MACHINE LEARNING DEVICE, LASER DEVICE, AND MACHINE LEARNING METHOD

PROBLEM TO BE SOLVED: To learn LDU drive condition data including a prediction result of a residual life of LDU.SOLUTION: A machine learning device (6) learning LD unit drive condition data, includes: a light output detection part (7) of a laser beam; a light output characteristic recording part (8) that records a hysteresis of a driving current and a measurement result of a light output characteristic; a state quantity observation part (10) that observes a state quantity of the laser device including output data from a drive condition and state quantity recording part (9) that records the hysteresis of the LD unit drive condition data and the state quantity of the laser device; an operation result acquisition part (11) that acquires a prediction result of the LD unit and the measurement result of light output character of each of the LD unit; a learning part (12) that learns the LD unit drive condition data while relating it to the results of the state quantity of the laser device and the LD unit drive condition data; and a decision determination part (13) that determines the LD unit drive condition data including the prediction result of the residual life while referring to a leaning result of the learning part.SELECTED DRAWING: Figure 1 【課題】ＬＤＵの残存寿命の予測結果を含むＬＤＵ駆動条件データを学習する。【解決手段】ＬＤユニット駆動条件データを学習する機械学習装置（６）は、レーザ光の光出力検出部（７）と、駆動電流と光出力特性の測定結果との履歴を記録する光出力特性記録部（８）と、ＬＤユニット駆動条件データの履歴と前記レーザ装置の状態量の履歴を記録する駆動条件・状態量記録部（９）からの出力データを含む前記レーザ装置の前記状態量を観測する状態量観測部（１０）と、ＬＤユニットの特性の予測結果と各ＬＤユニットの光出力特性の測定結果を取得する動作結果取得部（１１）と、ＬＤユニット駆動条件データを、前記レーザ装置の前記状態量および前記ＬＤユニット駆動条件データの結果に関連付けて学習する学習部（１２）と、学習部の学習結果を参照して、残存寿命の前記予測結果を含む前記ＬＤユニット駆動条件データを決定する意思決定部（１３）とを含む。【選択図】図１