 is operating in all operating conditions that the machine is normally used for.

(19) In step 12, after the learning period, the first calculating means calculate condition indicator values for the condition indicator CI for each recorded machine condition data measurement.

(20) In step 13, the determining means determine a graph GR representing the condition indicator values in function of a first process related parameter chosen from the measured process related parameters. It is assumed that in the following, the first process related parameter is the operating speed of the machine 1.

(21) Then, in step 14, the dividing means divide the graph GR into operating classes OC, each operating class being representative of different operating conditions of the machine.

(22) The lower and upper boundaries of each operating class are determined such that the variations of the condition indicator values in the considered operating class are smaller than an operating class definition threshold.

(23) For example, the operating class definition threshold is equal to 10% of the mean value calculated on all CI values obtained during the learning period.

(24) The graph GR is for example divided into 5 to 10 operating classes, each operating class comprises for example 10 to 20 measurements.

(25) In step 15, the second calculating means calculate an alarm level value AL for each operating class which is equal to:
AL=μ+X.σ  (1)

(26) where μ is the mean value and σ is the standard deviation of the condition indicator values in the considered operating class, and X is a detection factor.

(27) X is for example comprised between 1 and 10.

(28) In step 16, the determined alarm level value is set at the midpoint of each operating class by the setting means.

(29) In step 17, the interpolation means connect the alarm level values together by linear interpolation increasing even more the accuracy of the alarm level between the midpoints of the operating classes.

(30) The condition monitoring system 3 categorizes the operating conditions in multiple operating classes depending on process related parameter being monitored.

(31) FIG. 3 represents an example of the graph GR comprising the condition indicator CI versus the first process related parameter PP1.

(32) The measurements used to define the condition indicator CI are represented by dots.

(33) The graph GR is divided into 7 operating classes OC1 to OC7.

(34) Each operating classes OC1 to OC7 comprises at its midpoint a calculated alarm level value AL1 to AL7 represented by a cross, the alarm level values AL1 to AL7 being connected together by linear interpolation.

(35) When the machine 1 is operating and the learning period is over, if M out of N values of the condition indicator values are above the alarm level value AL1 to AL7, an alarm is triggered, M and N being integers.

(36) For example, M is equal to 4 and N is equal to 7.

(37) The alarm levels are automatically set for different operating conditions without any need of manual input, each operating condition being represented by an operating class on the graph GR.

(38) An alarm level is determined for each operating class enhancing the accuracy of the alarm level to detect a defect of the machine 1 at an early stage avoiding false alarms.

(39) In the illustrated example, one condition indicator reflecting the condition of the machine with respect to a defect to be monitored of the machine is analyzed.

(40) In order to detect more defects, more graphs divided into operating classes and comprising alarm level values are set up as explain above, each graph comprising a different condition indicator reflecting the condition of the machine with respect to a different defect to be monitored of the machine versus the first process related parameter PP1.