Prediction device and prediction method for predicting the state of a workpiece

Abstract

This technology makes it easier for operators to empirically learn the correlation between various measurement values ​​during grinding and the machining state of the workpiece. [Solution] A prediction device for predicting the state of a workpiece during grinding comprises: a sensor for measuring spark features, which are characteristic quantities of sparks generated during grinding of a workpiece; a storage unit for storing first data in which the correlation between spark features and the state of the workpiece is recorded in advance; a prediction unit for predicting the state of the workpiece based on the spark features obtained from the sensor and the first data; and an output unit for outputting the predicted state of the workpiece along with the spark features during grinding of the workpiece.

Description

Technical Field

[0001] The present disclosure relates to a prediction apparatus and a prediction method for predicting the state of a workpiece.

Background Art

[0002] Patent Document 1 discloses a technique in which a measuring element is brought into contact with a workpiece during grinding of the workpiece, and the dimension and surface roughness of the workpiece are measured in-process. By performing measurement in-process, time and labor can be reduced compared to performing measurement after processing.

Prior Art Literature

Patent Literature

[0003]

Patent Document 1

Summary of the Invention

Problem to be Solved by the Invention

[0004] When directly measuring a workpiece in-process, the presence of coolant and chips remaining on the surface of the workpiece may affect measurement accuracy. Accordingly, the inventor of the present application studied predicting the processing state of a workpiece from the vibration, processing load, temperature change, etc. of the workpiece using machine learning. However, the prediction accuracy by machine learning greatly depends on the quality and quantity of training data, and the prediction algorithm becomes a black box. As a result, there is a problem that an operator who is unfamiliar with machine tools has difficulty empirically learning the correlation between various measured values during processing and the processing state of the workpiece, and has difficulty mastering the adjustment work of grinding conditions such as the rotation speed of the grinding wheel, the depth of cut, and the feed rate.

Means for Solving the Problem

[0005] The present disclosure can be implemented in the following modes.

[0006] (1) According to a first embodiment of the present disclosure, a prediction device is provided for predicting the state of a workpiece during grinding. The prediction device includes a sensor for measuring a spark feature quantity, which is a characteristic quantity of sparks generated during grinding of a workpiece; a storage unit for storing first data in which the correlation between the spark feature quantity and the state of the workpiece is recorded in advance; a prediction unit for predicting the state of the workpiece based on the spark feature quantity obtained from the sensor and the first data; and an output unit for outputting the predicted state of the workpiece together with the spark feature quantity during grinding of the workpiece. In this configuration, the state of the workpiece, predicted based on pre-recorded first data showing the correlation between spark characteristics and the state of the workpiece, is output along with the spark characteristics. This makes it easier for operators unfamiliar with machine tools to learn the correlation between the workpiece processing state and the spark characteristics. (2) In the above embodiment, the state of the workpiece may include at least one of the surface roughness of the workpiece and the roundness of the workpiece. In such an embodiment, the surface roughness and roundness of the workpiece can be predicted as the state of the workpiece. (3) In the above configuration, the first data may include a correlation between the surface roughness of the workpiece and at least one of the number of sparks and the size of the sparks. In this configuration, the surface roughness of the workpiece can be predicted based on the number of sparks and the size of the sparks. (4) In the above configuration, the first data may include a correlation between the roundness of the workpiece and the spark generation period of the rotating workpiece. In this configuration, the roundness of the workpiece can be predicted based on the periodicity of spark generation. (5) In the above configuration, the prediction unit may use the spark feature quantity to predict the state of the tool used in the grinding process, and the output unit may output information representing the state of the tool. In this configuration, the state of the tool can be output based on the spark feature quantity. (6) In the above configuration, the prediction unit may predict the maintenance timing of the tool used for grinding based on the predicted state of the workpiece, and the output unit may output information regarding the predicted maintenance timing. In this configuration, the maintenance timing of the tool can be output based on the predicted state of the workpiece. This disclosure is not limited to the form of a prediction device as described above, but can also be realized in the form of, for example, a prediction method or computer program for predicting the state of a workpiece during grinding, or a tangible recording medium that does not have a temporary recording medium on which the computer program is stored. [Brief explanation of the drawing]

[0007] [Figure 1] This is an explanatory diagram showing the configuration of a workpiece condition prediction system. [Figure 2] This is a conceptual diagram illustrating the interrelationship between spark characteristics and the state of the workpiece. [Figure 3] This is a flowchart for predicting the state of a workpiece. [Figure 4] This diagram illustrates the concept of surface roughness of a workpiece. [Figure 5] This figure shows an example of outputting the surface roughness of a workpiece along with spark features. [Figure 6] This diagram illustrates the concept of roundness in a workpiece. [Figure 7] This figure shows an example of outputting the roundness of a workpiece along with spark features. [Figure 8] This graph shows the number of sparks as a spark feature. [Figure 9] This graph shows the length of a spark as a spark feature. [Figure 10] This is a graph showing the trajectory of a spark. [Figure 11] This is a polar coordinate graph showing the distribution of spark generation. [Figure 12] This is a flowchart of the workpiece state prediction process in the second embodiment. [Modes for carrying out the invention]

[0008] A. First Embodiment: Figure 1 is an explanatory diagram showing the configuration of the workpiece state prediction system 100. The workpiece state prediction system 100 comprises a prediction device 10 and a sensor 90. The workpiece state prediction system 100 is used in machining equipment such as a cylindrical grinding machine to predict the state of a workpiece W during machining. In this embodiment, the workpiece W is a cylindrical steel material. In this embodiment, the tool 200 is a disc-shaped grinding wheel used to grind the workpiece W. The workpiece W is also called a workpiece.

[0009] The prediction device 10 predicts the state of the workpiece W being ground by the tool 200 based on the sparks SP generated during the machining of the workpiece W by the tool 200. The sparks SP are generated by the contact between the workpiece W and the tool 200.

[0010] Sensor 90 is used to measure spark features, which are characteristic quantities of sparks SP generated during the grinding process of a workpiece W. In this embodiment, sensor 90 is an image sensor and is mounted on a camera. Sensor 90 generates multiple spark images by continuously capturing images of sparks SP generated during grinding in a time series. In this embodiment, sensor 90 continuously captures images of sparks SP at very short time intervals. A very short time interval means that, when focusing on one streamline of a spark SP, the process from the generation to the disappearance of that streamline can be captured by multiple images. More specifically, in this embodiment, the spark images are captured at an imaging speed of 10,000 frames / second. Sensor 90 is configured to communicate with the prediction device 10 by wired communication or wireless communication.

[0011] The prediction apparatus 10 is constituted by a computer including a CPU 11, a storage unit 12 including a ROM and a RAM, an input / output interface 13, and an internal bus 14. The CPU 11, the storage unit 12, and the input / output interface 13 are communicably connected bidirectionally via the internal bus 14. The CPU 11 implements functions as a prediction unit 21 and an output unit 22 by executing a program stored in the storage unit 12. A display device 300 is connected to the prediction apparatus 10.

[0012] The prediction unit 21 obtains spark feature values from a spark image captured by a sensor 90. The prediction unit 21 predicts the state of a workpiece W based on the spark feature values and first data D1 stored in the storage unit 12. In the first data D1, a correlation between the spark feature values and the state of the workpiece W is recorded in advance. In the present embodiment, the state of the workpiece W refers to at least one of the surface roughness and the roundness of the workpiece W. The spark feature value refers to at least one of the number of sparks SP, the size of the sparks SP, and the generation cycle of the sparks SP on the rotating workpiece W. In the present embodiment, a correlation between the surface roughness of the workpiece W and at least one of the number of sparks SP and the size of the sparks SP is recorded in the first data D1. Further, in the present embodiment, a correlation between the roundness of the workpiece W and the generation cycle of the sparks SP during one rotation of the workpiece W is recorded in the first data D1.

[0013] The output unit 22 outputs the state of the workpiece W predicted by the prediction unit 21 together with the spark feature values during the grinding of the workpiece W. In the present embodiment, the output unit 22 outputs the predicted state of the workpiece W and the spark feature values to the display device 300. The output unit 22 may output the state of the workpiece W and the spark feature values to another device connected via a printing device or a communication interface instead of the display device 300.

[0014] Fig. 2 is a conceptual diagram showing the correlation between spark feature quantities and the state of a workpiece W. The prediction device 10 according to the present embodiment predicts the surface roughness and roundness of the workpiece W as the state of the workpiece W. The surface roughness and roundness of the workpiece W change depending on the state of abrasive grains contained in a tool 200. That is, the state of the workpiece W changes depending on the degree of wear and falling-off of the abrasive grains. In addition, the amount, volume, shape, and temperature of chips generated during machining also change depending on the state of the abrasive grains. Since sparks SP generated during machining of the workpiece W are generated by oxidation of chips, spark feature quantities such as the number and size of sparks SP, generation cycle, and distribution of sparks SP in time or space change depending on the state of abrasive grains contained in a grinding wheel. Summarizing the above, it can be said that the surface roughness and roundness of the workpiece W correlate with the spark feature quantities during machining. For example, if abrasive grains are locally worn or fallen off in the tool 200, a distribution occurs in the grinding state on the surface of the workpiece W, which causes periodicity in chip generation and also causes periodicity in the generation of sparks SP. That is, if there is periodicity in the generation of sparks SP, the workpiece W has a shape deviating from a perfect circle.

[0015] In the present embodiment, such a correlation between the spark feature quantities and the state of the workpiece W is obtained in advance through experiments and recorded in first data D1 of a storage unit 12. The spark feature quantities included in the first data D1 are not limited to the above-mentioned feature quantities, and for example, also include the feature quantities shown in (1) to (6) below. All spark feature quantities are feature quantities obtained by analysis from spark images captured by a sensor 90. For example, a spark feature quantity can be obtained by extracting one group of regions having a pixel value equal to or higher than a predetermined luminance in a spark image as one spark SP, and analyzing the number, position, size, length, and movement distance of the extracted sparks SP for each spark image acquired in time series.

[0016] (1) A feature quantity based on the number of sparks SP generated per unit time. Examples include the maximum value, average value, variance, standard deviation, kurtosis, skewness, maximum value of first-order differentiation, and average value of first-order differentiation of the number of sparks SP. (2) Features based on the coordinates of sparks SP in the spark image. For example, the maximum value, mean, variance, standard deviation, kurtosis, and skewness of the size or length of the sparks SP. (3) Periodic characteristics of spark generation. For example, the generation period of spark SP, generation frequency, analysis results of the waveform using Fast Fourier Transform (FFT) representing the change in the number of spark SP over time, autocorrelation, etc. (4) Features of the polar coordinate plot of the spark SP. For example, aspect ratio, circularity, etc. (5) Feature quantities related to the streamlines from the generation to the extinction of the spark SP. For example, the length of the streamline, the velocity of the spark SP, the curvature of the streamline, the lifetime of the spark SP, the number of times the spark SP splits, etc. (6) Principal component analysis results for the features listed in (1) to (5).

[0017] The first data set, D1, pre-records the correlation between these spark features and the measured results of surface roughness and roundness of the workpiece W, either as measured values ​​or in the form of functions or maps. Surface roughness can be measured, for example, using a contact-type or non-contact-type surface roughness measuring instrument. Roundness can be measured by measuring the maximum and minimum diameters of the workpiece W using a roundness measuring instrument or micrometer, and dividing the difference between the maximum and minimum values ​​by 2. The closer the roundness value is to 0, the higher the roundness; the larger the value, the lower the roundness. Functions and maps can be derived by analyzing measured values.

[0018] Figure 3 is a flowchart of the workpiece state prediction process performed in the prediction device 10. This workpiece state prediction process is repeatedly performed while the workpiece W is being machined.

[0019] In step S10, the prediction unit 21 measures spark features based on the spark image acquired by the sensor 90.

[0020] In step S20, the prediction unit 21 predicts the surface roughness of the workpiece W based on the spark feature quantities measured in step S10 and the first data D1 stored in the storage unit 12. Specifically, it predicts the state of the workpiece W by deriving the surface roughness of the workpiece W corresponding to the spark feature quantities measured in step S10 from the correlation between the spark feature quantities recorded in the first data D1 and the state of the workpiece W.

[0021] In step S30, the prediction unit 21 predicts the roundness of the workpiece W based on the spark feature measured in step S10 and the first data D1 stored in the storage unit 12. Specifically, the roundness of the workpiece W is predicted by deriving the roundness of the workpiece W corresponding to the spark feature measured in step S10 from the correlation between the spark feature recorded in the first data D1 and the roundness of the workpiece W. The execution order of steps S20 and S30 may be reversed or simultaneous.

[0022] In step S40, the output unit 22 outputs the surface roughness and roundness of the workpiece W, which were predicted in steps S20 and S30, along with the spark characteristic quantity measured in step S10, to the display device 300 as the state of the workpiece W during the grinding process of the workpiece W.

[0023] As the workpiece state prediction process described above is repeatedly executed during the machining of the workpiece W, the display device 300 displays in real time the spark feature quantities measured in-process and the state of the workpiece W predicted based on those spark feature quantities.

[0024] Figure 4 illustrates the concept of surface roughness of a workpiece W. As shown in Figure 4, the surface roughness of a workpiece W represents the magnitude of irregularities along the axial direction on the surface of the workpiece W. For example, the surface roughness of a workpiece W can be expressed as arithmetic mean roughness (Ra) or maximum height roughness (Rz).

[0025] Figure 5 shows an example of outputting the surface roughness of the workpiece W, along with spark characteristics, to the display device 300. In this embodiment, the output unit 22 outputs to the display device 300 by plotting the surface roughness against the spark characteristics on a graph. In the graph shown in Figure 5, the horizontal axis represents the surface roughness of the workpiece W, and the vertical axis represents the spark characteristics (for example, the size of the spark). In this embodiment, the workpiece state prediction process described above is repeatedly executed to predict the surface roughness of the workpiece W against the spark characteristics measured in step S10, and the prediction result is plotted on the graph. If the surface roughness is plotted a predetermined number or more times, the output unit 22 may output an approximate straight line that approximates each plot, as shown in Figure 5.

[0026] Figure 6 illustrates the concept of roundness of a workpiece W. As shown in Figure 6, the roundness of a workpiece W refers to the degree of deviation of the irregular shape on the outer circumference of the workpiece W from a geometrically correct circle. Roundness can be expressed as the difference in radii between two concentric geometric circles that enclose the irregular shape in the circumferential direction of the workpiece W.

[0027] Figure 7 shows an example of outputting the roundness of the workpiece W, along with spark features, to the display device 300. In this embodiment, the output unit 22 outputs to the display device 300 by plotting the roundness relative to the spark features on a graph. In the graph shown in Figure 7, the horizontal axis represents the roundness of the workpiece W, and the vertical axis represents the spark features (for example, the spark generation period). In this embodiment, the workpiece state prediction process described above is repeatedly executed to predict the roundness of the workpiece W relative to the spark features measured in step S10, and the prediction result is plotted on the graph. If the roundness is plotted a predetermined number of times or more, the output unit 22 may output an approximate straight line that approximates each plot, as shown in Figure 7.

[0028] In step S40 described above, the output unit 22 may output the graphs shown in Figure 5 and Figure 7 simultaneously, or it may output only one of them. Alternatively, the operator may be able to switch between outputting the graph shown in Figure 5 and the graph shown in Figure 7.

[0029] The output unit 22 may output to the display device 300 during grinding any information related to sparks, such as the various spark characteristics or spark SPs shown in Figures 8 to 11, not limited to the graphs shown in Figures 5 and 7.

[0030] Figure 8 is a graph showing the number of sparks SP as a spark feature. The horizontal axis represents elapsed time, and the vertical axis represents the number of sparks SP counted in the spark image. The number of sparks SP tends to increase during the rough machining process in grinding and decrease during the finishing process performed after the rough machining process.

[0031] Figure 9 is a graph showing the length of the spark SP as a spark feature. The horizontal axis represents elapsed time, and the vertical axis represents the average value of the spark SP length. The curve shows the approximate curve of the measured values. Figure 9 shows that once the number of processing cycles exceeds a certain level, the length of the spark SP tends to increase significantly as the number of processing cycles increases.

[0032] Figure 10 is a graph showing the trajectory of the spark SP. The horizontal axis represents the X coordinate in the spark image, and the horizontal axis represents the Y coordinate. The output unit 22 plots the position of the spark SP generated during one rotation of the workpiece W, thereby outputting the trajectory shown in Figure 10 as a graph. The output unit 22 rewrites the graph in Figure 10 each time the workpiece W rotates. By referring to this graph, the operator can understand how the spark SP is flying.

[0033] Figure 11 is a polar coordinate graph showing the distribution of sparks SP. The output unit 22 counts the number of sparks SP generated for each rotation angle of the grinding wheel based on the spark image, and outputs the graph shown in Figure 11 by plotting the number of sparks SP generated at each angle of the grinding wheel. By referring to this graph, the operator can estimate the degree of roundness of the workpiece W or the roundness of the tool 200.

[0034] As described above, the prediction device 10 of this embodiment outputs the predicted state of the workpiece W based on the first data D1, which has a pre-recorded correlation between spark features and the state of the workpiece W, along with the spark features measured from the spark image. The output state of the workpiece W is predicted based on a simple correlation from the measured spark features. Therefore, it becomes easier for operators unfamiliar with machine tools to learn the correlation between the workpiece processing state and the spark features. As a result, operators can more easily predict the state of the workpiece W according to the spark generation situation SP, and it becomes easier for them to learn how to adjust grinding conditions such as the rotation speed of the grinding wheel, the depth of cut, and the feed rate. Furthermore, the correlation between spark features and the state of the workpiece W is determined in advance based on actual measurements. Therefore, the prediction reliability is higher than that of prediction results based on machine learning models, which depend heavily on the quality and quantity of training data, and outputting such prediction results to the display device 300 allows for appropriate training of operators unfamiliar with the work.

[0035] Furthermore, in this embodiment, the state of the workpiece W includes at least one of the surface roughness of the workpiece W and the roundness of the workpiece W. Therefore, the surface roughness and roundness of the workpiece W can be predicted based on the spark characteristics.

[0036] Furthermore, in this embodiment, the first data D1 records the correlation between the surface roughness of the workpiece W and at least one of the number of sparks SP and the size of the sparks SP. Therefore, the surface roughness of the workpiece W can be predicted based on the number of sparks SP and the size of the sparks SP measured using the sensor 90.

[0037] Furthermore, in this embodiment, the first data D1 records the correlation between the roundness of the workpiece W and the generation period of sparks SP during the rotation of the workpiece W. Therefore, the roundness of the workpiece W can be predicted based on the generation period of sparks SP measured using the sensor 90.

[0038] B. Second Embodiment: Figure 12 is a flowchart of the workpiece state prediction process in the second embodiment. The workpiece state prediction process in the second embodiment adds steps S50 and S60 to the workpiece state prediction process in the first embodiment shown in Figure 3. Steps S10 to S40 are the same as in the first embodiment, so their explanation is omitted. The configuration of the workpiece state prediction system 100 in the second embodiment is the same as in the first embodiment.

[0039] In the second embodiment, step S50 outputs information representing the state of the tool 200. Specifically, the prediction unit 21 predicts the state of the tool 200 used for grinding based on the spark feature measured in step S10. The output unit 22 then outputs the information representing the predicted state of the tool 200 to the display device 300. For example, the prediction unit 21 predicts that abrasive grain wear or detachment has occurred when the size of the spark SP, which is a spark feature, exceeds a predetermined threshold, and the output unit 22 outputs the prediction result. Alternatively, for example, data recording the correlation between the spark feature and the state of the tool 200 may be stored in the storage unit 12 as second data, and the prediction unit 21 may use this second data to predict the state of the tool 200 based on the spark feature.

[0040] In the second embodiment, step S60 outputs information regarding the maintenance timing. Specifically, the prediction unit 21 predicts the maintenance timing of the tool 200 used for grinding based on the state of the workpiece W predicted in steps S20 and S30. The output unit 22 then outputs the predicted maintenance timing information to the display device 300. For example, the prediction unit 21 predicts that the current time is maintenance time when the surface roughness of the workpiece W exceeds a predetermined threshold or when the roundness of the workpiece W exceeds a predetermined threshold, and the output unit 22 outputs to the display device 300 that the current time is maintenance time. Alternatively, for example, the prediction unit 21 may predict the timing when the values ​​of the surface roughness of the workpiece W or the roundness of the workpiece W reach a predetermined threshold based on the time change of these values, and the output unit 22 may output the predicted maintenance timing.

[0041] According to the second embodiment described above, the prediction device 10 outputs the state of the tool 200 based on the spark characteristics. The prediction device 10 also outputs information regarding the maintenance timing of the tool 200 based on the state of the workpiece W. Therefore, the operator can perform truing or replacement of the tool 200 at the appropriate time based on the output information. As a result, it is possible to suppress the occurrence of quality defects in the workpiece W and to extend the lifespan of the tool 200.

[0042] Note that both the process in step S50 and the process in step S60 shown in Figure 12 may be executed, or only one of them may be executed.

[0043] C. Other embodiments: (C1) In each of the above embodiments, the prediction unit 21 may predict the state of the workpiece W according to the spark feature quantity and determine an abnormality of the workpiece W based on the prediction result. If an abnormality is determined, the output unit 22 may output information to the display device 300 indicating that an abnormality has occurred. For example, the prediction unit 21 may determine that an abnormality has occurred if the surface roughness or roundness of the workpiece W deviates from a predetermined reference range.

[0044] (C2) In each of the above embodiments, the output unit 22 outputs the spark feature quantity and the state of the workpiece W, and may also output indicators that can be used as a reference for the operator to adjust the grinding conditions. For example, if the prediction result of the state of the workpiece W is good, the output unit 22 outputs a blue indicator (e.g., a blue rectangular area) to the display device 300, and if it is not good, it outputs a yellow indicator (e.g., a yellow rectangular area). The operator can easily adjust the grinding conditions by changing the grinding conditions while referring to these colors. In addition, the output unit 22 may output a blue indicator when there is little change in the spark feature quantity, such as the size and length of the spark SP, and a yellow indicator when there is a large change in the spark feature quantity. These indicators are not limited to the display device 300, but may also be displayed using indicator lights connected to the prediction device 10 or the machine tool, for example.

[0045] (C3) This disclosure is applicable not only to cylindrical grinding machines but also to various other grinding machines such as surface grinding machines and internal grinding machines.

[0046] This disclosure is not limited to the embodiments described above, and can be implemented in various configurations without departing from its spirit. For example, the technical features in the embodiments corresponding to the technical features in each form described in the summary of the invention can be replaced or combined as appropriate in order to solve some or all of the above-described problems, or to achieve some or all of the above-described effects. Furthermore, if a technical feature is not described as essential in this specification, it can be deleted as appropriate. [Explanation of Symbols]

[0047] 10…Prediction device, 11…CPU, 12…Storage unit, 13…Input / output interface, 14…Internal bus, 21…Prediction unit, 22…Output unit, 90…Sensor, 100…Workpiece state prediction system, 200…Tool, 300…Display device, D1…First data, SP…Spark, W…Workpiece

Claims

1. A predictive device for predicting the state of a workpiece during grinding, A sensor for measuring spark characteristics, which are characteristic quantities of sparks generated during the grinding process of a workpiece, A storage unit that stores first data in which the correlation between the spark characteristic quantity and the state of the workpiece is recorded in advance, A prediction unit predicts the state of the workpiece based on the spark characteristic quantity obtained from the sensor and the first data, During the grinding process of the workpiece, an output unit outputs the predicted state of the workpiece along with the spark characteristic quantity. A prediction device equipped with the following features.

2. A prediction device according to claim 1, A predictive device that includes at least one of the surface roughness and roundness of the workpiece as the condition of the workpiece.

3. A prediction device according to claim 1, A prediction device in which the first data records the correlation between the surface roughness of the workpiece and at least one of the number of sparks and the size of the sparks.

4. A prediction device according to claim 1, The prediction device records the correlation between the roundness of the workpiece and the spark generation period of the rotating workpiece in the first data.

5. A prediction device according to any one of claims 1 to 4, The prediction unit uses the spark feature quantity to predict the state of the tool used in the grinding process. The output unit is a prediction device that outputs information representing the state of the tool.

6. A prediction device according to any one of claims 1 to 4, Based on the predicted state of the workpiece, the prediction unit predicts the maintenance timing for the tool used in the grinding process. The output unit is a prediction device that outputs information regarding the predicted maintenance timing.

7. A prediction method for predicting the state of a workpiece during grinding, A step of measuring the characteristic quantity of a spark generated during grinding of a workpiece using a sensor for acquiring the characteristic quantity of the spark, A first data set in which the correlation between the characteristic quantity of the spark and the state of the workpiece is recorded in advance, and a step of predicting the state of the workpiece based on the characteristic quantity of the spark, The process of outputting the predicted state of the workpiece along with the characteristic quantities of the spark during the grinding process of the workpiece, A prediction method that includes the following features.

Images