TWM604901U - Intelligent monitoring system applied to grinding process - Google Patents

Abstract

An intelligent monitoring system applied to the polishing process includes a polishing module, a sensing module, a processing module, and a comparison module. The sensing module is used to monitor the dynamic signal generated by the action of the grinding module, the processing module converts the dynamic signal into a characteristic time map, and captures a repetitive target interval in the characteristic time map , And then store the target interval as a visualized map. The comparison module compares the visualized map with the real-time characteristic time map of the grinding module to obtain a judgment result. When the judgment result does not exceed one When the limit value is reached, the polishing module continues to operate, and when the judgment result exceeds the limit value, the processing module controls the polishing module to stop operating.

Description

Intelligent monitoring system applied to grinding process

This model relates to an intelligent monitoring system, especially an intelligent monitoring system applied to the grinding process.

With the rapid development of AI artificial intelligence, real-time monitoring and preventive perception of abnormalities in automated equipment or products are also new topics in industrial technology, and solutions can be completed by machine learning methods. The AI monitoring technology currently seen on the market mainly uses different types of sensors to capture a large amount of data, and then manually classifies whether the equipment is activated, process parameters, product types, etc., and then combines the manual records of the production line to collect The data is filtered, labeled, and then AI training is performed to obtain the so-called monitoring model.

However, before performing the monitoring task, a large amount of data must be collected by the sensor, and then the data must be collected and integrated with the production line information manually, which requires a lot of manpower and time, and then specific events such as machine damage Label features, machine quality, maintenance status, etc., before completing the pre-actions of AI training, and the subsequent training results will be based on the quality of artificial cleaning data. If the quality is not good, the model will not converge or the monitoring accuracy will be poor. , And need to be improved.

The above shows that AI monitoring is difficult to implement various problems, and mechanical damage is gradual. If minor mechanical failure characteristics can be found in the early stage, it will have a certain degree of benefit for product yield and production line capacity arrangements.

In view of this, the purpose of the present invention is to provide an intelligent monitoring system applied to the polishing process, which includes a polishing module, a sensing module, a processing module, and a comparison module.

The sensing module is connected with the grinding module signal to monitor the dynamic signal generated by the action of the grinding module, and the processing module is connected with the sensing module signal to convert the dynamic signal into a feature Time map, and capture a repetitive target interval in the characteristic time map, and then store the target interval as a visualization map. The comparison module is connected to the processing module signal, and the comparison module uses the visualization The graph is compared with the real-time characteristic time graph of the grinding module to obtain a judgment result. When the judgment result does not exceed a limit value, the grinding module continues to act, and when the judgment result exceeds the limit value , The processing module controls the polishing module to stop operating.

Preferably, the target interval is a result obtained by artificial judgment based on the display content of the characteristic time map.

Preferably, the visualized map is decomposed into dynamic signal similarity correlation, amplitude correlation, frequency correlation, phase correlation, short-term change correlation, glitch correlation and other characteristic value maps by different mathematical formulas.

Preferably, the comparison module includes an actuation database storing a plurality of characteristic time maps of the grinding module at different time points, and a plurality of judgment results at different time points obtained from the characteristic time maps of different time points.

Preferably, the action database further stores a plurality of state trend graphs made with characteristic time maps at different time points. The state trend graph includes a dynamic signal similarity graph, an amplitude graph, a frequency graph, and a Phase diagram, a short-term change diagram, a burst diagram, etc.

Preferably, the intelligent monitoring system applied to the polishing process further includes a learning module connected with the signal of the comparison module, and the learning module executes a supervised learning method to use the feature value map of the visualization map and The dynamic signal similarity graph, the amplitude graph, the frequency graph, the phase graph, the short-term change graph, and the burst graph are compared to generate a difference value. When the difference value result is lower than the limit value, The processing module sends out a predictive signal.

Preferably, the sensing module includes a sensor disposed on the polishing module, and the sensor is selected from accelerometers, multiaxial accelerometers, gyroscopes, strain gauges, pressure sensing, One or a combination of thermometer, voltmeter, ammeter, etc.

Preferably, the intelligent monitoring system applied to the grinding process further includes a warning module connected to the processing module signal, and when the judgment result exceeds the limit value, the processing module sends out a warning through the warning module Warning signal.

Preferably, the sensing module performs feature capture and monitoring of the specific operating point of the polishing module, and the specific operating point is selected from the mechanical quality of the polishing wheel, polishing pad, rotating shaft, and Z moving axis.

Preferably, the intelligent monitoring system applied to the polishing process further includes a display module connected with the processing module signal to display the judgment result, and the judgment result records the pass or not, alarm record, and failure reason , Measurement time, dynamic similarity, amplitude, frequency, surge changes and other information.

The beneficial effect of the new model is that by applying the artificial inspection threshold, not only can the equipment monitoring task be started at the same time as the installation, but also can automatically classify and clean the data on the production line, without the need for special personnel and extra time to Data labeling and time sequence classification, you only need to automatically classify the measurement data into groups and then directly learn from the learning module. At the same time, the monitoring model can be optimized online to achieve the dual functions of detection and prediction of equipment status.

The characteristics and technical content of the patent application related to this new model will be clearly presented in the following detailed description of the preferred embodiment with reference to the drawings.

Refer to Figure 1, which is a preferred embodiment of the new intelligent monitoring system applied to the polishing process, which includes a polishing module 1, a sensing module 2, a processing module 3, a comparison module 4, a Learning module 5, a warning module 6, and a display module 7. The new model has the first stage function of monitoring the real-time status of the production line machine and the second stage function of predicting machine damage.

The sensing module 2 is signal-connected to the polishing module 1 and includes a sensor 21 arranged on the polishing module 1 to monitor the dynamic signal generated by the operation of the polishing module 1. Wherein, the sensor 21 is selected from one or a combination of accelerometers, multi-axial accelerometers, gyroscopes, strain gauges, pressure sensors, thermometers, voltmeters, current meters, etc., as long as they can be used to detect signals Sensors with dynamic changes can be implemented without limitation. The action signal can be one or a combination of acceleration value, angular momentum value, strain value, pressure value, temperature value, voltage value, current value, etc., to implement manufacturing equipment with different characteristics.

The grinding process mainly uses an accelerometer (monitoring the amount of vibration) and a current-related sensor 21, which detects changes in the signal during the cycle, and there is a signal difference between the signal and the calibration cycle, which means that some abnormalities in the combined components occur, which will cause process variations. This leads to product failure.

Among them, the sensor 21 of the sensing module 2 can perform feature capture and monitoring on a specific actuation point of the polishing module 1, and the specific actuation point is selected from the group consisting of a polishing wheel, a polishing pad, a rotating shaft, and a Z moving axis machine. Quality, product quality, these specific operating point monitoring will produce abnormal spindle quality, polishing pad wear and tear, moving axis (feed) motor, reducer, drive circuit and other defective items.

The processing module 3 is connected to the sensing module 2 with signals, converts the dynamic signal into a characteristic time map, and captures a repetitive target interval in the characteristic time map, and then stores the target interval as a visualization Map, and the processing module 3 normalizes the characteristic time map to obtain the visualized map.

Wherein, the target interval is a result obtained by artificial judgment based on the display content of the characteristic time map, and the characteristic time map is a repetitive production behavior.

In addition, the visualization map is decomposed into dynamic signal similarity correlation, amplitude correlation, frequency correlation, phase correlation, short-term change correlation, glitch correlation and other eigenvalue maps using different mathematical algorithms.

The comparison module 4 is signal-connected to the processing module 3. The comparison module 4 compares the visualized map with the real-time feature time map of the grinding module 1 to obtain a judgment result. When the judgment result does not exceed a limit value, the polishing module 1 continues to operate, and when the judgment result exceeds the limit value, the processing module 3 controls the polishing module 1 to stop operating.

Wherein, the similarity comparison is the calculation of dynamic signal similarity, amplitude change, frequency change, phase difference, short-term change, surge change, etc. The limit value is 75% or more.

Refer to Figure 2, which is the first stage of the new monitoring of the real-time status of the production line machine, including the following steps.

S1: The sensor 21 arranged on the polishing module 1 monitors the dynamic signal generated by the operation of the polishing module 1.

S2: The processing module 3 converts the dynamic signal into the characteristic time map, and extracts a repetitive target interval in the characteristic time map, and stores it as the visual map. The visualized map is decomposed into at least six maps by different mathematical algorithms, including dynamic signal similarity correlation, amplitude correlation, frequency correlation, phase correlation, short-term change correlation, and burst correlation. With reference to the red dashed box in Figure 3, in order to capture a repetitive target interval in the characteristic time map, Figure 4 shows the visualized map decomposed into dynamic signal similarity correlation, amplitude correlation, frequency correlation, phase correlation, and short-term Diagrams of change correlation and surge correlation.

S3: The sensor 21 continuously measures the dynamic signal generated by the action of the grinding module 1 to obtain a real-time dynamic signal, and the processing module 3 converts the real-time dynamic signal into the characteristic time map. The comparison module 4 compares the visual atlas with the characteristic time atlas of the grinding module 1 for similarity, that is, when looking for the same characteristic time atlas with the visual atlas as the target, the characteristic time atlas is captured and combined Perform comparison, and at the same time, the extracted characteristic time map is decomposed into dynamic signal similarity correlation, amplitude correlation, frequency correlation, phase correlation, short-term change correlation, glitch correlation, etc., which are respectively related to the dynamic signal similarity of the visual map , Amplitude correlation, frequency correlation, phase correlation, short-term change correlation, and surge correlation for comparison. As shown in Figure 5, in the process of real-time monitoring of dynamic signals, the calibrated visual map is automatically detected, and simple or complex actions can be automatically tracked and identified.

Here, dynamic signal similarity correlation, amplitude correlation, frequency correlation, phase correlation, short-term change correlation, and glitch correlation have their own weights. The comparison module 4 will calculate the individual scores and add them together to get The total score, and the total score standard (referred to as the limit value here) can be adjusted by the customer according to the product's longitude requirements. For example, 90 points are the standard. When the score is 92, the result is qualified, when the score is 85, it is unqualified. When the judgment result does not exceed the limit value, the polishing module 1 continues to operate, and when the judgment result exceeds the limit value, the processing module 3 controls the polishing module 1 to stop operating. If the first comparison score is not good, it is judged that the machine has deteriorated.

Further, the comparison module includes an actuation database 41 that stores a plurality of characteristic time maps of the polishing module 1 at different time points, and a plurality of judgment results at different time points obtained from the characteristic time maps of different time points. And the plural state trend graphs made with the characteristic time graphs at different time points. Wherein, the state trend graph includes a dynamic signal similarity graph, an amplitude graph, a frequency graph, a phase graph, a short-term change graph, a burst graph, etc.

The learning module 5 is signal-connected to the comparison module 4, and the learning module 5 executes a supervised learning method to use the six eigenvalue maps of the visualized map with the dynamic signal similarity map, the amplitude map, the The frequency map, the phase map, the short-term change map, and the burst map are compared to generate a difference value. When the difference value result is lower than the limit value, the processing module sends out a predictive signal.

If executed N times, different types of state trend graphs will mark N points and make individual trend graphs, as shown in Figure 6. If the equipment is stable and there is no deterioration, it will display a linear steady state, if the equipment is deteriorated, It shows obvious tortuous amplitude, and Figure 7 shows the long-term deterioration trend of the equipment, allowing users to know the aging phenomenon of the equipment from the trend graph under different conditions, so as to achieve the predictive maintenance effect of controlling the deterioration trend state of the equipment for a long time.

The warning module 6 is connected to the processing module 3 with a signal. When the judgment result exceeds the limit value, the processing module 3 controls the grinding module 1 to stop operating, and sends a warning signal through the warning module 6. Further, the warning signal can be sent to the electronic device of the preset user through the warning module 6, such as mobile device, personal computer, industrial computer, etc.

The display module 7 is connected with the processing module 3 signal to display the judgment result, and the judgment result records the pass or not, alarm record, failure reason, measurement time, dynamic similarity, amplitude, frequency, short time , Surge changes and other information, as shown in Figure 8. The specification in the figure is the visual map, the red is the real-time measurement data, and the dynamic similarity is the customer-defined limit value.

Refer to Figure 9 for the second stage of the new prediction machine damage, including the following steps.

S4: Store the characteristic time map information of the grinding module 1 at different time points, and the judgment results at different time points based on the characteristic time map information at different time points, and then use the experience threshold set by the customer as a supervisory method Learn.

S5: The learning module 5 executes a supervised learning method, using the feature values of the visualized map to compare with the dynamic signal similarity map, the amplitude map, the frequency map, the phase map, the short-term change map, and the burst map For comparison, when the result of the difference value is lower than the limit value, the processing module sends out a predictive signal to achieve the effect of detecting equipment failures in advance, while protecting the health of the equipment structure and product quality. Among them, the difference value can also be adjusted by the customer. Refer to Figure 10, which shows the application of this intelligent monitoring system to grinding equipment.

To sum up, the present invention is applied to the intelligent monitoring system of the polishing process, through the polishing module 1, the sensing module 2, the processing module 3, the comparison module 4, the learning module 5, The warning module 6 and the display module 7 apply artificial inspection thresholds to not only start the equipment monitoring task at the same time as the installation, but also to automatically classify and clean the data on the production line, without the need for special personnel to follow. The extra time is needed to label the data and classify the time sequence. It is only necessary to automatically classify the measurement data into groups and then directly learn from the learning module 5. At the same time, the monitoring model can be optimized online to achieve the double detection and prediction of the equipment status. Effectiveness, so it can indeed achieve the purpose of new cost.

However, the above are only the preferred embodiments of the present model, and should not be used to limit the scope of implementation of the present model, that is, simple equivalent changes and modifications made in accordance with the scope of the patent application for the present model and the description of the model, All are still within the scope of this new patent.

1: Grinding module 2: Sensing module 21: Sensor 3: Processing module 4: Comparison module 41: Action Database 5: Learning module 6: Warning module 7: Display module S1~S5: steps

Figure 1 is a schematic diagram illustrating a preferred embodiment of the intelligent monitoring system of the present invention applied to the grinding process; 2 is a schematic diagram illustrating the steps of monitoring the real-time status of the production line machine of the grinding process in the preferred embodiment; FIG. 3 is a schematic diagram illustrating a repetitive target interval in a characteristic time map captured in the preferred embodiment; Figure 4 is a schematic diagram illustrating the decomposition of a visual map into six eigenvalue maps; Figure 5 is a schematic diagram illustrating the visual map of automatic detection and calibration during real-time monitoring of dynamic signals; Figure 6 is a schematic diagram illustrating the state trend diagram of the device; Figure 7 is a schematic diagram illustrating the long-term deterioration trend of the equipment; Figure 8 is a schematic diagram illustrating the display state of the intelligent monitoring system; Figure 9 is a schematic diagram illustrating the steps of predicting machine damage in the preferred embodiment; and Figure 10 is a schematic diagram illustrating the application of the intelligent monitoring system to grinding equipment.

1: Grinding module

2: Sensing module

21: Sensor

3: Processing module

4: Comparison module

41: Action Database

5: Learning module

6: Warning module

7: Display module

Claims (10)

An intelligent monitoring system applied to the grinding process, including: A grinding module; A sensing module connected to the signal of the grinding module for monitoring the dynamic signal generated by the action of the grinding module; A processing module is connected to the sensing module signal, converts the dynamic signal into a characteristic time map, and captures a repetitive target interval in the characteristic time map, and then stores the target interval as a visual map ;and A comparison module is connected to the processing module signal, and the comparison module compares the visual map with the real-time characteristic time map of the grinding module to obtain a judgment result. When the judgment result is not When a limit value is exceeded, the polishing module continues to operate, and when the judgment result exceeds the limit value, the processing module controls the polishing module to stop operating. According to the intelligent monitoring system applied to the grinding process according to the first item of the scope of patent application, the target interval is the result of artificial judgment based on the display content of the characteristic time map. According to the intelligent monitoring system applied to the grinding process according to the second item of the scope of patent application, the visual map is decomposed into dynamic signal similarity correlation, amplitude correlation, frequency correlation, phase correlation, short-term change correlation, and sudden change by different mathematical formulas. Eigenvalue map of wave correlation. According to the intelligent monitoring system applied to the polishing process according to item 3 of the scope of patent application, the comparison module includes an actuation database storing a plurality of characteristic time maps of the polishing module at different time points, and plural basis Judgment results at different time points obtained from the characteristic time atlas at different time points. According to the intelligent monitoring system applied to the grinding process according to item 4 of the scope of patent application, the action database further stores a plurality of state trend graphs made with characteristic time maps at different time points, and the state trend graph includes A dynamic signal similarity diagram, an amplitude diagram, a frequency diagram, a phase diagram, a short-term change diagram, a burst diagram, etc. According to item 5 of the scope of patent application, the intelligent monitoring system applied to the polishing process further includes a learning module connected to the signal of the comparison module. The learning module executes a supervised learning method and uses the visual map The eigenvalue map is compared with the dynamic signal similarity map, the amplitude map, the frequency map, the phase map, the short-term change map, and the burst map to generate a difference value. When the difference value result is lower than For this limit value, the processing module sends out a predictive signal. According to item 1 of the scope of patent application, the intelligent monitoring system applied to the polishing process, wherein the sensing module includes a sensor arranged on the polishing module, and the sensor is selected from accelerometers, One or a combination of multi-axial accelerometers, gyroscopes, strain gauges, pressure sensors, thermometers, voltmeters, and current meters. The intelligent monitoring system applied to the grinding process according to item 1 of the scope of patent application further includes a warning module connected to the processing module signal. When the judgment result exceeds the limit value, the processing module transmits the warning The module sends out a warning signal. The intelligent monitoring system applied to the polishing process according to the first item of the scope of patent application, wherein the sensing module performs feature capture and monitoring of a specific operating point of the polishing module, and the specific operating point is selected from the grinding wheel , Grinding pad, rotating shaft, Z moving shaft mechanical quality. The intelligent monitoring system applied to the grinding process according to item 1 of the scope of patent application, further includes a display module connected with the processing module signal to display the judgment result, and the judgment result records the pass or not, alarm Record, failure reason, measurement time, dynamic similarity, amplitude, frequency, surge change and other information.

Images