JP2003019643A - Processing state detection device, processing state detection method, recording medium, and program - Google Patents

Abstract

(57) [Summary] [PROBLEMS] To reduce the number of teacher data (learning data) required at the time of learning and estimate a machining state efficiently and with high accuracy. SOLUTION: A machining state detecting device 10 is provided with a cutting torque T.
or a torque sensor 16 for detecting the
, A sensor-side input parameter calculator 17 for calculating a power spectrum value TF, a cutting torque integrated value TI, and a cutting torque peak value TP in the vicinity of the natural vibration component of the tool system, and data on a machining state. A data storage unit 18, a data-side input parameter calculation unit 19 that calculates a data-side input parameter TD based on data obtained from the data storage unit 18, and a fuzzy inference between the sensor-side input parameter and the data-side input parameter TD And the output of fuzzy inference, drill 13
And a fuzzy inference unit 20 that uses an estimated value Es for determining whether the tool life has elapsed as a recurrent input.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention detects a machining state, such as a tool wear amount, a finished surface roughness of a work material, a finished dimensional accuracy, etc., when machining a work material using a tool such as a drill. The present invention relates to a machining state detecting device, a machining state detecting method, a program, and a recording medium having the program recorded therein.

[0002]

2. Description of the Related Art Conventionally, as a device for detecting damage to a tool of a machine tool, such as a tool damage detecting device disclosed in Japanese Patent Laid-Open No. 3-49850, A generated during machining is used.
2. Description of the Related Art There is known a damage detection device that includes an AE sensor that detects an E (Acoustic Emission) signal in the vicinity of a tool and that detects the damage to the tool by frequency analysis of the AE signal output from the AE sensor. With this tool damage detection device,
The AE signal output from the E sensor is converted into frequency components by FFT (Fast Fourier Transform) to generate a frequency spectrum. This frequency spectrum is input to a neural network that has been learned in advance with predetermined teacher data, and a predetermined signal corresponding to the presence or absence of damage to the tool is output.

[0003]

However, in the tool damage detection apparatus according to the above-mentioned conventional example, a predetermined signal from the output layer of the neural network according to the presence or absence of tool damage, for example, 0. It is necessary to perform learning of the neural network using a large amount of teacher data in advance so that an output value of 5 or more is output, and an output value of -0.5 or less is output in other cases.
There arises a problem that the complicated work of performing many preliminary experiments is required. The present invention has been made in view of the above circumstances, and it is possible to reduce the number of pieces of teacher data (learning data) required at the time of learning and efficiently and highly accurately estimate a processing state. An object of the present invention is to provide a detection device, a processing state detection method, a program, and a recording medium recording the program.

[0004]

In order to solve the above problems and achieve the object, a machining state detecting apparatus according to the present invention is a machine tool (for example, a machine tool in an embodiment described later). A cutting tool (for example, a drill in the embodiment to be described later) that is provided in a main shaft portion (for example, the main shaft portion 12 in the embodiment to be described later) or a table portion (for example, a table 14 in the embodiment to be described later) of the machine 11). 13) Information related to the processing state during cutting (for example, the cutting torque To in the embodiment described later)
r) information detecting means (for example, a torque sensor 16 in an embodiment described later) and an input parameter of fuzzy inference (for example, an embodiment described later) based on a detection signal output from the information detecting means. Power spectrum value TF near the natural vibration component of the tool system at
Sensor-side input parameter calculation means (for example, a sensor-side input parameter calculation unit 17 in an embodiment described later) for calculating the cutting torque integrated value TI and the cutting torque peak value TP, and machining during cutting by the cutting tool. A data storage unit that stores in advance data related to the state (for example, data that indicates a change between the machining time T and the tool wear amount A during drilling by the drill 13 in the embodiment described later), and the data storage unit (for example, A data-side input parameter calculation unit that calculates an input parameter for fuzzy inference based on the data stored in a data storage unit 18) in the later-described embodiment (for example, a data-side input parameter calculation unit in the later-described embodiment). 1
9) and the output data of the fuzzy inference (for example, a predetermined estimated value Es for determining whether or not the tool wear amount A of the drill 13 in the embodiment described later is a predetermined tool wear amount A0 or less). As an input, a recurrent fuzzy inference means that infers a processing state during cutting by the cutting tool based on the input parameter and the recurrent input and outputs the inference result as the output data (for example, an embodiment described later). And a fuzzy inference unit 20).

According to the processing state detecting apparatus having the above-mentioned configuration, the processing state during cutting by the cutting tool is, for example, the amount of wear of the tool, and further, the occurrence of an abnormal state at the time of processing due to this amount of wear, The finished surface roughness, finished dimensional accuracy, etc. of the work material can be accurately inferred based on the detection signal related to the working state and the data stored in advance. That is, as a detection signal related to the processing state, for example, a cutting torque or a thrust force during cutting is detected, and an input parameter on the sensor side is calculated based on these detection signals. Desired data is searched for and acquired from the accumulated data relating to the processing state, and the input parameter on the data side is calculated based on this data. Then, both of these input parameters are input to the recurrent fuzzy inference means. As a result, the accuracy of the recurrent fuzzy inference can be improved by using the data accumulated in advance, as compared with the case of inputting only the input parameters on the sensor side to the recurrent fuzzy inference means.

Further, in the machining state detecting apparatus of the present invention as set forth in claim 2, the sensor side input parameter calculating means and the data side input parameter calculating means synchronize the input parameters with each other to perform the recurrent operation. It is characterized by inputting to fuzzy inference means.

According to the processing state detecting apparatus having the above-mentioned structure, the data side input parameter calculating means inputs the input side in advance according to the sensor side input parameter inputted from the sensor side input parameter calculating means to the recurrent type fuzzy inference means. Predetermined data associated with the parameter is acquired from the data storage means to calculate the input parameter on the data side. Then, the calculated input parameter on the data side is input to the recurrent fuzzy inference means in synchronization with the input parameter on the sensor side. Accordingly, for example, even when each input parameter for the recurrent fuzzy inference changes according to the machining time, the machining state can be accurately detected in real time during cutting.

Further, in the machining state detecting apparatus of the present invention as set forth in claim 3, the information detecting means detects the cutting torque during cutting by the cutting tool and outputs it as a torque signal, and the sensor side. The input parameter calculation means calculates the power spectrum value in the vicinity of the natural vibration component of the tool system based on the torque signal output from the information detection means (for example, power spectrum value calculation means).
An FFT processor 33) in an embodiment described later,
And an integral value calculating means for calculating the cutting torque integral value by integrating the cutting torque for a predetermined time (for example, the waveform peak / waveform area calculating device 3 in the embodiment described later).
8), and a peak value calculating means for calculating a cutting torque peak value in the predetermined time (for example, the waveform peak / waveform area calculating device 38 also serves in an embodiment described later), and the power spectrum value and the cutting The torque integrated value and the cutting torque peak value are output as input parameters of the fuzzy inference, and the data storage means relates to the state of the cutting tool as data relating to the processing state during cutting by the cutting tool. Data (for example, the tool life due to the tool wear amount A in the embodiment described later) and / or data regarding the state of the work material (for example, the roundness of the through hole and the surface in the embodiment described later) Roughness and dimensional tolerance) are stored, and the data-side input parameter calculating means stores
Data relating to the state of the cutting tool and / or data relating to the state of the work material at a predetermined machining time are output as input parameters of the fuzzy inference.

According to the processing state detecting device having the above-mentioned configuration, the cutting tool can detect the power spectrum value near the natural vibration component of the tool system based on the detection signal of the cutting torque, the cutting torque integrated value and the cutting torque peak value. The processing state during cutting, for example, the amount of tool wear as the state of the cutting tool,
It is possible to estimate the finished surface roughness and the finished dimensional accuracy as the state of the work material. Here, in addition to the input parameter on the sensor side, the input parameter on the data side calculated based on the data on the state of the cutting tool and / or the data on the state of the work material is input. As a result, it is possible to improve the accuracy of recurrent fuzzy inference for estimating the machining state.

Further, in the machining state detecting method of the present invention according to claim 4, information relating to a machining state during cutting by a cutting tool (for example, a drill 13 in the embodiment to be described later) (for example, to be described later). Information detection step for detecting the cutting torque Tor in the above-mentioned embodiment, and an input parameter for fuzzy inference (for example, natural vibration component of the tool system in the embodiment described later) based on the detection signal output in the information detection step. A sensor-side input parameter calculating step for calculating a power spectrum value TF in the vicinity, a cutting torque integrated value TI, and a cutting torque peak value TP, and data relating to a processing state during cutting by the cutting tool (for example, described later). Drill 13 in the embodiment
A data storage step of pre-storing data indicating a change between the machining time T and the tool wear amount A in the drilling process by the step of calculating the input parameter of the fuzzy inference based on the data stored in the data storage step. Data-side input parameter calculation step and output data of the fuzzy inference (for example, a predetermined estimation for determining whether the tool wear amount A of the drill 13 in the embodiment described later is a predetermined tool wear amount A0 or less). A value Es) as a recurrent input, and a recurrent fuzzy inference step of inferring a machining state during cutting by the cutting tool based on the input parameter and the recurrent input and outputting the inference result as the output data. It is characterized by that.

According to such a machining state detecting method, as the machining state during the cutting by the cutting tool, for example, the wear amount of the tool, and further, the abnormal state at the time of machining and the tool life due to this wear amount are generated. The finished surface roughness and finished dimensional accuracy of the work material can be accurately estimated based on the detection signal related to the working state and the data stored in advance.

Further, in the machining state detecting method according to the present invention, the sensor side input parameter calculating step and the data side input parameter calculating step output the input parameters in synchronization with each other. Is characterized by.

According to such a machining state detecting method, for example, even when each input parameter for the recurrent fuzzy inference changes according to the machining time, the machining state is accurately detected in real time during cutting. be able to.

Further, in the machining state detecting method of the present invention as set forth in claim 6, in the information detecting step, the cutting torque during cutting by the cutting tool is detected and output as a torque signal. The side input parameter calculation step is based on the torque signal output in the information detection step, based on the power spectrum value near the natural vibration component of the tool system (for example, near the natural vibration component of the tool system in the embodiment described later. Power spectrum value of
F), a power spectrum value calculating step, and an integral value calculating step of calculating a cutting torque integrated value (for example, a cutting torque integrated value TI in an embodiment described later) by integrating the cutting torque for a predetermined time. and,
A peak value calculation step of calculating a cutting torque peak value (for example, a cutting torque peak value TP in an embodiment described later) at the predetermined time is included, and the power spectrum value, the cutting torque integrated value, and the cutting torque peak value are included. , Output as an input parameter of the fuzzy inference, the data storage step, as data relating to a processing state during cutting by the cutting tool, data relating to the state of the cutting tool (for example, in the embodiment described later. The tool life resulting from the tool wear amount A) and / or the data regarding the state of the work material (for example, the roundness of the through hole, the surface roughness, and the dimensional tolerance in the embodiments described later) are stored. And the data-side input parameter calculation step includes Data relating to, and / or is characterized by outputting data regarding the conditions of the workpiece as the input parameters of the fuzzy inference.

According to such a processing state detecting method, the cutting tool is operated according to the power spectrum value in the vicinity of the natural vibration component of the tool system based on the cutting torque detection signal, the cutting torque integrated value and the cutting torque peak value. The processing state during cutting, for example, the amount of tool wear as the state of the cutting tool,
It is possible to estimate the finished surface roughness and the finished dimensional accuracy as the state of the work material. Here, in addition to the input parameter on the sensor side, the input parameter on the data side calculated based on the data on the state of the cutting tool and / or the data on the state of the work material is input. As a result, it is possible to improve the accuracy of recurrent fuzzy inference for estimating the machining state.

According to a seventh aspect of the present invention, there is provided a computer-readable recording medium in which a computer is used as a cutting tool (for example, a drill 1 in an embodiment described later).
Information related to the processing state during cutting by 3) (for example,
An input parameter for fuzzy inference (for example, an operation described below) based on a detection signal output from an information detection unit (for example, a torque sensor 16 in an embodiment described below) that detects a cutting torque Tor in an embodiment described below. Sensor side input parameter calculating means for calculating a power spectrum value TF in the vicinity of the natural vibration component of the tool system, a cutting torque integrated value TI, and a cutting torque peak value TP (for example, the sensor side in the embodiment described later). The input parameter calculation unit 17) and data related to the processing state during cutting by the cutting tool (for example, a change in the processing time T and the tool wear amount A in the drilling by the drill 13 in the embodiment described later are shown. Data storage means for storing data in advance, and the data storage means (for example, described later) Data storage unit in the facilities of the Embodiment 1
8) Data-side input parameter calculation means for calculating an input parameter of fuzzy inference based on the data stored in 8) (for example, a data-side input parameter calculation unit 19 in an embodiment described later), and output data of the fuzzy inference. (For example, the drill 1 in the embodiment described later.
During the cutting process by the cutting tool based on the input parameter and the recurrent input, with a predetermined estimated value Es) for determining whether or not the tool wear amount A of 3 is a predetermined tool wear amount A0 or less. Is recorded, and a program for functioning as a recurrent fuzzy inference means (for example, a fuzzy inference unit 20 in an embodiment to be described later) which infers the processing state and outputs the inference result as the output data is recorded. .

According to the recording medium having the above-mentioned structure, the machining state during cutting by the cutting tool is, for example, the wear amount of the tool, and further, the occurrence of an abnormal state at the time of machining due to this wear amount, the tool life, The finished surface roughness, finished dimensional accuracy, etc. of the work material can be accurately inferred based on the detection signal related to the working state and the data stored in advance.

Further, in the computer-readable recording medium of the present invention as set forth in claim 8, the sensor-side input parameter calculating means and the data-side input parameter calculating means synchronize the input parameters with each other and It is characterized by inputting to recurrent fuzzy inference means.

According to the recording medium having the above-mentioned configuration, the data-side input parameter calculating means preliminarily sets the input parameter in accordance with the sensor-side input parameter input from the sensor-side input parameter calculating means to the recurrent fuzzy inference means. The associated predetermined data is acquired from the data storage means and the input parameter on the data side is calculated. Then, the calculated input parameter on the data side is input to the recurrent fuzzy inference means in synchronization with the input parameter on the sensor side. Accordingly, for example, even when each input parameter for the recurrent fuzzy inference changes according to the machining time, the machining state can be accurately detected in real time during cutting.

Further, in the computer-readable recording medium of the present invention according to claim 9, the information detecting means detects the cutting torque during cutting by the cutting tool and outputs it as a torque signal. The sensor-side input parameter calculation means calculates the power spectrum value in the vicinity of the natural vibration component of the tool system based on the torque signal output from the information detection means (for example, an embodiment described later). FFT processor 33), and an integral value calculating means for calculating the integrated value of the cutting torque by integrating the cutting torque for a predetermined time (for example, a waveform peak / waveform area calculating device 38 in an embodiment described later), Peak value calculating means for calculating the peak value of the cutting torque at the predetermined time (for example, as described below In this state, the waveform peak / waveform area calculating device 38 also serves as the input device, and outputs the power spectrum value, the cutting torque integrated value, and the cutting torque peak value as input parameters of the fuzzy inference, and the data storage means. Is data related to the state of the cutting tool (for example, the tool life due to the tool wear amount A in the embodiment described later) as data relating to the machining state during cutting by the cutting tool, and / or the workpiece. The data regarding the state of the material (for example, the roundness of the through hole, the surface roughness, and the dimensional tolerance in the embodiments described later) is stored, and the data-side input parameter calculating means stores the cutting in a predetermined processing time. The fuzzy inference is performed on data regarding the state of the tool and / or data regarding the state of the work material. It is characterized in that output as an input parameter.

According to the recording medium having the above structure, the cutting process by the cutting tool is performed by the power spectrum value in the vicinity of the natural vibration component of the tool system based on the detection signal of the cutting torque, the cutting torque integral value, and the cutting torque peak value. It is possible to estimate the amount of tool wear as the state of machining, for example, the state of the cutting tool, and the finished surface roughness, finished dimensional accuracy, etc. as the state of the work material. Here, in addition to the input parameters on the sensor side, data regarding the state of the cutting tool, and /
Alternatively, the input parameter on the data side calculated based on the data regarding the state of the work material is input. As a result, it is possible to improve the accuracy of recurrent fuzzy inference for estimating the machining state.

Further, the program of the present invention according to claim 10 is such that information relating to a processing state during cutting by a cutting tool (for example, a drill 13 in an embodiment described later) is executed by a computer (for example, an operation described later). Of the cutting torque Tor in the above embodiment, based on the detection signal output from the information detecting means (for example, the torque sensor 16 in the embodiment described later), the input parameter of the fuzzy inference (for example, in the embodiment described later). A sensor-side input parameter calculation means (for example, a sensor-side input parameter calculation in an embodiment to be described later) that calculates a power spectrum value TF near the natural vibration component of the tool system, a cutting torque integrated value TI, and a cutting torque peak value TP). Part 17) and data relating to a processing state during cutting by the cutting tool (for example, Data storage means for storing in advance a data indicating a change in the machining time T and the tool wear amount A in the drilling process by the drill 13 in the embodiment described later, and the data storage means (for example, in the embodiment described later. Data-side input parameter calculating means for calculating an input parameter for fuzzy inference based on the data stored in the data storage unit 18) (for example, a data-side input parameter calculation unit 19 in an embodiment described later).
And output data of the fuzzy inference (for example, a predetermined estimated value Es for determining whether or not the tool wear amount A of the drill 13 in the embodiment described later is a predetermined tool wear amount A0 or less) as a recurrent input. , A recurrent fuzzy inference means that infers a machining state during cutting by the cutting tool based on the input parameter and the recurrent input, and outputs the inference result as the output data (for example, fuzzy in the embodiment described later). The feature is that it functions as an inference unit 20).

According to the program having the above-mentioned configuration, the machining state during cutting by the cutting tool is, for example, the wear amount of the tool, and further, the occurrence of an abnormal state at the time of machining due to this wear amount, the tool life, and the cutting. The finished surface roughness and finished dimensional accuracy of the material can be accurately inferred based on the detection signal related to the processing state and the data stored in advance.

Further, in the program of the present invention as set forth in claim 11, the sensor side input parameter calculating means,
The data-side input parameter calculating means is characterized in that the input parameters are input to the recurrent fuzzy inference means in synchronization with each other.

According to the program having the above-mentioned configuration, the data-side input parameter calculating means corresponds in advance to this input parameter in accordance with the sensor-side input parameter input from the sensor-side input parameter calculating means to the recurrent fuzzy inference means. The attached predetermined data is acquired from the data storage means and the input parameter on the data side is calculated. Then, the calculated input parameter on the data side is input to the recurrent fuzzy inference means in synchronization with the input parameter on the sensor side. Accordingly, for example, even when each input parameter for the recurrent fuzzy inference changes according to the machining time, the machining state can be accurately detected in real time during cutting.

Further, in the program of the present invention as set forth in claim 12, the information detecting means detects the cutting torque during cutting by the cutting tool and outputs it as a torque signal, and the sensor side input parameter calculation The means is a power spectrum value calculating means for calculating a power spectrum value in the vicinity of the natural vibration component of the tool system based on the torque signal output from the information detecting means (for example, the FFT processor 33 in the embodiment described later). , And an integral value computing means for computing the cutting torque integral value by integrating the cutting torque for a predetermined time (for example, a waveform peak / waveform area computing device 38 in an embodiment described later),
And a peak value calculating means for calculating the cutting torque peak value in the predetermined time (for example, the waveform peak / waveform area calculating device 38 also serves as an embodiment described later).
The power spectrum value and the cutting torque integral value and the cutting torque peak value are output as an input parameter of the fuzzy inference, and the data storage means changes the processing state during cutting by the cutting tool. As such data, data relating to the state of the cutting tool (for example, tool wear amount A in the embodiment described later)
Data) and / or data relating to the state of the work material (for example, the roundness of the through hole, the surface roughness, and the dimensional tolerance in the embodiments described later) are stored. The side input parameter calculating means is characterized by outputting data regarding a state of the cutting tool and / or data regarding a state of the work material at a predetermined machining time as input parameters of the fuzzy inference.

According to the program having the above-described configuration, during the cutting process by the cutting tool, the power spectrum value near the natural vibration component of the tool system based on the detection signal of the cutting torque, the cutting torque integrated value, and the cutting torque peak value are used. It is possible to estimate the tool wear amount as the machining state, for example, the state of the cutting tool, and the finished surface roughness and the finished dimensional accuracy as the state of the work material. Here, in addition to the input parameters on the sensor side, data regarding the state of the cutting tool, and /
Alternatively, the input parameter on the data side calculated based on the data regarding the state of the work material is input. As a result, it is possible to improve the accuracy of recurrent fuzzy inference for estimating the machining state.

[0028]

BEST MODE FOR CARRYING OUT THE INVENTION An embodiment of a machining state detecting device and a machining state detecting method of the present invention will be described below with reference to the accompanying drawings. FIG. 1 is a block diagram showing a processing state detecting device 10 according to an embodiment of the present invention.
FIG. 6 is a graph showing an example of data stored in the data storage unit 18.

Processing state detecting apparatus 10 according to the present embodiment
Is a work material 15 placed on a table 14 by means of a drill 13 mounted on the spindle 12 of the machine tool 11, for example.
A predetermined input parameter is calculated based on a sensor (for example, a torque sensor 16 that detects a cutting torque Tor) that detects various physical quantities related to a processing state when drilling a workpiece and a detection signal that is output from the torque sensor 16. A sensor side sensor side input parameter calculation section 17, a data storage section 18, and a data side input parameter calculation section 19 for calculating a predetermined input parameter based on data acquired by a search in the data storage section 18, The fuzzy inference unit 20 and the output unit 21 are provided. In addition,
In FIG. 1, the torque sensor 16 is replaced by the spindle unit 1 of the machine tool 11.
The torque sensor 16 is incorporated in the work material 15
It may be incorporated between the table 14 and the table 14, or the table 14 or the like.

The sensor-side input parameter calculation unit 17 calculates, for example, the cutting torque Tor of the drill 13 output from the torque sensor 16 when performing drilling for forming a plurality of through holes in the work material 15. Based on the detection signal, the power spectrum value TF near the natural vibration component of the tool system,
The cutting torque integrated value TI and the cutting torque peak value TP are calculated. Then, these values are output as sensor-side input parameters to the fuzzy inference unit 20. Here, the sensor side input parameter calculation unit 17 includes a band pass filter 31, an A / D converter 32, an FFT processor 33, a low pass filter 36, and an A / D converter 37.
And a waveform peak / waveform area calculation device 38.

The detection signal of the cutting torque Tor output from the torque sensor 16 is input to the bandpass filter 31 and the lowpass filter 36. The bandpass filter 31 extracts only a predetermined frequency component of the input detection signal, particularly, only the vicinity of the natural vibration component of the tool system, and A / D
Input to the converter 32. The A / D converter 32 performs A / D conversion on the detection signal of the cutting torque Tor input via the bandpass filter 31, and FFs the converted detection signal.
Input to the T processor 33. FFT processor 33
Is the cutting torque T input through the A / D converter 32.
A fast Fourier transform is performed on the detection signal of or to calculate the vibration component and the power spectrum value TF.

The low-pass filter 36 removes a high frequency component, which is a noise component, from the input detection signal to obtain A /
Input to the D converter 37. The A / D converter 37 performs A / D conversion on the detection signal of the cutting torque Tor input via the low-pass filter 36, and inputs the converted detection signal to the waveform peak / waveform area calculation device 38. The waveform peak / waveform area calculation device 38 uses the waveform indicating the time change of the detection signal of the cutting torque Tor input via the A / D converter 37 to determine the time per predetermined rotation speed of the drill 13. Integral value of the cutting torque Tor obtained by integrating in, for example, the cutting torque integrated value TI obtained by integrating the time per one rotation of the drill 13, and the peak value of the waveform within this integration range, that is, the cutting torque peak Calculate the value TP.

The data storage unit 18 stores, for example, data relating to a predetermined machining state accumulated in the past machining work in association with a predetermined parameter. For example,
As shown in FIG. 2, the data storage unit 18 stores data indicating a change in the machining time T and the tool wear amount A in the drilling by the drill 13, and in this data, the predetermined machining time T0 is stored. At, a predetermined tool wear amount A0, which is the life of the drill 13, is reached. The data-side input parameter calculator 19 is a sensor-side input parameter calculator 17.
Data related to various input parameters (sensor-side input parameters) input to the fuzzy inference unit 20 from the data storage unit 18 is acquired, and processing such as synchronization with the sensor-side input parameter is performed to perform data-side input parameter processing. TD
Is input to the fuzzy inference unit 20. For example, the data-side input parameter calculation unit 19 indicates a change between the processing time T and the tool wear amount A (for example, the wear amount of the outer peripheral corner portion of the drill 13) in the drilling process by the drill 13 as shown in FIG. When the data is acquired, the tool wear amount A corresponding to the machining time T at this point is calculated and synchronized with the sensor side input parameter input from the sensor side input parameter calculation unit 17 to the fuzzy inference unit 20. Input to the fuzzy inference unit 20.

The fuzzy inference unit 20 is, as will be described later,
The sensor side input parameter calculated by the sensor side input parameter calculation unit 17, for example, the power spectrum value TF,
The three input parameters including the cutting torque integrated value TI and the cutting torque peak value TP, and the data side input parameter TD calculated by the data side input parameter calculation unit 19, for example, the input parameter including the tool wear amount A are fuzzy. This is a parameter output from the fuzzy inference unit 20 as an inference input, and a predetermined estimated value Es for determining whether the tool wear amount A of the drill 13 is less than or equal to a predetermined tool wear amount A0 is recurrently input. As a fuzzy reasoning. Here, the predetermined estimated value Es for determining whether the tool wear amount A of the drill 13 is equal to or less than the predetermined tool wear amount A0 is, for example, a value of either Es = 0.9 or Es = 0.1. When it is estimated that the tool wear amount A of the drill 13 exceeds the predetermined tool wear amount A0 and reaches the tool life, Es = 0.9 is output and the drill 13
Is estimated not to reach the tool life, Es =
It is designed to output 0.1. The output unit 21 outputs, based on the estimated value Es output from the fuzzy inference unit 20, whether or not the drill 13 has reached the tool life.

Processing state detecting apparatus 10 according to the present embodiment
Is provided with the above configuration, and the operation of the processing state detecting apparatus 10 will be described below. First, the recurrent fuzzy inference processing executed by the fuzzy inference unit 20 will be described with reference to the accompanying drawings. FIG. 3 is a conceptual diagram showing the process of recurrent fuzzy inference according to the present embodiment, and FIG. 4 is a graph diagram showing a membership function expressed by a radial basis function.

As shown in FIG. 3, in the recurrent fuzzy inference according to the present embodiment, the output y _t from the fuzzy inference (for example, a predetermined value for determining whether or not the drill 13 has reached the tool life) is output. Is used as a recurrent input for fuzzy inference, and a predetermined input parameter (for example, power spectrum value TF) and cutting torque integrated value TI based on measurement information from a sensor (for example, torque sensor 16). , The cutting torque peak value TP, and the data-side input parameter TD) are the fuzzy inference inputs x ₁ , ..., X _i (i is an arbitrary natural number), the fuzzy inference output (for example, the drill 13 is a tool). predetermined estimate recurrent input y consisting _{Es) t-1 ,, y to} (o is call any natural number) for determining whether or not the reached life To use as information from the sensor.

Here, as shown in FIG. 4, the coefficients a _nj and b
_{Let nj} and c _nj (j is an arbitrary natural number) be a coefficient that determines the shape of the membership function, and let I _n be a fuzzy reasoning input and a recurrent input, the membership function of the i th input in the j th fuzzy rule. (That is, a function that represents the degree to which an element belongs to a fuzzy set) is
It is expressed as the following mathematical formulas (1) and (2).

[0038]

[Equation 1]

[0039]

[Equation 2]

Where μ _ij is j for the i-th input
Represents the goodness of fit of the second fuzzy rule. Also, the membership function of the o-th recurrent input in the j-th fuzzy rule is expressed by the following mathematical expressions (3) and (4).

[0041]

[Equation 3]

[0042]

[Equation 4]

Here, μoj represents the fitness of the jth fuzzy rule with respect to the oth recurrent input. Then, the goodness of fit and the output value of the j-th fuzzy rule are expressed by the following mathematical expressions (5) and (6). Note that w _j is a real value of the consequent part of the fuzzy rule.

[0044]

[Equation 5]

[0045]

[Equation 6]

The learning of recurrent fuzzy inference according to the present embodiment, in particular, the processing of rule addition type learning will be described below with reference to the accompanying drawings. FIG. 5 shows learning of recurrent fuzzy inference according to the present embodiment, in particular,
6 is a flowchart showing a process of rule addition type learning, and FIG. 6 is a graph showing a change in the shape of the membership function according to the goodness of fit of the fuzzy rule.
FIG. 6A is a graph showing a membership function when the degree of conformity of an existing fuzzy rule is relatively low, and FIG.
FIG. 7B is a graph showing a membership function when the existing fuzzy rule has a relatively high conformity, and FIG. 7 shows the shape of a new fuzzy rule to be added according to the existing fuzzy rule. It is a graph which shows the change of.

Learning of the recurrent fuzzy inference according to this embodiment is performed by the so-called steepest descent method. First,
In step S01 shown in FIG. 5, fuzzy inference is executed by using predetermined input parameters based on measurement information from a plurality of sensors as fuzzy inference inputs x ₁ , ..., X _i , that is, learning data. Next, in step S02, the membership function is updated based on the output value y _t and the target output value * y _t for the learning data. That is, as shown in the following formula (7), the error function E _t is represented by the output value y _t and the target output value * y _t for the t-th learning data.

[0048]

[Equation 7]

Then, the consequent part real value w _j is updated as shown in the following mathematical expression (8). Note that k _w is a learning coefficient, and the previous value of the consequent real number w _j is w _j ′.

[0050]

[Equation 8]

Further, as shown in the following formulas (9) to (11), the membership functions in the antecedent part are updated by updating the coefficients a _nj , b _nj , and c _nj . Note that k _a , k _b ,
The k _c a learning coefficient, the coefficients a _nj, b _nj, respectively a _nj preceding value of c _nj ', b _nj', and the c _nj '.

[0052]

[Equation 9]

[0053]

[Equation 10]

[0054]

[Equation 11]

Next, in step S03, it is determined whether or not a new fuzzy rule is added. If this determination result is "YES", the process proceeds to step S04, while if the determination result is "NO", the series of processes is terminated. Then, when adding a new fuzzy rule, at the learning point where the error described by the error function Et is the maximum, the value λ of the highest degree of conformity of the fuzzy rule is searched and the following formula (12) is obtained. As shown, it is determined whether or not a predetermined condition for each of the predetermined threshold values Th _E and Th _t set in advance for the error function Et and the fitness value λ is satisfied.

[0056]

[Equation 12]

In step S04, a new fuzzy rule is added and the series of processing is completed. here,
Each coefficient indicating the shape of the membership function of the new fuzzy rule is set in advance with the fuzzy inference input x _i and the recurrent input y _to at the learning point where the error is maximum, as shown in the following mathematical expression (13). It is represented by the predetermined coefficient C and the value λ of the highest fitness of the fuzzy rule.

[0058]

[Equation 13]

At this time, the conformity of the existing fuzzy rule
Is relatively low, for example, as shown in FIG.
In addition, the membership function with wide dead zone and gentle slope
Is generated. On the other hand, the compatibility of existing fuzzy rules is
When it is relatively high, for example, as shown in FIG.
Membership functions with narrow dead zones and steep slopes are generated.
Be done. That is, for example, as shown in FIG.
If it is relatively low, a new fuzzy rule (μ_{nj + 2}: L
ow) and existing fuzzy rules (μ_nj) Is
If the goodness of fit is relatively high, new
Na fuzzy rule (μ _{nj + 1}: High) and existing fuzzy
Irrule (μ_nj) And the distance between. others
For example, as shown in FIG.
Rule (μ_{nj + 1}: High, μ_{nj + 2}: Low) shape
Existing fuzzy rules (μ_nj) Suitable for
Change as appropriate.

As described above, according to the machining state detecting apparatus 10 of the present embodiment, the occurrence of an abnormal state during machining due to wear of the drill 13 is detected by the torque sensor 16 as a detection signal. It is possible to make an accurate inference based on the data stored in advance in the data storage unit 18. That is, as compared with the case where only the sensor-side input parameter is input to the fuzzy inference unit 20, the accuracy of the recurrent fuzzy inference can be improved by inputting the data-side input parameter based on the data accumulated in advance together with the sensor-side input parameter. Can be improved.

Further, according to the above-described machining state detecting method according to the present embodiment, the power spectrum value TF and the cutting torque integrated value TI in the vicinity of the natural vibration component of the tool system calculated based on the detection signal of the cutting torque Tor. By using only four pieces of information, the cutting torque peak value TP and the data-side input parameter TD including the tool wear amount A, as the input parameters of the fuzzy inference, the fuzzy inference can be executed efficiently with high accuracy. Therefore, it is possible to simplify the process of determining the tool life due to tool wear, reduce the calculation load, and efficiently execute fuzzy inference.

In the present embodiment described above,
Although the torque sensor 16 is provided as the sensor, the present invention is not limited to this, and another sensor such as a thrust force sensor for detecting the thrust force of the drill 13 may be provided. Further, when a plurality of different sensors are provided, the accuracy of fuzzy inference can be improved.

In the present embodiment described above,
Although the tool life due to the tool wear amount A of the drill 13 was estimated by fuzzy reasoning, the present invention is not limited to this. For example, the roundness of the through hole, the surface roughness, the dimensional tolerance, etc. The state of the cutting material may be estimated. In this case, the data storage unit 18 may store, for example, processing time T in drilling with the drill 13 and data indicating changes in the roundness of the through hole, the surface roughness, the dimensional tolerance, and the like. .

In the above-described embodiment,
In the fuzzy reasoning unit 20, the drill 13 has a predetermined tool wear amount A.
Es = 0.9 is output when the tool life is estimated to exceed 0 and Es = 0.9 is output when the drill 13 is not reached the tool life.
= 0.1 is output, but the present invention is not limited to this, and the tool life and two values corresponding to less than the tool life (that is, E
In addition to s = 0.1 and Es = 0.9, for example, a value corresponding to a state between tool life and less than tool life (eg, E
(s = 0.5, etc.) may be output, and in this case, fuzzy inference is performed using at least three different values as recurrent inputs.

The machining state detecting apparatus 10 for realizing the processing and learning of the recurrent fuzzy inference according to the embodiment of the present invention may be realized by dedicated hardware, and the memory and The function may be realized by loading a program for realizing the function of the processing state detecting device 10 into a memory and executing the program, which is configured by a CPU.

Further, a program for realizing the above-mentioned machining state detecting method according to the present invention is recorded in a computer-readable recording medium, and the program recorded in this recording medium is read into a computer system and executed. Therefore, recurrent fuzzy inference processing and learning may be performed. The computer system referred to herein may include an OS and hardware such as peripheral devices.

Computer-readable recording media include flexible disks, magneto-optical disks, and ROs.
M, a portable medium such as a CD-ROM, and a storage device such as a hard disk built in a computer system. Further, a computer-readable recording medium is one that dynamically holds a program for a short time, such as a communication line when transmitting the program through a network such as the Internet or a communication line such as a telephone line, In this case, the volatile memory inside the computer system that serves as a server or a client in this case includes those holding a program for a certain period of time. Further, the program may be for realizing a part of the above-described functions, and may be one that can realize the above-mentioned functions in combination with a program already stored in the computer system. .

[0068]

As described above, according to the machining state detecting apparatus of the present invention as set forth in claim 1, the machining state during cutting by the cutting tool is, for example, the wear amount of the tool, and further, this wear amount. Accurately infer the occurrence of abnormal conditions during machining, tool life, finished surface roughness of workpieces and finished dimensional accuracy, etc., based on detection signals related to this machining state and pre-stored data. You can Further, according to the machining state detecting apparatus of the present invention as set forth in claim 2, even when each input parameter for the recurrent fuzzy inference changes according to the machining time,
It is possible to accurately detect the processing state in real time during cutting. Further, according to the machining state detecting apparatus for machining state of the present invention as set forth in claim 3, the power spectrum value in the vicinity of the natural vibration component of the tool system based on the detection signal of the cutting torque, the cutting torque integrated value, and the cutting torque peak. Depending on the value, the processing state during cutting by the cutting tool,
For example, the amount of tool wear can be estimated as the state of the cutting tool, and the finished surface roughness, the finished dimensional accuracy, etc. can be estimated as the state of the work material. Here, in addition to the input parameters on the sensor side, by inputting the data on the state of the cutting tool and / or the input parameter on the data side calculated based on the data on the state of the work material, the machining state can be changed. The accuracy of the estimated recurrent fuzzy reasoning can be improved.

Further, according to the machining state detecting method of the machining state of the present invention as set forth in claim 4, the machining state during cutting by the cutting tool is caused by, for example, the wear amount of the tool, and further, the wear amount. Occurrence of abnormal state during machining, tool life, finish surface roughness of work material, finish dimensional accuracy, etc.
It is possible to make an accurate inference based on the detection signal related to the processing state and the data stored in advance. Further, according to the processing state detecting method of the processing state of the present invention as set forth in claim 5, even when each input parameter for the recurrent fuzzy inference changes according to the processing time, for example, in real time during cutting. The processing state can be detected with high accuracy. Further, according to the machining state detecting method of the machining state of the present invention as set forth in claim 6, the power spectrum value in the vicinity of the natural vibration component of the tool system based on the detection signal of the cutting torque, the cutting torque integrated value, and the cutting torque peak. Based on the value, it is possible to estimate the processing state during cutting by the cutting tool, for example, the amount of tool wear as the state of the cutting tool, and the finished surface roughness and the finished dimensional accuracy as the state of the work material. Here, in addition to the input parameters on the sensor side, by inputting the data on the state of the cutting tool and / or the input parameter on the data side calculated based on the data on the state of the work material, the machining state can be changed. The accuracy of the estimated recurrent fuzzy reasoning can be improved.

According to the computer-readable recording medium of the present invention as set forth in claim 7, as the processing state during cutting by the cutting tool, for example, the amount of wear of the tool, and further the processing caused by this amount of wear Occurrence of an abnormal state, tool life, finished surface roughness of the work material, finished dimensional accuracy, etc. can be accurately inferred based on the detection signal related to the working state and prestored data.
Further, according to the computer-readable recording medium of the present invention as set forth in claim 8, for example, even when each input parameter for the recurrent fuzzy inference changes according to the machining time, the accuracy is real-time during cutting. The processing state can be detected well. further,
According to the computer-readable recording medium of the present invention described in claim 9, the power spectrum value near the natural vibration component of the tool system based on the detection signal of the cutting torque, the cutting torque integrated value, and the cutting torque peak value are used. It is possible to estimate the machining state during cutting by the cutting tool, for example, the amount of tool wear as the state of the cutting tool, and the finished surface roughness and the finished dimensional accuracy as the state of the work material. Here, in addition to the input parameters on the sensor side, by inputting the data on the state of the cutting tool and / or the input parameter on the data side calculated based on the data on the state of the work material, the machining state can be changed. The accuracy of the estimated recurrent fuzzy reasoning can be improved.

According to the program of the present invention as set forth in claim 10, as the processing state during the cutting by the cutting tool, for example, the amount of wear of the tool, and further, the abnormal state at the time of processing due to this amount of wear Occurrence, tool life, finished surface roughness of the work material, finished dimensional accuracy, etc. can be accurately inferred based on the detection signal related to this working state and prestored data. Further, according to the program of the present invention as set forth in claim 11, even when each input parameter for the recurrent fuzzy inference changes according to the machining time, for example, the machining state is accurately detected in real time during cutting. can do. Furthermore, according to the program of the present invention according to claim 12,
The power spectrum value near the natural vibration component of the tool system based on the detection signal of the cutting torque, the cutting torque integral value, and the cutting torque peak value are used to determine the machining state during cutting by the cutting tool, for example, the state of the cutting tool. It is possible to estimate the finished surface roughness, the finished dimensional accuracy, etc. as the wear amount and the state of the work material. Here, in addition to the input parameters on the sensor side, by inputting the data on the state of the cutting tool and / or the input parameter on the data side calculated based on the data on the state of the work material, the machining state can be changed. The accuracy of the estimated recurrent fuzzy reasoning can be improved.

[Brief description of drawings]

FIG. 1 is a configuration diagram showing a processing state detection device according to an embodiment of the present invention.

FIG. 2 is a graph showing an example of data stored in a data storage unit.

FIG. 3 is a conceptual diagram showing a process of recurrent fuzzy inference according to the present embodiment.

FIG. 4 is a graph showing a membership function represented by a radial basis function.

FIG. 5 is a flowchart showing learning of recurrent fuzzy inference according to the present embodiment, particularly processing of rule addition type learning.

FIG. 6A is a graph showing a membership function in the case where the existing fuzzy rules have a relatively low conformance, and FIG. 6B shows the existing fuzzy rules that have a relatively good conformance. It is a graph which shows the membership function when it is very high.

FIG. 7 is a graph showing a change in the shape of a new fuzzy rule to be added according to the conformity of an existing fuzzy rule.

[Explanation of symbols]

10 Processing State Detection Device 11 Machine Tool 12 Spindle Part 13 Drill (Cutting Tool) 14 Table (Table Part) 16 Torque Sensor (Information Detecting Unit) 17 Sensor-Side Input Parameter Calculation Unit (Sensor-side Input Parameter Calculation Means) 18 Data Storage Unit (Data storage means) 19 Data side input parameter calculation section (Data side input parameter calculation means) 20 Fuzzy inference section (Recurrent type fuzzy inference means) 33 FFT processor (Power spectrum value calculation means) 38 Waveform peak / waveform area calculation device ( (Integral value calculation means, peak value calculation means)

Claims (12)

[Claims] 1. An information detecting means, which is provided on a main shaft portion or a table portion of a machine tool, for detecting information on a processing state during cutting by a cutting tool, and based on a detection signal output from the information detecting means. ,
Based on the sensor side input parameter calculating means for calculating the input parameter of the fuzzy reasoning, the data storing means for pre-storing the data relating to the processing state during the cutting by the cutting tool, and the data stored in the data storing means. A data-side input parameter calculating means for calculating an input parameter of fuzzy inference, and using the output data of the fuzzy inference as a recurrent input, infer a machining state during cutting by the cutting tool based on the input parameter and the recurrent input. Then
A machining state detecting device comprising: a recurrent fuzzy inference means for outputting the inference result as the output data. 2. The sensor-side input parameter calculation means and the data-side input parameter calculation means input the input parameters to the recurrent fuzzy inference means in synchronization with each other. The processing state detection device described. 3. The information detecting means detects a cutting torque during cutting by the cutting tool and outputs it as a torque signal, and the sensor-side input parameter calculating means outputs the information from the information detecting means. A power spectrum value calculating means for calculating a power spectrum value near the natural vibration component of the tool system based on the torque signal; and an integral value calculating means for calculating the cutting torque integral value by integrating the cutting torque in a predetermined time period, And, comprising a peak value calculating means for calculating the cutting torque peak value in the predetermined time, the power spectrum value and the cutting torque integrated value and the cutting torque peak value is output as an input parameter of the fuzzy inference, The data storage means stores data relating to a processing state during cutting by the cutting tool, Data about the state of the cutting tool, and / or stores data relating to the state of the workpiece, the data side input parameter calculating means, data relating to the state of the cutting tool in a predetermined processing time, and /
Alternatively, the processing state detecting apparatus according to claim 1 or 2, wherein data regarding a state of the work material is output as an input parameter of the fuzzy inference. 4. A sensor side for calculating an input parameter of fuzzy inference based on an information detecting step for detecting information on a processing state during cutting by a cutting tool, and a detection signal output in the information detecting step. An input parameter calculation step, a data storage step of pre-storing data relating to a machining state during cutting by the cutting tool, and an input parameter of fuzzy inference based on the data stored in the data storage step Data-side input parameter calculation step, the output data of the fuzzy inference as a recurrent input, infer the processing state during cutting by the cutting tool based on the input parameter and the recurrent input,
And a recurrent type fuzzy inference step of outputting the inference result as the output data. 5. The machining state detecting method according to claim 4, wherein the sensor-side input parameter calculating step and the data-side input parameter calculating step output the input parameters in synchronization with each other. 6. The information detecting step detects a cutting torque during cutting by the cutting tool and outputs it as a torque signal, and the sensor-side input parameter calculating step is output in the information detecting step. Based on the torque signal, a power spectrum value calculating step for calculating a power spectrum value in the vicinity of the natural vibration component of the tool system, and an integrated value calculation for calculating the cutting torque integral value by integrating the cutting torque for a predetermined time. Step, and a peak value calculation step of calculating a cutting torque peak value in the predetermined time, the power spectrum value and the cutting torque integrated value and the cutting torque peak value is output as an input parameter of the fuzzy reasoning. The data storage step is performed during cutting by the cutting tool. Data relating to the state of the cutting tool, and / or data relating to the state of the work material are stored as data relating to the machining state, and the data-side input parameter calculation step is the state of the cutting tool at a predetermined machining time. 6. The machining state detecting method according to claim 4, wherein data relating to the state of the work material and / or data relating to the state of the work material are output as an input parameter of the fuzzy inference. 7. A sensor-side input parameter calculation means for calculating an input parameter of fuzzy inference based on a detection signal outputted from an information detection means for detecting information on a processing state during cutting by a cutting tool by a computer. And a data storage unit for pre-storing data relating to a processing state during cutting by the cutting tool, and a data-side input parameter calculation for calculating an input parameter of fuzzy inference based on the data stored in the data storage unit. Means, the output data of the fuzzy inference as a recurrent input, infers the processing state during cutting by the cutting tool based on the input parameter and the recurrent input,
A computer-readable recording medium having recorded thereon a program for functioning as a recurrent fuzzy inference means for outputting the inference result as the output data. 8. The sensor-side input parameter calculating means and the data-side input parameter calculating means input the input parameters to the recurrent fuzzy inference means in synchronization with each other. The computer-readable recording medium as described above. 9. The information detecting means detects a cutting torque during cutting by the cutting tool and outputs it as a torque signal, and the sensor-side input parameter calculating means outputs the information from the information detecting means. A power spectrum value calculating means for calculating a power spectrum value near the natural vibration component of the tool system based on the torque signal; and an integral value calculating means for calculating the cutting torque integral value by integrating the cutting torque in a predetermined time period, And, comprising a peak value calculating means for calculating the cutting torque peak value in the predetermined time, the power spectrum value and the cutting torque integrated value and the cutting torque peak value is output as an input parameter of the fuzzy inference, The data storage means stores data relating to a processing state during cutting by the cutting tool, Data about the state of the cutting tool, and / or stores data relating to the state of the workpiece, the data side input parameter calculating means, data relating to the state of the cutting tool in a predetermined processing time, and /
Alternatively, the computer-readable recording medium according to claim 7 or 8, wherein data regarding the state of the work material is output as an input parameter of the fuzzy inference. 10. A sensor-side input parameter calculating means for calculating an input parameter of fuzzy inference based on a detection signal outputted from an information detecting means for detecting information on a processing state during cutting by a cutting tool by a computer. And a data storage unit for pre-storing data relating to a processing state during cutting by the cutting tool, and a data-side input parameter calculation for calculating an input parameter of fuzzy inference based on the data stored in the data storage unit. Means, the output data of the fuzzy inference as a recurrent input, infers the processing state during cutting by the cutting tool based on the input parameter and the recurrent input,
A program which functions as recurrent fuzzy inference means for outputting the inference result as the output data. 11. The sensor-side input parameter calculating means and the data-side input parameter calculating means input the input parameters to the recurrent fuzzy inference means in synchronization with each other. The listed program. 12. The information detecting means detects a cutting torque during cutting by the cutting tool and outputs it as a torque signal, and the sensor side input parameter calculating means outputs the information from the information detecting means. A power spectrum value calculating means for calculating a power spectrum value near the natural vibration component of the tool system based on the torque signal; and an integral value calculating means for calculating the cutting torque integral value by integrating the cutting torque in a predetermined time period, And, comprising a peak value calculating means for calculating the cutting torque peak value in the predetermined time, the power spectrum value and the cutting torque integrated value and the cutting torque peak value is output as an input parameter of the fuzzy inference, The data storage means, as the data related to the processing state during cutting by the cutting tool, Data about the state of the serial cutting tools, and / or stores data relating to the state of the workpiece, the data side input parameter calculating means, data relating to the state of the cutting tool in a predetermined processing time, and /
Alternatively, the program according to claim 10 or 11, wherein data regarding a state of the work material is output as an input parameter of the fuzzy inference.