JPH0552712A - Fault precognition apparatus for production machine - Google Patents

Abstract

PURPOSE:To perform preventive maintenance for a production machine by detecting the vibration pattern of a moving member constituting the production machine. CONSTITUTION:The waveform from a vibration sensory which is provided at the working end of a robot, i.e., the free end, is converted into the digital signal with an a A/D converter 4. The power spectrum is obtained from the digital signal with an FFT means 5. In a judging means 6, the digital data outputted from the A/D converter 4 and the power spectrum obtained with the FFT means are compared with the data and the spectrum, which are obtained at the normal time and stored in a memory circuit 7. When the detected values are largely different from the values at the normal time in this comparison, the information is displayed a display means 9 and an alarm generating means 10.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a failure prediction device for predicting a failure of a production machine, and more particularly to a failure prediction device suitable for use in a multi-axis type industrial robot.

[0002]

2. Description of the Related Art Although an industrial robot has been introduced into a production process for a long time, various preventive maintenance is usually performed for this robot having a large number of constituent parts in order to improve the operating rate. One of the preventive maintenance is preventive maintenance of the mechanical system around the rotating shaft. In an industrial robot, this part is the most fatigued, and once a failure occurs in this part, repairing it takes a great deal of time and money. For this reason, conventionally, in performing this preventive maintenance in an industrial robot having 6 degrees of freedom as shown in FIG. 1, vibration sensors 20 to 24 are attached to the joints forming the first joint to the sixth joint, respectively.
A single vibration pattern can be obtained at each joint. The signal from the vibration sensor is input to the FFT 25 and subjected to the fast Fourier transform processing, and the waveform after the processing is displayed on the oscilloscope 26. On the basis of the waveform output from the oscilloscope 26, the maintenance staff applies the general rotating machine vibration diagnosis theory to determine that the bearing is abnormal, the bearing is becoming loose, or the like.

[0003]

However, in such a conventional fault diagnosis device, since the general rotary machine vibration theory is applied to perform the fault diagnosis, the vibration is detected for each joint. It is necessary to attach a sensor, and it is necessary to independently drive the joints one by one to obtain a vibration pattern for each joint. In other words, this theory is effective for failure diagnosis of rotating equipment. Further, since the failure diagnosis based on the above general theory of rotation mechanical vibration is effective when the driven body rotates at a high speed, normally, for the failure diagnosis of the axis of the robot that moves far from the high speed rotation, Although a certain degree of tendency can be recognized, its accuracy is not always good. Therefore, in reality, it is difficult to understand the deterioration status of the bearing of the shaft of the robot, and it has always been the case that the failure prediction is diagnosed at the time when the rotation of the shaft begins to be hindered. Further, since the failure diagnosis is performed based on the statistical processing of past results, accurate diagnosis cannot be performed for a production machine having a small amount of data as past results. The present invention has been made in view of the above problems of the prior art, and detects a vibration pattern of each joint by a vibration sensor provided at a free end of an industrial robot, and detects the detected vibration pattern. An object of the present invention is to provide a failure prediction device that predicts a failure of a production machine based on the above.

[0004]

According to the present invention for achieving the above object, a vibration pattern for each moving part generated from a plurality of moving parts constituting a production machine is detected at a free end of the moving part. Vibration detecting means, storage means for storing a reference vibration pattern at the free end during normal operation of the production machine, and vibration pattern detected at any time by the vibration detecting means in the storage means. And a failure prediction means for predicting a failure of the production machine as compared with a reference vibration pattern that is present.

[0005]

The present invention thus constructed operates as follows. The vibration detecting means detects the vibration pattern generated from each moving part of the production machine at any free end of the moving part. For example, in the case of an industrial robot, its vibration detecting means is attached to the arm working end of the robot. The storage means stores the vibration pattern detected at its free end during normal operation of the production machine as a reference vibration pattern. This reference vibration pattern is a vibration pattern for each moving part obtained when each moving part is independently moved. The failure prediction means compares the vibration pattern detected by the vibration detection means when a certain moving part is moved with the reference vibration pattern for that part stored in the storage means, and if the difference in the pattern is large, it is close to the reference vibration pattern. Predict that failures will occur in the future. By this prediction, preventive maintenance can be performed, and it becomes possible to improve the operating rate and reliability of the production machine.

[0006]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of the present invention will be described below with reference to the drawings. FIG. 2 is an external view of an industrial robot as an example of a production machine to which a vibration sensor is attached when performing failure prediction, and FIGS. 3A and 3B are side views of a tool attachment to which the vibration sensor is attached. It is a figure and a front view. In this embodiment, an industrial robot as shown in FIG. 2 will be described as an example of the production machine. The vibration sensor 1 is a sensor that detects acceleration, and is attached to the working end of the arm of the robot 12, that is, the tip of the wrist rotation shaft, as shown in the figure. The tip of the wrist rotation shaft is a portion that moves if any arm of the robot 12 moves, and is referred to as a free end. If a vibration sensor is attached to this free end, it is possible to capture the vibration waveform generated when each arm moves independently.

The vibration sensor 1 is not directly attached to the arm of the robot 12, but is attached to the tip of its rotation shaft via a tool attachment 14 as shown in FIG. This tool attachment 14
Is attached in place of the work tool attached to the wrist rotation tip of the robot 12 when attempting preventive maintenance of the robot 12. The vibration sensor 1 does not have the characteristic of being able to detect acceleration in any direction, that is, vibration, and the detection direction is usually determined to avoid the influence of gravity, so vibration can be detected. Specifically, specifically, the position of the vibration sensor 1 is specified by moving each arm of the robot 12 so that the movement direction is the tangential direction of the vibration sensor 1. By doing so, it is possible to eliminate the adverse effect of the backlash phenomenon of other mechanical components as much as possible. Then, for example, in the case of performing preventive maintenance of the wrist shaft, the wrist is moved so that the vibration sensor 1 is located at the position shown in FIG. 2, and the position of the vibration sensor 1 is maintained so that the wrist shaft is maintained. Only slowly rotates in the horizontal direction, and the vibration pattern output from the vibration sensor 1 at that time is output to the device described later. Since the operation at this time is taught in advance, when performing the preventive maintenance, it is only necessary to instruct the playback of the taught operation. Further, when reciprocating instead of rotating, for example, the vibration pattern is measured in a stable region during forward movement. Although this vibration sensor 1 is not shown in the figure, it is the same as in FIG.
It is connected to the FT25.

FIG. 4 is a block diagram of the preventive diagnosis apparatus according to the present invention. The vibration sensor 1 is attached to the wrist rotation shaft end of the robot 12 via the tool attachment 14 as described above. From the vibration sensor 1, the rotation of each axis constituting the robot 12 or An analog signal is output according to the degree of vibration of the end of the rotating shaft that vibrates with the moving operation. This analog signal is, for example, a signal that changes with time as shown at the left end of the lower left waveform diagram in FIG. The analog waveform output from the vibration sensor 1 is recorded by the analog vibrometer 2. The trigger circuit 3 detects a signal output from a control device of the robot 12 (not shown) in a stable region on the forward path taught for vibration measurement. The reason why the stable region is provided in this way is that, for example, a stable vibration state is not obtained in the initial stage of movement of the shaft, so that the vibration pattern in such an unstable region is not taken into consideration as failure prediction data. Is. Analog vibrometer 2
, The analog waveform from the vibration sensor is always input, but the conversion of this waveform by the A / D converter 4 is started when the signal is output from the trigger circuit 3.
That is, from the time when the control device of the robot 12 outputs a signal that the stable range has been reached. The waveform diagram after being converted into a digital signal by the A / D converter 4 becomes a digital signal which changes with time as shown in the center of the waveform diagram, for example. The FFT means 5 performs high-speed Fourier transform on the digital signal output from the A / D converter 4 to calculate the power spectrum. For example, when the waveform shown in the center of the waveform chart is subjected to the fast Fourier transform, the waveform of the spectrum distribution of the frequency components included in the waveform, that is, the power spectrum as shown at the right end of the waveform is obtained.

The judging means 6 is based on the digital signal output from the A / D converter 4 and the power spectrum obtained by the FFT means 5 and, via the personal computer interface 8, an average value and a peak value thereof. The previous value stored in the storage circuit 7 and the reference value, which is the data obtained in the normal time, are compared with each other to determine the bearing abnormality, the gear abnormality, and the wear deterioration related to the harmonic speed reducer. The memory circuit 7 stores, as a reference vibration pattern, the A / D-converted vibration waveform obtained when each axis is moved and the data regarding the previous vibration waveform obtained by the movement of each axis. There is. The reference vibration pattern may be updated to a normal vibration pattern detected every arbitrary period. The bearing abnormality and the gear abnormality are determined based on the digital digital signal output from the A / D converter 4, and the wear deterioration related to the harmonic speed reducer is determined based on the power spectrum output from the FFT means 6. It This judgment can be automatically made by providing the judging means 6 with a comparator function for comparing the area of the level value and the peak value of the designated spectral bandwidth. The display means 9 is, for example, L
It is like a CD display and has an alarm generating means 10.
Is, for example, a buzzer or a lamp. When the judgment means 6 judges that there is an abnormality, a display to that effect is displayed on the display means 9 and an alarm or visual alarm is displayed by the alarm generation means. The data measured this time is stored in the storage circuit 7, but this data is stored and managed in an external storage device via the personal computer interface 8. Also, the FFT means 5 and the judging means 6 may be constituted by a microprocessor.

Next, how to actually use the failure prediction device for an industrial machine according to the present invention based on the data when the bearing abnormality of the robot wrist rotary joint mechanism and the situation when the harmonic reducer is worn are detected. Then, it will be explained whether an abnormality is detected. The waveform shown in FIG. 5A is a normal vibration waveform output from the vibration sensor 1 to the analog vibrometer 2 when only a certain axis is moved. Also, FIG.
The waveform shown in is the power spectrum obtained by the normal waveform at the time of the vibration. This power spectrum was obtained in the 500HZ range, but has a waveform with a peak near 150HZ. The fundamental rotation frequency at this time is about 2HZ. 6 (a) and 6 (b)
[Fig. 4] is a waveform and power spectrum obtained when one of the rolling elements of the wrist rotary joint mechanism is slightly scratched. As is clear from this figure, the vibration sensor 1
The detection level of the vibration waveform detected by is larger than that of the normal vibration waveform. It can also be seen that the power spectrum has a higher level at the same frequency than that detected in the normal state. The frequency due to the scratches on the rolling elements at this time is 9 Hz. When such a waveform is inputted to the judging means 6, the judging means 6 compares the waveform shown in FIG. 5 stored in the memory circuit 7 with the inputted waveform. Specifically, the average value and the peak value of the levels obtained from the respective waveforms are compared with each other. Whether these values are abnormal or normal is determined depending on whether or not these values deviate significantly from the values obtained under normal conditions. 7 (a) and 7 (b)
Are the waveforms and power spectra obtained when the harmonic reducer of the wrist rotary joint mechanism has a backlash of 1.5 minutes in lost motion. It can be seen from this figure that the waveform obtained by the vibration sensor 1 has a longer cycle than the waveform obtained under normal conditions. In the case of the power spectrum, the peak frequency is around 100HZ. Is to shift to. Since this peak shift is 30% or more, if the peak in this frequency band can be detected, lost motion can be easily detected. When this diagnosis is actually carried out, it is carried out for all axes constituting the robot 12. That is, in the case of diagnosing a 6-axis robot as shown in FIG. 1, first, only the first joint is moved for this measurement, then only the second joint is moved, and then only the third joint is called. Only one joint is sequentially moved in order, and the above waveform is taken for all joints. The movement locus of the robot at the time of this diagnosis is taught, and of course the posture of the vibration sensor 1 can be maintained at any posture regardless of the movement of each joint. As described above, the waveform is taken in the stable region on each trajectory.

The above three examples are the detection results when the same portion is normal, when the bearing is abnormal, and when there is play in the speed reducer. The differences are the average value of the vibration waveform, the peak value, and the peak of the power spectrum. It can be clearly distinguished from shifts and the like. Therefore, by managing these data, robot diagnosis by trend management can be performed,
As a result, preventive maintenance can be performed.
In the above embodiments, the case of detecting the abnormality of each axis of the industrial robot has been described as an example. However, the present invention is not limited to this, and the present invention can be applied to a commonly used machine tool or the like. it can. Further, the data obtained by this diagnosis may be temporarily stored in an IC card or the like on the spot, and the computer in the management department may read the data of the IC card for automatic analysis or precise diagnosis. This is very easy because it is not necessary to bring a diagnostic device to a narrow site. By using the device of the present invention, it is possible to easily verify the restoration accuracy after repair by preventive maintenance,
You will always be in the best condition.

[0012]

As described above, according to the present invention, the vibration pattern generated from each moving part of the production machine is detected by the vibration sensor provided at its free end, and the detected vibration pattern is I tried to compare it with
Preventive maintenance can be performed even if you are not a specialist in equipment diagnosis.

[Brief description of drawings]

FIG. 1 is a configuration diagram of a device for detecting a vibration pattern in a conventional industrial robot.

FIG. 2 is an external view of an industrial robot as an example of a production machine to which a vibration sensor is attached when failure prediction is performed.

3 (A) and 3 (B) are a side view and a front view of a tool attachment to which a vibration sensor is attached.

FIG. 4 is a configuration block diagram of a signal processing system of the failure prediction device according to the present invention.

5A is a vibration waveform in a normal state, and FIG. 5B is a power spectrum obtained by the normal waveform in a vibrating state.

6 (a) and 6 (b) are waveforms and power spectra obtained when one of the bearing rolling elements of the wrist rotary joint mechanism is slightly scratched.

7 (a) and 7 (b) are waveforms and power spectra obtained when the harmonic reducer of the wrist rotary joint mechanism has a backlash of 1.5 minutes in lost motion.

[Explanation of symbols]

 1 ... Vibration sensor (vibration detection means) 2 ... Analog vibrometer (vibration detection means) 3 ... Trigger circuit 4 ... A / D converter (vibration detection means) 5 ... FFT means (vibration detection means) 6 ... Judgment means (failure) Prediction device 7 ... Storage circuit (storage means) 12 ... Industrial robot (production machine) 14 ... Tool attachment 20-24 ... Vibration sensor 25 ... FFT 26 ... Oscilloscope

Claims (1)

[Claims] 1. A vibration detecting means for detecting a vibration pattern of each moving part generated from a plurality of moving parts constituting a production machine at a free end of the moving part, and a vibration detecting means during normal operation of the manufacturing machine. A storage unit that stores a reference vibration pattern at the free end and a vibration pattern that is detected by the vibration detection unit at any time are compared with a reference vibration pattern that is stored in the storage unit to predict failure of the production machine. And a failure prediction device for a production machine.