JP7700466B2 - Tool life prediction system - Google Patents

Description

The present invention relates to a tool life prediction system.

Predicting the lifespan of tools used in machining is important from the perspective of tool costs. Conventionally, tool lifespans were determined in advance based on the number of machining operations required to determine the end of a tool's lifespan, taking into account individual tool variations and safety factors. However, with this method of determining tool lifespan, even when the tool had not actually reached its end of life, it was deemed to have reached its end of life and the tool was replaced.

Therefore, Patent Document 1 proposes predicting tool life using machining information when machining is being performed with a tool, for example, the drive current of the motor of the spindle device. In this prediction method, a calculation model determined based on the machining information is used to predict the tool life. In addition, the calculation model used to predict the tool life is selected from a plurality of calculation models depending on, for example, the material of the workpiece, the machining conditions, etc.

JP 2019-82836 A

The state in which a tool reaches the end of its life varies depending on factors such as the type of tool, the material of the workpiece, and the machining conditions. Therefore, it is effective to set a calculation model to be used to predict the tool's life for each life case depending on factors such as the type of tool, the material of the workpiece, and the machining conditions.

However, it is not easy to select one calculation model from among multiple calculation models. For example, if a calculation model exists in which all factors such as the tool, workpiece, and processing conditions are the same, that calculation model can be used. However, if some of the factors are different, for example, if the type of tool is different, it is not easy to determine which calculation model should be selected. Furthermore, even if all factors such as the tool, workpiece, and processing conditions are the same, the selected calculation model may not be appropriate due to the influence of individual differences in the tool, workpiece, processing machine body, etc.

When applying machine learning, consider the case where one trained model is selected from multiple trained models. In this case, if the explanatory variables of the multiple trained models are all homogeneous, it is also possible to select one trained model to apply based on the explanatory variables in the estimation phase.

In general, the selection of features as explanatory variables in machine learning has a significant impact on the estimation accuracy of machine learning. Therefore, it may be effective to use different features as explanatory variables for each learning model. However, when the features as explanatory variables of multiple trained models differ for each model, it is not easy to select one trained model from among the multiple trained models.

The present invention aims to provide a tool life prediction system that can predict tool life with high accuracy by using multiple trained models without having to select one trained model from among multiple trained models.

The tool life prediction system includes a processing machine body that uses a tool to machine a workpiece, a detector that detects observable state data in the processing machine body while the workpiece is being machined, a life prediction model storage unit that stores, for each of a plurality of life cases, a life prediction model that is generated by performing machine learning using a training data set including the explanatory variables and the objective variable, with the state data detected by the detector as an explanatory variable and the remaining number of times the workpiece is to be machined until the life of one tool is reached as an objective variable, and the life prediction model that is generated by performing machine learning using a training data set including the explanatory variables and the objective variable, a scoring model storage unit that stores, for each of a plurality of life cases, a scoring model that is generated by performing machine learning using a training data set with the state data for each of the life cases as an explanatory variable, the scoring model that outputs a score value of the state data, when the cases for each of the tools are treated as separate life cases , a remaining number of times of machining prediction unit that outputs a remaining number of times of machining prediction value by each of the plurality of life prediction models based on the state data in the estimation phase, a score value output unit that outputs a score value by each of the plurality of scoring models based on the state data in the estimation phase, and an overall remaining number of times of machining prediction unit that calculates an overall remaining number of times of machining based on each of the plurality of remaining number of times of machining predicted values and each of the plurality of score values.

According to the above tool life prediction system, a life prediction model and a scoring model are stored in correspondence with each life case. The scoring model outputs a score value of the condition data as an explanatory variable. The score value output by the scoring model corresponds to the score value of the corresponding life prediction model, i.e., the degree of suitability that indicates how well the life prediction model suits the condition data. For example, when the score value is large, it is estimated that the degree of suitability of the corresponding life prediction model is high, and when the score value is small, it is estimated that the degree of suitability of the corresponding life prediction model is low.

Then, the overall remaining number of machining operations prediction unit calculates the overall remaining number of machining operations based on each of the multiple remaining number of machining operations prediction values and each of the multiple score values. In other words, the overall remaining number of machining operations prediction unit can use multiple life prediction models by taking the score values into consideration. In other words, the overall remaining number of machining operations prediction unit does not need to select only one life prediction model from among the multiple life prediction models. In this way, the overall remaining number of machining operations prediction unit predicts the overall remaining number of machining operations using multiple life prediction models and multiple score values, and therefore can perform tool life prediction with high accuracy.

FIG. 1 is a diagram illustrating a configuration of a tool life prediction system. FIG. 2 is a diagram illustrating an example of a processing machine. FIG. 1 is a schematic diagram of a tool life prediction process using machine learning in a tool life prediction system. FIG. 11 is an explanatory diagram of score values output by a scoring model. FIG. 2 is a diagram showing a functional block configuration of the tool life prediction system. FIG. 13 is a diagram for explaining the calculation of the remaining number of machining operations in the remaining number of machining operations calculation unit, and is also a diagram showing a training data set for life prediction. FIG. 13 is a diagram showing a correspondence table between tools, life prediction models, and scoring models. 4 is a diagram showing a plurality of life prediction models stored in a life prediction model storage unit and a plurality of scoring models stored in a scoring model storage unit; FIG.

(1. Types of machining machines and tools to which the tool life prediction system is applicable)
The tool life prediction system predicts the life of a tool in a processing machine that uses the tool to machine a workpiece. Here, the tool life includes not only a state in which the tool is completely unusable, but also a state in which the tool needs to be modified. For example, tool modification includes resharpening the tool, truing the grinding wheel, dressing, etc.

Processing machines include machining centers, lathes, milling machines, boring machines, gear processing machines, and other machines that perform cutting processing. In these cases, the tools are cutting tools. For example, tools in machining centers include drills, milling tools, boring tools, gear processing tools, turning tools, and the like. Tools in lathes include at least turning tools, and tools in multi-purpose lathes include turning tools as well as tools similar to those in machining centers, such as drills and milling tools. Gear processing machines include gear skiving machines, hob processing machines, shaper processing machines, and the like, and tools in the gear processing machines include gear skiving tools, hob tools, shaper tools, and other tools for gear processing.

Processing machines include grinding machines that perform grinding. Tools for grinding machines include grinding wheels. Processing machines include forging machines, such as presses and rolling machines. Tools for presses and rolling machines include forging punches, forging dies, etc.

(2. Overview of the configuration of tool life prediction system 1)
An overview of the configuration of the tool life prediction system 1 will be described with reference to Fig. 1. The tool life prediction system 1 includes a processing machine 10 and a computing device (20, 30). The processing machine 10 may be a single machine, or may be a plurality of machines as shown in Fig. 1. In this example, the tool life prediction system 1 includes a plurality of processing machines 10.

The machining device 10 includes at least a machine body 11 that uses a tool T to machine a workpiece W, and a detector 13 that detects observable status data of the machine body 11 while the workpiece W is being machined.

The calculation device (20, 30) predicts the life of the tool T by applying machine learning using the condition data detected by the detector 13. In FIG. 1, the calculation device (20, 30) is composed of a learning processing device 20 and a predictive calculation device 30, and although the learning processing device 20 and the predictive calculation device 30 are shown as independent configurations, they can also be one device. In addition, a part or all of the calculation device (20, 30) can be an embedded system in the processing machine 10.

In this example, the learning processing device 20 and the prediction calculation device 30 are independent. The learning processing device 20 executes the learning phase processing in machine learning, and the prediction calculation device 30 executes the estimation phase processing in machine learning. The learning processing device 20 also has a so-called server function and is connected to multiple processing machines 10 so as to be able to communicate with each other. On the other hand, the prediction calculation device 30 is provided in a one-to-one correspondence with each processing machine 10 and is connected to each processing machine 10 so as to be able to communicate with each other. In other words, the multiple prediction calculation devices 30 function as so-called edge computers and are able to realize high-speed calculation processing. In other words, in this example, the tool life prediction system 1 includes one learning processing device 20 having a server function and multiple prediction calculation devices 30, 30 corresponding to each of the multiple processing machines 10, 10.

(3. Detailed Configuration of Tool Life Prediction System 1)
The detailed configuration of the tool life prediction system 1 will be described in more detail with reference to Fig. 1. The tool life prediction system 1 includes a plurality of processing machines 10, 10, one learning processing device 20 functioning as a part of the calculation device, and a plurality of prediction calculation devices 30, 30 functioning as other parts of the calculation device.

As described above, various types of processing machines can be applied to each processing machine 10. The processing machine 10 includes a processing machine body 11 that uses a tool T to machine a workpiece W, a control device 12 that controls the processing machine body 11, a detector 13, and an interface 14.

The machine body 11 has a tool T, supports a workpiece W, and is configured to move the tool T and the workpiece W relative to one another. In other words, the machine body 11 is equipped with a structure and a drive device that drives the structure. The control device 12 includes a CNC device and a PLC device, and is configured with a processor, a storage device, etc. The control device 12 controls the drive device and the like in the machine body 11. The interface 14 is a device that enables communication between the machine body 11, the control device 12, and the detector 13 and the outside.

The detector 13 detects observable status data in the machine body 11 while the workpiece W is being machined. The detector 13 detects time-series status data relating to, for example, the machining load and the drive load of the drive device. The detector 13 is, for example, a current sensor that detects drive current data of a motor serving as a drive device, a vibration sensor that detects vibration data of components of the machine body 11, a microphone that detects sound data during machining, etc. In other words, the status data is, for example, drive current data, vibration data, sound data, etc.

The learning processing device 20 is configured with a processor 21, a storage device 22, an interface 23, etc. The learning processing device 20 executes the processing of the learning phase in machine learning. In this example, the learning processing device 20 also has a server function and is connected to multiple processing machines 10 so that it can communicate with them.

The learning processing device 20 generates a life prediction model for predicting the life of the tool T by performing machine learning based on the condition data detected by the detector 13. In particular, the learning processing device 20 generates a life prediction model for each of multiple life cases. In other words, the learning processing device 20 generates multiple life prediction models.

For example, in the learning processing device 20, a life prediction model is generated for each tool T. For tools of the same type as well as for different types of tools T, a life prediction model is generated for each life case, with the life cases for each tool T being treated as separate life cases. In other words, a life prediction model is generated for each tool T, regardless of whether it is the same type or different. In this way, the learning processing device 20 generates a life prediction model for each of multiple life cases, with each tool T treated as a respective life case.

Furthermore, the learning processing device 20 performs machine learning based on the state data for each lifespan case detected by the detector 13 to generate a scoring model that outputs a score value for the state data. In other words, the learning processing device 20 generates a scoring model for each of the multiple lifespan cases. Therefore, the same number of scoring models as the lifespan prediction models are generated, and each scoring model corresponds to each lifespan prediction model.

Here, the score value is a numerical value that indicates the degree of normality of the condition data input to the scoring model. The higher the degree of normality, the larger the score value, and the lower the degree of normality, the smaller the score value. Furthermore, the score value may indicate a value that indicates an abnormality in addition to a value that indicates normality for the condition data input to the scoring model. For example, the score value indicates a positive value that indicates normality and a negative value that indicates an abnormality.

And, as described above, each scoring model corresponds to a respective lifespan prediction model. Therefore, when a scoring model outputs a large score value, it means that the lifespan prediction model corresponding to that scoring model has a high degree of compatibility with the input state data. Conversely, when a scoring model outputs a small score value, it means that the lifespan prediction model corresponding to that scoring model has a low degree of compatibility with the input state data. In other words, the scoring model outputs a score value as the degree to which the input state data matches (goodness of compatibility) with the lifespan prediction model corresponding to that scoring model. In other words, the score value represents the degree of compatibility between a lifespan prediction model and the state data input to that lifespan prediction model.

Each prediction calculation device 30 is configured with a processor 31, a storage device 32, an interface 33, etc. The prediction calculation device 30 executes the processing of the estimation phase in machine learning. In addition, the prediction calculation device 30 is connected so as to be able to communicate with the learning processing device 20 as a server and the corresponding processing machine 10.

Each prediction calculation device 30 is disposed in close proximity to each processing machine 10 and functions as a so-called edge computer. Using a plurality of life prediction models and a plurality of scoring models generated by the learning processing device 20, the prediction calculation device 30 predicts the life of the tool T based on the status data detected by the detector 13 during processing of the workpiece W. In this example, the prediction calculation device 30 predicts the remaining number of times of processing, i.e., the remaining number of times that the workpiece W can be processed.

The tool life prediction system 1 further includes a common display device 40 and a plurality of individual display devices 50. However, the tool life prediction system 1 may be configured without the common display device 40, or without the individual display devices 50. The common display device 40 is arranged in correspondence with the learning processing device 20. Furthermore, the individual display devices 50 are arranged in correspondence with each of the processing machines 10.

(4. Examples of Processing Machine Body 11)
As an example of the processing machine body 11, a machining center will be described with reference to Fig. 2. As described above, the processing machine body 11 is an example of a processing machine capable of performing various processes including gear processing, and may be applied to other processing machines.

As shown in FIG. 2, the machine body 11 as a machining center is configured as a five-axis machining center having three linear axes and two rotational axes as drive axes that change the relative position and posture of the workpiece W and the tool T. In this example, the machine body 11 has three orthogonal axes (X-axis, Y-axis, Z-axis) as linear axes, and the B-axis and Cw-axis as rotational axes. The B-axis is a rotational axis around the Y-axis, and the Cw-axis is a rotational axis around the central axis of the workpiece W.

The processing machine body 11 is provided with a tool spindle 61 that supports a tool T (rotary tool) and can rotate on the Ct axis, and can move in the Y axis and Z axis directions. The processing machine body 11 further includes a workpiece spindle 62 that supports a workpiece W and can rotate on the Cw axis, can rotate on the B axis, and can move in the X axis direction. The processing machine body 11 has a motor as a drive device for driving each axis (X axis, Y axis, Z axis, B axis, Cw axis, Ct axis).

(5. Overview of Tool Life Prediction Processing)
An overview of the tool life prediction process of the tool T in the tool life prediction system 1 will be described with reference to Fig. 3 and Fig. 4. Here, the following description is given by way of an example in order to understand the overview of the tool life prediction process. Therefore, the tool life prediction process is not limited to the following description. The tool life prediction process applies machine learning, and is a process in which a learning phase is performed by the learning processing device 20 and an estimation phase is performed by the prediction calculation device 30.

As shown in FIG. 3, in the learning phase by the learning processing device 20, status data Da during machining by tool Ta, status data Db during machining by tool Tb, and status data Dc during machining by tool Tc are used. In other words, the cases using each of the tools Ta, Tb, and Tc are treated as separate life cases.

First, the learning processing device 20 generates a life prediction model Ma1 by performing machine learning. The life prediction model Ma1 is generated by performing machine learning using a life prediction training dataset including explanatory variables and objective variables, with the status data Da during machining by the tool Ta as explanatory variables and the remaining number of times the workpiece W is machined until the tool Ta reaches its life as objective variables. In other words, the life prediction model Ma1 is a trained model that outputs a remaining number of times machining, Na, which is the objective variable, when the status data Da, which is the explanatory variable, is input.

The learning processing device 20 generates a life prediction model Ma1, for example, by supervised learning. For example, linear regression, ridge regression, Lasso, elastic net, random forest, support vector machine, etc. can be applied to the life prediction model Ma1. The life prediction model Ma1 may be any model that can define the relationship between the state data Da, which is an explanatory variable, and the remaining number of machining operations, which is an objective variable.

Furthermore, the learning processing device 20 generates a scoring model Ma2 by performing machine learning using a training dataset for scoring in which the state data Da is an explanatory variable. The learning processing device 20 generates a scoring model Ma2 that outputs a score value Va for the state data Da by performing machine learning.

The learning processing device 20 generates a scoring model Ma2, for example, by unsupervised learning. For example, a one-class support vector machine (SVM) can be applied to the scoring model Ma2. As shown in FIG. 4, the one-class SVM performs machine learning using normal data, can set the boundary between normal and abnormal, and outputs a score value representing normal and a score value representing abnormality as different values. For example, the one-class SVM outputs a score value represented by a positive value in the case of normality, and outputs a score value represented by a negative value in the case of abnormality. Furthermore, the one-class SVM outputs a score value representing the degree of normality in the case of normality. Each range of the score value representing the degree of normality is illustrated by a dashed line in FIG. 4.

In addition to the one-class SVM, the scoring model Ma2 can also be a model capable of outputting the degree of normality, such as an autoencoder, Mahalanobis distance, or likelihood. The Mahalanobis distance is an index of distance obtained based on the variance of the data. The likelihood is an index of "plausibility" used in Bayesian estimation, etc.

Here, it is preferable that the lifespan prediction model Ma1 and the scoring model Ma2 corresponding to the lifespan prediction model Ma1 use the same type of feature as explanatory variables. By using the same type of feature as explanatory variables, it is possible to increase the correlation between the two models Ma1 and Ma2. However, the feature amounts serving as explanatory variables of the two models Ma1 and Ma2 do not all have to be the same type, and some of them may be different.

Various statistics can be applied as the feature quantity. For example, the feature quantity can be a statistical quantity such as the mean, variance, standard deviation, skewness, kurtosis, maximum value, minimum value, median, first quartile, third quartile, or difference between the maximum and minimum values. The feature quantity may also include the above statistical quantities for data obtained by differentiating state data, the above statistical quantities for data obtained by performing frequency analysis on state data, etc.

The learning processing device 20 also generates a life prediction model Mb1 and a scoring model Mb2 for the condition data Db related to the tool Tb. The learning processing device 20 also generates a life prediction model Mc1 and a scoring model Mc2 for the condition data Dc related to the tool Tc.

In this way, the learning processing device 20 generates life prediction models Ma1, Mb1, Mc1 and scoring models Ma2, Mb2, Mc2 for each of the tools Ta, Tb, Tc, i.e., for each life case.

Here, each of the life prediction models Ma1, Mb1, and Mc1 can use different feature quantities for each of the state data Da, Db, and Dc as explanatory variables. In other words, at least some of the feature quantities serving as explanatory variables for the life prediction model Ma1, the feature quantities serving as explanatory variables for the life prediction model Mb1, and the feature quantities serving as explanatory variables for the life prediction model Mc1 can be different types of feature quantities.

Of course, the explanatory variables of the life prediction models Ma1, Mb1, and Mc1 may all be the same type of feature. However, by allowing different features to be used as explanatory variables, each of the life prediction models Ma1, Mb1, and Mc1 can be a good model, i.e., a model that can perform highly accurate predictions.

Next, in the estimation phase by the prediction calculation device 30, status data Dx during machining by the new tool Tx is used. The prediction calculation device 30 uses the status data Dx regarding the tool Tx as input data, and outputs remaining machining cycle prediction values Na, Nb, Nc as life prediction results using multiple life prediction models Ma1, Mb1, Mc1. For example, the remaining machining cycle prediction values Na, Nb, Nc are 500 times, 480 times, and 510 times, respectively.

Furthermore, the prediction calculation device 30 uses the state data Dx regarding the tool Tx as input data, and outputs the score values Va, Vb, and Vc as the scoring results using multiple scoring models Ma2, Mb2, and Mc2. For example, the score values Va, Vb, and Vc are 5, 4, and 2, respectively.

Then, the prediction calculation device 30 calculates the total remaining number of processing times Nx by using each of the multiple score values Va, Vb, and Vc as a weight for each of the multiple remaining number of processing times predicted values Na, Nb, and Nc. For example, the prediction calculation device 30 calculates the total remaining number of processing times Nx by a weighted average by using each of the multiple score values Va, Vb, and Vc as a weight. In this case, the total remaining number of processing times Nx is calculated by Equation 1.
Nx=500×5/11+480×4/11+510×2/11=495 (Formula 1)

(6. Functional Block Configuration of Tool Life Prediction System 1)
The functional block configuration of the tool life prediction system 1 for realizing the above-mentioned tool life prediction process will be described with reference to Figures 5 to 8. As shown in Figure 5, the tool life prediction system 1 includes a detector 13, a counter 12a, a learning processing device 20, a prediction calculation device 30, and display devices 40 and 50.

As described above, the detector 13 detects observable status data in the machine body 11 while machining the workpiece W. The status data includes, for example, driving load data in the motor that drives the rotating tool T. The status data further includes, for example, driving load data in the motor that drives the workpiece W. The driving load data corresponds to the driving current data of the motor. The status data may also include vibration data, machining sound data, and the like. The status data is time-series data from the start to the end of machining of one workpiece W.

The counter 12a is included in the control device 12 of the processing machine 10, and counts the number of times the workpiece W has been processed for each tool T. In other words, the counter 12a counts the number of times the workpiece W has been processed for each tool T since the tool T was first used. The counter 12a can also be provided in the detector 13 itself, in addition to the control device 12.

The learning processing device 20 generates a life prediction model for predicting the life of the tool T based on the condition data detected by the detector 13 and the number of machining operations counted by the counter 12a. Furthermore, the learning processing device 20 generates a scoring model that outputs a score value of the input condition data based on the condition data detected by the detector 13.

As shown in FIG. 5, the learning processing device 20 includes a training dataset acquisition unit 71, a training dataset storage unit 72 for life prediction, a training dataset storage unit 73 for scoring, a life prediction model generation unit 74, and a scoring model generation unit 75.

The training dataset acquisition unit 71 acquires a training dataset for performing machine learning for each of a plurality of tools T. The plurality of tools T includes tools T of the same type, tools T of different types, etc. Regardless of whether the tools T are of the same type or different types, the life cases of each tool T are considered to be separate life cases, and acquiring a training dataset for each of the tools T means acquiring a training dataset for each of a plurality of life cases.

The training data set acquisition unit 71 includes a state data acquisition unit 71a and a feature calculation unit 71b. The state data acquisition unit 71a acquires state data detected by the detector 13 during machining of the workpiece W. The state data acquisition unit 71a acquires state data for the number of detectors 13 in one machining session with one tool T (e.g., machining of one workpiece W).

Then, the status data acquisition unit 71a acquires status data for each of the multiple machining operations until one tool T reaches the end of its life. In other words, the status data acquisition unit 71a acquires status data for multiple detectors 13 for the number of machining operations until one tool T reaches the end of its life. The status data acquisition unit 71a then acquires the above status data for each of the multiple tools T. In other words, the status data acquisition unit 71a acquires status data for each life case.

The feature amount calculation unit 71b calculates multiple feature amounts of the state data acquired by the state data acquisition unit 71a. As described above, the feature amounts are average values, etc.

The training data set acquisition unit 71 further includes a machining count acquisition unit 71c, a lifespan information acquisition unit 71d, and a remaining machining count calculation unit 71e. The machining count acquisition unit 71c acquires the number of machining operations from the start of machining for each tool T from the counter 12a. The lifespan information acquisition unit 71d is input by, for example, an operator who judges whether the target tool T has reached its lifespan. The judgment of whether the tool T has reached its lifespan can be made, for example, by whether scratches called tool marks have been formed on the surface of the workpiece W. In addition, the judgment can also be made by whether the machining accuracy of the workpiece W has significantly decreased. In addition to being input by an operator, the lifespan information acquisition unit 71d may be input by an inspection device when the judgment is made by an inspection device.

The remaining number of machining operations calculation unit 71e determines the remaining number of machining operations for the tool T based on the number of machining operations acquired by the number of machining operations acquisition unit 71c and the life end information acquired by the life end information acquisition unit 71d. The remaining number of machining operations is set to zero when the tool's life end is reached, and increases in ascending order going back to the time before the end of the tool's life. As shown in the left column of FIG. 6, if the tool is determined to have reached the end of its life when the number of machining operations is N, the remaining number of machining operations will be zero when the tool's life end is reached, and will start at (N-1), as shown in the right column of FIG. 6.

As shown in the right column of FIG. 6, the training dataset storage unit 72 for life prediction stores the feature quantities D(1) to D(N) and the remaining number of machining operations (N-1) to (0) of the condition data acquired by the training dataset acquisition unit 71 as a training dataset for life prediction for each of multiple life cases. Furthermore, the training dataset storage unit 72 for life prediction stores the feature quantities D(1) to D(N) of the condition data and the remaining number of machining operations (N-1) to (0) in association with each other. The contents shown in the right column of FIG. 6 are the training dataset for one tool T. The training dataset storage unit 72 for life prediction stores a training dataset for each of multiple tools T, i.e., for each of multiple life cases.

The scoring training dataset storage unit 73 stores the feature quantities D(1) to D(N) of the condition data acquired by the training dataset acquisition unit 71 as a scoring training dataset for each of the multiple life cases. In other words, the scoring training dataset storage unit 73 stores a training dataset for each of the multiple tools T, i.e., for each of the multiple life cases.

Note that the life prediction training dataset storage unit 72 and the scoring training dataset storage unit 73 are described separately for convenience, but they may be integrated. In other words, the integrated training dataset storage unit may store the feature quantities D(1) to D(N) of the status data and the remaining number of machining operations (N-1) to (0) in association with each other as a training dataset for life prediction, and store the feature quantities D(1) to D(N) of the status data as a training dataset for scoring. The feature quantities D(1) to D(N) of the status data are common data.

The life prediction model generation unit 74 performs machine learning using the life prediction training dataset stored in the life prediction training dataset storage unit 72. Specifically, for each life case, i.e., for each individual tool T, the life prediction model generation unit 74 performs machine learning using the feature quantities of the state data detected by the detector 13 as explanatory variables and the remaining number of machining operations as a target variable. Then, the life prediction model generation unit 74 generates a life prediction model for predicting the life of the tool T.

Here, a life prediction model is generated for each life case, i.e., for each individual tool T. Therefore, the life prediction model generation unit 74 generates life prediction models for each of the multiple tools T. In other words, the correspondence between the tools T and the life prediction models is as shown in FIG. 7. For example, tool Ta corresponds to life prediction model Ma1, and so on.

For example, regression may be used as a machine learning technique. For example, linear regression, ridge regression, Lasso, elastic net, random forest, support vector machine, etc. are useful. In particular, by using these techniques, the influence of multiple feature quantities can be grasped, and they can also be used for feature selection, etc., as necessary. Note that machine learning techniques other than regression can also be applied.

The scoring model generation unit 75 performs machine learning using the scoring training dataset stored in the scoring training dataset storage unit 73. Specifically, the scoring model generation unit 75 performs machine learning for each life case, i.e., for each individual tool T, using the feature quantities of the condition data detected by the detector 13 as explanatory variables. Then, the scoring model generation unit 75 generates a scoring model that outputs a score value of the condition data.

Here, the scoring model is generated for each life case, that is, for each individual tool T. Therefore, the scoring model generation unit 75 generates scoring models for each of the multiple tools T. That is, the correspondence between the tools T and the scoring models is as shown in FIG. 7. For example, tool Ta corresponds to scoring model Ma2, and so on. Furthermore, as shown in FIG. 7, the scoring model also corresponds to a life prediction model. As a machine learning technique, a model capable of outputting the degree of normality, such as one-class SVM, autoencoder, Mahalanobis distance, likelihood, etc., may be used.

The prediction calculation device 30 predicts the life of the tool T used in processing based on status data during processing in the corresponding processing machine 10. The prediction calculation device 30 includes a life prediction model storage unit 81, a scoring model storage unit 82, a prediction data acquisition unit 83, a remaining number of processing operations prediction unit 84, a score value output unit 85, a total remaining number of processing operations prediction unit 86, and an output unit 87.

As shown in FIG. 8, the life prediction model storage unit 81 stores multiple life prediction models Ma1, Mb1, Mc1, ... generated by the life prediction model generation unit 74. As described above, the multiple life prediction models Ma1, Mb1, Mc1, ... correspond to each life case, and in this example, correspond to each tool T.

As shown in FIG. 8, the scoring model storage unit 82 stores multiple scoring models Ma2, Mb2, Mc2, ... generated by the scoring model generation unit 75. As described above, the multiple scoring models Ma2, Mb2, Mc2, ... correspond to each life case, and in this example, correspond to each tool T, and also correspond to each life prediction model Ma1, Mb1, Mc1, ....

The prediction data acquisition unit 83 acquires prediction data during machining by the tool Tx to be predicted. The prediction data acquisition unit 83 includes a state data acquisition unit 83a, a feature amount calculation unit 83b, and a machining count acquisition unit 83c.

The status data acquisition unit 83a acquires status data detected by the detector 13 during machining of the workpiece W by the tool Tx to be predicted. Here, the tool Tx to be predicted may be the same type as the tool T used when it was acquired by the training data set acquisition unit 71, or may be a different type.

The feature amount calculation unit 83b calculates the feature amount of the state data acquired by the state data acquisition unit 83a. Here, the feature amount uses various statistics in the state data. The feature amount is the same type as the feature amount calculated by the feature amount calculation unit 71b of the training data set acquisition unit 71. The machining count acquisition unit 83c acquires the number of machining times from the start of machining with the tool Tx to be predicted from the counter 12a.

Here, the state data acquisition unit 83a and the feature amount calculation unit 83b perform the same processing as the state data acquisition unit 71a and the feature amount calculation unit 71b of the training data set acquisition unit 71. In this example, the state data acquisition unit 83a and the feature amount calculation unit 83b have been described as separate elements from the state data acquisition unit 71a and the feature amount calculation unit 71b of the training data set acquisition unit 71. However, it is also possible for the elements 71a and 71b of the training data set acquisition unit 71 to double as the elements 83a and 83b of the prediction data acquisition unit 83. In other words, the functions of the elements 71a and 71b in the learning processing device 20 are shared with some of the functions of the prediction calculation device 30.

The remaining number of machining operations prediction unit 84 outputs each of the remaining number of machining operations prediction values Na, Nb, Nc, ... of the tool Tx to be predicted by each of the multiple life prediction models Ma1, Mb1, Mc1, ... based on the state data of the target tool Tx acquired by the prediction data acquisition unit 83. Specifically, the remaining number of machining operations prediction unit 84 outputs multiple remaining number of machining operations prediction values Na, Nb, Nc, ... for the tool Tx to be predicted by using at least a part of the feature values calculated by the feature value calculation unit 83b and the number of machining operations from the start of machining acquired by the machining number acquisition unit 83c as input data for each of the multiple life prediction models Ma1, Mb1, Mc1, .... In this way, the remaining number of machining operations prediction unit 84 predicts the number of times the workpiece W can be machined by the tool Tx to be predicted for multiple patterns.

The score value output unit 85 outputs score values Va, Vb, Vc, ... by each of the multiple scoring models Ma2, Mb2, Mc2, ... based on the state data of the target tool Tx acquired by the prediction data acquisition unit 83. Specifically, the score value output unit 85 outputs score values Va, Vb, Vc, ... for the state data of the target tool Tx for each of the multiple scoring models Ma2, Mb2, Mc2, ... by using at least a part of the feature values calculated by the feature calculation unit 83b as input data. In other words, the score value output unit 85 outputs score values Va, Vb, Vc, ... for each of the multiple remaining machining count prediction values Na, Nb, Nc, ....

The total remaining number of processing times prediction unit 86 calculates the total remaining number of processing times Nx based on each of the multiple remaining number of processing times predicted values Na, Nb, Nc, ... predicted by the remaining number of processing times prediction unit 84 and each of the multiple score values Va, Vb, Vc, ... calculated by the score value output unit 85. For example, the total remaining number of processing times Nx is calculated by a weighted average, with each of the multiple score values Va, Vb, Vc, ... serving as weights for each of the multiple remaining number of processing times predicted values Na, Nb, Nc, .... The total remaining number of processing times Nx is not limited to a weighted average, and may be calculated using the multiple score values Va, Vb, Vc, ... as weights.

In addition, when all of the multiple score values are positive values indicating normality, the total remaining number of machining operations prediction unit 86 can perform a weighted average calculation using all the score values as described above. However, when some of the multiple score values have negative values indicating abnormality, a different processing method may be used. For example, the remaining number of machining operations predicted values whose score values are negative values indicating abnormality may be excluded, and the total remaining number of machining operations Nx may be calculated using a weighted average using the remaining remaining number of machining operations predicted values and the score values corresponding to the remaining number of machining operations predicted values. In addition, when a predetermined number of score values (a number equal to or greater than 1) are negative values indicating abnormality, it is also possible to predict that the target tool Tx is not in a normal wear state but in an abnormal state. An abnormal state of the target tool Tx is, for example, chipping, chipping, etc.

The output unit 87 outputs the total remaining number of machining operations Nx of the tool Tx to be predicted, which is predicted by the total remaining number of machining operations prediction unit 86, to the display device 40, 50. In this case, the display device 40, 50 displays the total remaining number of machining operations Nx, and the operator can grasp the total remaining number of machining operations Nx of the tool Tx to be predicted. Furthermore, the output unit 87 can also output the current number of machining operations, the score values of each of the life prediction models, and the like, in addition to the total remaining number of machining operations Nx, to the display device 40, 50.

The output unit 87 may also output the total remaining number of machining operations Nx of the tool Tx to be predicted to the control device (control unit) 12, and the control device 12 may execute a tool replacement or tool correction operation based on the total remaining number of machining operations Nx. For example, when the total remaining number of machining operations Nx reaches a life value (e.g., zero), the control device 12 executes tool replacement or tool correction. Note that tool replacement and tool correction are set according to the type of tool T, and in the case of a tool T that can be corrected, tool replacement is performed when the number of tool corrections reaches a predetermined number.

Here, tool correction may be performed by removing the tool T from the processing machine body 11, or may be performed in the processing machine body 11. For example, if the tool T is a cutting tool, the tool correction is a resharpening process of the tool T, which is usually performed by removing the tool T from the processing machine body 11. Also, if the tool T is a grinding wheel, the tool correction is truing or dressing of the grinding wheel, and the tool correction process is performed on a grinding machine equipped with a grinding wheel.

The output unit 87 may also notify the operator of the timing for tool replacement or tool correction based on the total remaining number of times of machining Nx. For example, the output unit 87 may notify the operator of the timing when the total remaining number of times of machining Nx approaches the life value in addition to the timing when it has reached the life value. This allows the operator to prepare for tool replacement or tool correction, or to prepare the tool T to be replaced.

According to the tool life prediction system 1, the life prediction models Ma1, Mb1, Mc1 and the scoring models Ma2, Mb2, Mc2 are stored in correspondence with each life case. The scoring models Ma2, Mb2, Mc2 output score values Va, Vb, Vc of the state data as explanatory variables. The score values Va, Vb, Vc output by the scoring models Ma2, Mb2, Mc2 correspond to the score values of the corresponding life prediction models Ma1, Mb1, Mc1, that is, the degree of conformance indicating how well the life prediction models Ma1, Mb1, Mc1 conform to the state data. For example, when the score values Va, Vb, Vc are large, it is estimated that the degree of conformance of the corresponding life prediction models Ma1, Mb1, Mc1 is high, and when the score values Va, Vb, Vc are small, it is estimated that the degree of conformance of the corresponding life prediction models Ma1, Mb1, Mc1 is low.

Then, the total remaining number of machining operations prediction unit 86 calculates the total remaining number of machining operations Nx based on each of the multiple remaining number of machining operations prediction values Na, Nb, Nc and each of the multiple score values Va, Vb, Vc. In other words, the total remaining number of machining operations prediction unit 86 can use multiple life prediction models Ma1, Mb1, Mc1 by considering the score values Va, Vb, Vc. In other words, the total remaining number of machining operations prediction unit 86 does not need to select only one life prediction model from the multiple life prediction models Ma1, Mb1, Mc1. In this way, the total remaining number of machining operations prediction unit 86 predicts the total remaining number of machining operations Nx using multiple life prediction models Ma1, Mb1, Mc1 and multiple score values Va, Vb, Vc, so that the life prediction of the tool T can be performed with high accuracy.

1: Tool life prediction system, 10: Processing machine, 11: Processing machine main body, 12: Control device (control unit), 13: Detector, 20: Learning processing device, 30: Prediction calculation device, 40, 50: Display device, 71: Training data set acquisition unit, 72: Training data set storage unit for life prediction, 73: Training data set storage unit for scoring, 74: Life prediction model generation unit, 75: Scoring model generation unit, 81: Life prediction model storage unit, 82: Scoring model storage unit, 83: Prediction data acquisition unit, 84: Remaining number of machining operations prediction unit, 85: Score value output unit, 86: Total remaining number of machining operations prediction unit, 87: Output unit, Ma1, Mb1, Mc1: Life prediction model, Ma2, Mb2, Mc2: Scoring model, Na, Nb, Nc: Remaining number of machining operations predicted value, Nx: Total remaining number of machining operations, T: Tool, Va, Vb, Vc: score value, W: workpiece

Claims (10)

A processing machine body that performs machining of a workpiece using a tool;
a detector for detecting observable status data of the processing machine body during processing of the workpiece;
a life prediction model storage unit that stores a life prediction model generated by performing machine learning using a training data set including the explanatory variables and a training data set including the explanatory variables and the objective variables, the life prediction model being generated by using the state data detected by the detector as explanatory variables and a remaining number of times the workpiece is machined until the tool reaches its life as an objective variable , for each of a plurality of life cases, where cases for each of the tools are considered as separate life cases; and
a scoring model storage unit that stores, for each of the plurality of life cases, a scoring model that is generated by performing machine learning using a training data set in which each of the state data in each of the life cases is an explanatory variable, and outputs a score value of the state data; and
a remaining machining count prediction unit that outputs a remaining machining count prediction value by each of the plurality of life prediction models based on the state data in the estimation phase;
a score value output unit that outputs a score value by each of the plurality of scoring models based on the state data in the estimation phase;
A total remaining processing count prediction unit that calculates a total remaining processing count based on each of the plurality of remaining processing count prediction values and each of the plurality of score values;
A tool life prediction system comprising: The tool life prediction system according to claim 1, wherein the total remaining number of machining operations prediction unit calculates the total remaining number of machining operations by a weighted average, with each of the multiple score values being used as a weight for each of the multiple remaining number of machining operations prediction values. The tool life prediction system according to claim 1 or 2, wherein the total remaining machining times prediction unit predicts that the tool is not in a normal wear state but in an abnormal state when a predetermined number of the multiple score values indicate an abnormality. The tool life prediction system according to any one of claims 1 to 3, wherein at least some of the multiple life prediction models use different feature quantities of the condition data as explanatory variables. The tool life prediction system according to claim 4, wherein the scoring model uses the same type of feature as the life prediction model for the corresponding life case as explanatory variables. The tool life prediction system according to any one of claims 1 to 5, wherein the scoring model is a one-class support vector machine and outputs the degree of normality as a score value. The tool life prediction system according to claim 6, wherein the scoring model is a model that further outputs the score value representing normality and the score value representing an abnormality as different values. The tool life prediction system further comprises:
The tool life prediction system according to any one of claims 1 to 7, further comprising a display device for displaying the total remaining number of machining operations. The tool life prediction system further comprises:
The tool life prediction system according to any one of claims 1 to 8, further comprising a control unit that executes an operation of replacing or correcting a tool based on the total remaining machining count. The tool life prediction system further comprises:
The tool life prediction system according to any one of claims 1 to 9, further comprising an output unit that notifies the timing of tool replacement or tool correction based on the total remaining number of machining operations.

Images