Tool wear monitoring and predicting method

Patent metadata

Field	Value
Publication number	US-10695884-B2
Application number	US-201815933380-A
Country	US
Kind code	B2
Filing date	Mar 23, 2018
Priority date	Mar 24, 2017
Publication date	Jun 30, 2020
Grant date	Jun 30, 2020

Abstract

A tool wear monitoring and predicting method is provided, and uses a hybrid dynamic neural network (HDNN) to build a tool wear prediction model. The tool wear prediction model adopts actual machining (cutting) conductions, sensing data detected at the current tool run of operation and the predicted tool wear value at the previous tool run of operation to predict a predicted tool wear value at the current tool run. A cyber physical agent (CPA) is adopted for simultaneously monitoring and predicting tool wear values of plural machines of the same machine type.

First claim

What is claimed is:

1. A tool wear monitoring and predicting method, comprising:
obtaining ranges of a plurality of sets of factory machining conditions regarding a cutting tool product, wherein the ranges of the sets of factory machining conditions have a plurality of boundary conditions enclosing an area of feasible machining conditions;
respectively performing a plurality of life-determining operations on a plurality of first cutting tools in accordance with the boundary conditions, thereby obtaining a plurality of actual maximum tool life values of the cutting tool product that are operated at the boundary conditions respectively, wherein the first cutting tools have the same type as the cutting tool product, and in each of the life-determining operations, one of the first cutting tools is continuously operated from its brand new condition until it is completely inoperable;
obtaining a maximum tool life (MTL) of a second cutting tool under a set of actual machining conditions in accordance with a Taylor's tool life equation by using the actual maximum tool life values and the boundary conditions, wherein the second cutting tool has the same type as the cutting tool product;
sequentially performing a plurality of historical tool runs of operation using the second cutting tool under the set of actual machining conditions, thereby obtaining a relationship of actual tool wear to tool life, a plurality of sets of historical sensing data and a plurality of historical tool wear values, wherein the historical tool wear values are corresponding to the sets of historical sensing data and the historical runs of operation in a one-to-one manner;
building a tool wear prediction model in accordance with a hybrid dynamic neural network (HDNN) algorithm by using the set of historical sensing data and the historical tool wear values;
obtaining a plurality of sets of sensing data of a third cutting tool that is sequentially performing a plurality of tool runs of operation under the set of actual machining conditions, wherein the third cutting tool has the same type as the cutting tool product, and the tool runs of operation are corresponding to the sets of sensing data in a one-to-one manner;
inputting the sets of sensing data and the maximum tool life into the tool wear prediction model, thereby obtaining a tool wear predicted value of the third cutting tool after each of the tool runs of operation, wherein, when the tool wear predicted value of the third cutting tool after the each of the tool runs of operation is desired to be predicted, the tool wear predicted value of the third cutting tool at the tool run of operation right before the each of the tool runs of operation is required to be inputted into the tool wear prediction model;
calculating a remaining useful life (RUL) of the third cutting tool by using the tool wear predicted value, the maximum tool life and the maximum tool wear threshold; and
replacing the third cutting tool with a new cutting tool when the tool wear predicted value is greater than or equal to a maximum tool wear threshold.

2. The tool wear monitoring and predicting method of claim 1 , further comprising:
obtaining a tool life of the third cutting tool from the tool wear predicted value in accordance with the relationship of actual tool wear to tool life.

3. The tool wear monitoring and predicting method of claim 1 , further comprising:
obtaining the maximum tool wear threshold from the maximum tool life in accordance with the relationship of actual tool wear to tool life.

4. The tool wear monitoring and predicting method of claim 1 , wherein the historical tool runs of operation are the same as the tool runs of operation.

5. The tool wear monitoring and predicting method of claim 1 , wherein the HDNN algorithm comprises a logistic regression (LR) algorithm and a dynamic neural network (DNN) algorithm.

6. The tool wear monitoring and predicting method of claim 1 , further comprising:
storing the sets of factory machining conditions, the actual maximum tool life values, the relationship of actual tool wear to tool life, the sets of historical sensing data and the historical tool wear values into a database on a cloud layer;
performing an operation of building the tool wear prediction model by using a cloud sever connected to the database, wherein the cloud server is located on the cloud layer;
downloading the tool wear prediction model into a cyber-physical agent (CPA) from the cloud server, wherein the cyber-physical agent is located on a factory layer, and the cyber-physical agent is communicatively connected to the cloud server through a networking layer; and
obtaining and inputting the sets of sensing data into the tool wear prediction model from a tool machine mounted with the third cutting tool by using the cyber-physical agent, thereby obtaining the tool wear predicted value of the third cutting tool after each of the tool runs of operation.

7. The tool wear monitoring and predicting method of claim 1 , further comprising:
filtering and converting the sets of historical sensing data and the set of sensing data into data corresponding to at least one feature type.

8. The tool wear monitoring and predicting method of claim 7 , wherein the at least one feature type comprises a time domain, a frequency domain and/or a time-frequency domain.

9. The tool wear monitoring and predicting method of claim 7 , wherein operations of filtering and converting the sets of historical sensing data and the set of sensing data are performed using a wavelet de-noising method and fast Fourier transform (FFT) or discrete wavelet transform (DWT).

Assignees

• Univ Nat Cheng Kung

Inventors

• Yang Haw-Ching
• Cheng Fan-Tien

Classifications

• H04L61/5007

  Internet protocol [IP] addresses · CPC title

• Y02P90/80

  Management or planning · CPC title

• B23Q17/0971 Primary

  by measuring mechanical vibrations of parts of the machine (arrangements for measuring vibrations B23Q17/12) · CPC title

• G05B23/0213 Primary

  Modular or universal configuration of the monitoring system, e.g. monitoring system having modules that may be combined to build monitoring program; monitoring system that can be applied to legacy systems; adaptable monitoring system; using different communication protocols · CPC title

• G01N3/58

  Investigating machinability by cutting tools; Investigating the cutting ability of tools · CPC title