\u5468\u4e09\u591c\u73ed，\u673a\u52a0\u8f66\u95f4\u3002<\/p>\n
"\u8fd9\u6279\u4ef6\u5e72\u5230 180 \u4ef6，\u5200\u53c8\u5d29\u4e86，"\u5200\u5177\u7ba1\u7406\u5458\u8001\u5468\u628a\u51e0\u6839\u786c\u8d28\u5408\u91d1\u5200\u6446\u684c\u4e0a，"\u7cfb\u7edf\u91cc\u8bb0\u7684\u662f'\u6362\u5200\u4e8b\u4ef6'，\u53ef\u6211\u60f3\u77e5\u9053：\u540c\u724c\u53f7\u7684\u5200，\u5e72\u8fd9\u4e2a\u6750\u6599\u5230\u5e95\u80fd\u6491\u591a\u5c11\u4ef6？\u662f 150 \u4ef6\u5c31\u5230\u5934，\u8fd8\u662f\u80fd\u5230 220？\u73b0\u5728\u5168\u9760\u8001\u5e08\u5085\u62cd\u8111\u888b，\u6362\u65e9\u4e86\u6d6a\u8d39\u5200，\u6362\u665a\u4e86\u51fa\u5e9f\u54c1\u3002"<\/p>\n
\u6211\u6253\u5f00 MES \u7684\u5bfc\u51fa\u8868\u3002<\/p>\n
"\u8fd9\u8868\u91cc\u6709\u4ec0\u4e48？"\u8001\u5468\u95ee\u3002<\/p>\n
"\u6bcf\u628a\u5200\u4e00\u4e2a\u7f16\u53f7，\u52a0\u5de5\u96f6\u4ef6\u65f6\u8bb0\u5f00\u59cb\u4ef6\u53f7\u3001\u7ed3\u675f\u4ef6\u53f7\u3001\u52a0\u5de5\u6750\u6599\u3001\u5207\u524a\u53c2\u6570\u3001\u662f\u5426\u5d29\u5203，"\u6211\u6307\u7740 CSV，"\u4e00\u628a\u5200\u4e00\u884c，\u6216\u8005\u6309\u5bff\u547d\u4e8b\u4ef6\u4e00\u884c\u3002\u53ef\u7cfb\u7edf\u4e0d\u805a\u5408'\u540c\u578b\u53f7\u5200\u5e73\u5747\u80fd\u52a0\u5de5\u51e0\u4ef6'，\u4e5f\u4e0d\u544a\u8bc9\u4f60\u662f\u540e\u5200\u9762\u78e8\u635f\u5148\u5230，\u8fd8\u662f\u5d29\u5203\u5148\u5230\u3002"<\/p>\n
"\u6211\u5c31\u60f3\u95ee\u4e00\u53e5，"\u8001\u5468\u8bf4，"WC-Co \u7ec6\u6676\u7684\u3001\u8d85\u7ec6\u6676\u7684\u3001\u6d82\u5c42 PVD \u7684，\u5404\u7edf\u8ba1\u4e00\u4e0b\u5e73\u5747\u5bff\u547d\u4ef6\u6570，\u753b\u4e2a\u5206\u5e03，\u518d\u544a\u8bc9\u6211\u54ea\u6279\u5200\u5bff\u547d\u79bb\u6563\u5927，\u54ea\u6279\u7a33\u3002\u6700\u597d\u80fd\u6309\u52a0\u5de5\u6750\u6599\u5206\u5f00\u770b，\u94a2\u4ef6\u548c\u4e0d\u9508\u94a2\u4e0d\u80fd\u6df7\u7740\u7b97\u3002"<\/p>\n
"\u6240\u4ee5\u4f60\u8981\u7684\u4e0d\u662f\u6362\u5200\u8bb0\u5f55\u8868，\u662f'\u540c\u578b\u53f7\u5200\u5177\u8010\u7528\u5ea6\u7edf\u8ba1 + \u5bff\u547d\u5206\u5e03 + \u79bb\u6563\u5ea6\u8bc4\u4f30 + \u5f02\u5e38\u5200\u8bc6\u522b'？"<\/p>\n
"\u5bf9，"\u8001\u5468\u70b9\u5934，"\u6bd4\u5982\u540c\u662f PVD \u6d82\u5c42\u5200\u5e72 45 \u94a2，\u5e73\u5747 195 \u4ef6，\u4f46\u6709\u4e00\u628a\u53ea\u5e72\u4e86 80 \u4ef6\u5c31\u5d29\u4e86，\u6211\u60f3\u628a\u5b83\u6311\u51fa\u6765，\u770b\u662f\u4e0d\u662f\u90a3\u53f0\u673a\u5e8a\u4e3b\u8f74\u8df3\u52a8\u4e86\u3002"<\/p>\n
"\u660e\u767d\u4e86，"\u6211\u5f00 VS Code，"\u7528 pandas \u8bfb\u5200\u5177\u52a0\u5de5\u8bb0\u5f55，numpy \u7b97\u5bff\u547d\u4ef6\u6570，scipy \u505a\u5206\u5e03\u68c0\u9a8c\u548c\u7f6e\u4fe1\u533a\u95f4，matplotlib \u753b\u5bff\u547d\u7bb1\u7ebf\u56fe+\u5a01\u5e03\u5c14\u56fe，scikit-learn \u805a\u7c7b\u627e\u5f02\u5e38\u5200，networkx \u628a'\u5200\u5177\u578b\u53f7\u2192\u52a0\u5de5\u6750\u6599\u2192\u673a\u5e8a'\u5efa\u5173\u7cfb\u7f51\u770b\u74f6\u9888\u3002\u6570\u636e\u81ea\u5305\u542b\u2014\u2014\u5408\u6210 5 \u7c7b\u786c\u8d28\u5408\u91d1\u5200\u00d7\u591a\u628a，\u6a21\u62df\u6b63\u5e38\u78e8\u635f\u4e0e\u5c11\u91cf\u65e9\u635f，\u4e0b\u8f7d\u5c31\u80fd\u8dd1\u3002"<\/p>\n
\u6211\u6572\u4e86\u6bb5\u539f\u578b：<\/p>\n
import pandas as pd<\/p>\n
life = df.groupby("tool_id").apply(<\/p>\n
\u00a0 \u00a0 lambda g: g["end_part"].max() – g["start_part"].min() + 1)<\/p>\n
\u00a0<\/p>\n
"\u5b8c\u6574\u7248\u7528 OOP \u5c01\u597d，"\u6211\u8bf4，"\u4e00\u4e2a\u7c7b\u7ba1\u5200\u5177\u8bb0\u5f55\u52a0\u8f7d，\u4e00\u4e2a\u7c7b\u7b97\u5355\u5200\u5bff\u547d，\u4e00\u4e2a\u7c7b\u505a\u578b\u53f7\u7ea7\u7edf\u8ba1，\u4e00\u4e2a\u7c7b\u505a\u5a01\u5e03\u5c14\u5206\u6790，\u4e00\u4e2a\u7c7b\u627e\u5f02\u5e38\u5200，\u4e00\u4e2a\u7c7b\u753b\u5173\u7cfb\u7f51，\u4e00\u4e2a\u7c7b\u51fa\u56fe\u3002\u8f93\u51fa\u6bcf\u79cd\u5200\u5e73\u5747\u4ef6\u6570\u3001\u5bff\u547d\u5206\u5e03\u3001\u65e9\u635f\u6e05\u5355\u3001\u5a01\u5e03\u5c14\u659c\u7387\u3002"<\/p>\n
\u8001\u5468\u51d1\u8fd1\u770b："\u90a3\u6211\u4ee5\u540e\u770b\u62a5\u544a：PVD \u6d82\u5c42\u5e72 45 \u94a2，\u5747\u503c 195 \u4ef6，\u5a01\u5e03\u5c14\u659c\u7387 8.2，\u8bf4\u660e\u5bff\u547d\u633a\u96c6\u4e2d；\u6709\u4e00\u628a 82 \u4ef6\u5d29\u5203，\u6807\u7ea2，\u5173\u8054\u673a\u5e8a M3，\u5c31\u77e5\u9053\u53bb\u67e5 M3 \u4e3b\u8f74\u4e86\u3002"<\/p>\n
"\u5bf9，"\u6211\u63a5\u8bdd，"\u5200\u5177\u8010\u7528\u5ea6\u4e0d\u662f\u6807\u79f0\u503c，\u662f\u7edf\u8ba1\u51fa\u6765\u7684\u5206\u5e03\u3002\u6570\u5b57\u5b6a\u751f\u91cc\u5200\u662f\u6d88\u8017\u54c1\u6a21\u578b，\u5f97\u5148\u6709\u8fd9\u5f20\u7edf\u8ba1\u5e95\u8868，\u624d\u80fd\u505a\u6362\u5200\u9884\u6d4b\u3002"<\/p>\n
\u00a0<\/p>\n
\u4e00\u3001\u5b9e\u9645\u5e94\u7528\u573a\u666f（\u771f\u5b9e\u75db\u70b9）<\/p>\n
\u00a0<\/p>\n
\u573a\u666f\u8bbe\u5b9a：\u591a\u578b\u53f7\u786c\u8d28\u5408\u91d1\u5200\u5177（\u4e0d\u540c\u724c\u53f7\/\u6d82\u5c42）\u52a0\u5de5\u4e0d\u540c\u6750\u6599\u96f6\u4ef6，MES \u8bb0\u5f55\u6bcf\u628a\u5200\u7684\u542f\u7528\u4ef6\u53f7\u3001\u505c\u7528\u4ef6\u53f7\u3001\u52a0\u5de5\u6750\u6599\u3001\u673a\u5e8a\u53f7\u3001\u505c\u7528\u539f\u56e0（\u6b63\u5e38\u78e8\u635f\/\u5d29\u5203\/\u5d29\u89d2\/\u8fbe\u5230\u5bff\u547d）\u3002\u73b0\u573a\u53ea\u8bb0\u5f55\u4e8b\u4ef6，\u4e0d\u63d0\u4f9b\u540c\u578b\u53f7\u5200\u5177\u5bff\u547d\u805a\u5408\u7edf\u8ba1，\u4e5f\u4e0d\u533a\u5206\u6750\u6599\u5de5\u51b5，\u5bfc\u81f4\u6362\u5200\u7b56\u7565\u5168\u51ed\u7ecf\u9a8c\u3002<\/p>\n
\u00a0<\/p>\n
\u73b0\u573a\u539f\u8bdd（\u53d9\u4e8b\u5316）：<\/p>\n
\u00a0<\/p>\n
"\u6211\u4e0d\u662f\u820d\u4e0d\u5f97\u6362\u5200，"\u8001\u5468\u8bf4，"\u662f\u6ca1\u6570\u3002\u540c\u662f PVD \u6d82\u5c42\u5200，\u6709\u7684\u5e72 210 \u4ef6，\u6709\u7684 160 \u4ef6\u5c31\u5d29，\u7cfb\u7edf\u91cc\u5c31\u662f\u4e24\u6761\u6362\u5200\u8bb0\u5f55，\u6ca1\u4eba\u5e2e\u6211\u7b97\u5747\u503c\u3002\u6709\u6b21\u6211\u6309 150 \u4ef6\u6362，\u7ed3\u679c\u540e\u9762\u5200\u90fd\u8fd8\u80fd\u7528，\u4e00\u5e74\u767d\u6254\u51e0\u767e\u6839；\u53e6\u4e00\u6b21\u6309 220 \u4ef6\u6362，\u51fa\u4e86\u6279\u632f\u7eb9\u5e9f\u54c1\u3002\u540e\u6765\u6211\u628a\u6bcf\u628a\u5200\u5bff\u547d\u624b\u8bb0\u5728\u672c\u5b50\u4e0a，\u53ef\u672c\u5b50\u8ddf MES \u5bf9\u4e0d\u4e0a，\u4e5f\u5206\u4e0d\u4e86\u6750\u6599\u3002"<\/p>\n
\u6838\u5fc3\u77db\u76fe："\u6362\u5200\u4e8b\u4ef6\u6d41\u6c34"\u4e0e"\u540c\u578b\u53f7\u5200\u5177\u8010\u7528\u5ea6\u7edf\u8ba1 + \u5bff\u547d\u5206\u5e03 + \u5a01\u5e03\u5c14\u8bc4\u4f30 + \u65e9\u635f\u5f02\u5e38\u8bc6\u522b + \u5de5\u51b5\u5f52\u56e0"\u4e4b\u95f4\u7684\u65ad\u5c42\u3002\u9700\u8981\u4e00\u4e2a"\u786c\u8d28\u5408\u91d1\u5200\u5177\u8010\u7528\u5ea6\u7edf\u8ba1\u4e0e\u8bc4\u4f30\u7a0b\u5e8f"，\u7528\u00a0<\/p>\n
"pandas" \u805a\u5408\u3001<\/p>\n
"numpy" \u7b97\u5bff\u547d\u3001<\/p>\n
"scipy" \u505a\u5206\u5e03\u62df\u5408\u4e0e\u68c0\u9a8c\u3001<\/p>\n
"matplotlib" \u753b\u7bb1\u7ebf\u56fe\/\u5a01\u5e03\u5c14\u56fe\u3001<\/p>\n
"scikit-learn" \u805a\u7c7b\u5f02\u5e38\u5200\u3001<\/p>\n
"networkx" \u5efa\u5de5\u51b5\u5173\u7cfb\u7f51\u3002<\/p>\n
\u00a0<\/p>\n
\u4e8c\u3001\u75db\u70b9\u5206\u6790（\u6620\u5c04\u5230\u6ee8\u5dde\u804c\u4e1a\u5b66\u9662\u300a\u5148\u8fdb\u5236\u9020\u6280\u672f\u300b\u8bfe\u7a0b\u6a21\u578b）<\/p>\n
\u00a0<\/p>\n
\u300a\u5148\u8fdb\u5236\u9020\u6280\u672f\u300b\u6a21\u5757 \u672c\u7bc7\u75db\u70b9\u5bf9\u5e94<\/p>\n
\u6570\u63a7\u52a0\u5de5\u4e0eCAD\/CAM\u6280\u672f：\u5200\u5177\u6750\u6599\u4e0e\u5207\u524a\u53c2\u6570\u4f18\u5316 \u786c\u8d28\u5408\u91d1\u724c\u53f7\/\u6d82\u5c42\u8010\u7528\u5ea6\u91cf\u5316\u7edf\u8ba1<\/p>\n
\u5148\u8fdb\u5236\u9020\u6280\u672f\u57fa\u7840：\u5236\u9020\u8fc7\u7a0b\u8d44\u6e90\u6d88\u8017\u7ba1\u7406 \u5200\u5177\u4f5c\u4e3a\u5173\u952e\u6d88\u8017\u8d44\u6e90\u5bff\u547d\u5efa\u6a21<\/p>\n
FMS\u4e0e\u5148\u8fdb\u751f\u4ea7\u7ba1\u7406：\u5355\u5143\u7ea7\u5200\u5177\u7ba1\u7406\u4e0e\u6362\u5200\u7b56\u7565 \u6309\u578b\u53f7\u805a\u5408\u5bff\u547d，\u652f\u6491\u6362\u5200\u9608\u503c\u8bbe\u5b9a<\/p>\n
\u667a\u80fd\u5236\u9020\u4e0e\u6570\u5b57\u5b6a\u751f：\u5200\u5177\u6570\u5b57\u5b6a\u751f \u5bff\u547d\u5206\u5e03\u2192\u53ef\u9884\u6d4b\u6362\u5200\u6a21\u578b\u5e95\u5ea7<\/p>\n
\u5148\u8fdb\u5236\u9020\u65b0\u6a21\u5f0f：\u6210\u672c\u4e0e\u8d28\u91cf\u534f\u540c（\u96f6\u7f3a\u9677） \u65e9\u635f\u8bc6\u522b\u907f\u514d\u6279\u91cf\u632f\u7eb9\u5e9f\u54c1<\/p>\n
\u00a0<\/p>\n
\u4e00\u53e5\u8bdd\u603b\u7ed3：\u6211\u4eec\u9700\u8981\u6784\u5efa\u4e00\u4e2a"\u786c\u8d28\u5408\u91d1\u5200\u5177\u8010\u7528\u5ea6\u7edf\u8ba1\u4e0e\u8bc4\u4f30\u7a0b\u5e8f"，\u7528\u00a0<\/p>\n
"pandas" \u6309\u578b\u53f7\u805a\u5408，<\/p>\n
"scipy" \u505a\u5a01\u5e03\u5c14\u62df\u5408，<\/p>\n
"scikit-learn" \u627e\u65e9\u635f\u5f02\u5e38，<\/p>\n
"networkx" \u5173\u8054\u5de5\u51b5，\u5b9e\u73b0\u4ece"\u6362\u5200\u6d41\u6c34\u8d26"\u5230"\u53ef\u7edf\u8ba1\u3001\u53ef\u9884\u6d4b\u3001\u53ef\u5f52\u56e0\u7684\u5200\u5177\u5bff\u547d\u6a21\u578b"\u3002<\/p>\n
\u00a0<\/p>\n
\u4e09\u3001\u6838\u5fc3\u903b\u8f91\u8bb2\u89e3（\u5927\u767d\u8bdd）<\/p>\n
\u00a0<\/p>\n
3.1 \u95ee\u9898\u672c\u8d28：\u628a\u5200\u5177\u60f3\u6210"\u706f\u6ce1\u5bff\u547d"<\/p>\n
\u00a0<\/p>\n
\u628a\u6bcf\u628a\u5200\u60f3\u6210\u4e00\u4e2a\u706f\u6ce1：<\/p>\n
\u00a0<\/p>\n
* \u5200\u5177\u578b\u53f7 = \u706f\u6ce1\u54c1\u724c（\u7ec6\u6676 WC-Co \/ \u8d85\u7ec6\u6676 \/ PVD\u6d82\u5c42\u2026）<\/p>\n
* \u52a0\u5de5\u4ef6\u6570 = \u4eae\u4e86\u591a\u5c11\u5c0f\u65f6<\/p>\n
* \u505c\u7528\u539f\u56e0 = \u600e\u4e48\u706d\u7684（\u81ea\u7136\u6697\u4e0b\u53bb=\u6b63\u5e38\u78e8\u635f，\u70b8\u4e86=\u5d29\u5203）<\/p>\n
* \u540c\u578b\u53f7\u591a\u628a\u5200 = \u540c\u4e00\u54c1\u724c\u4e00\u5806\u706f\u6ce1，\u6709\u7684\u4eae\u4e45\u70b9\u6709\u7684\u77ed\u70b9<\/p>\n
* \u5e73\u5747\u4ef6\u6570 = \u54c1\u724c\u6807\u79f0\u5bff\u547d<\/p>\n
* \u5a01\u5e03\u5c14\u659c\u7387 = \u8fd9\u6279\u706f\u6ce1"\u9f50\u4e0d\u9f50\u5fc3"：\u659c\u7387\u5927（>5）\u8bf4\u660e\u5bff\u547d\u5f88\u96c6\u4e2d，\u659c\u7387\u5c0f\u8bf4\u660e\u5ffd\u957f\u5ffd\u77ed<\/p>\n
* \u65e9\u635f\u5200 = \u540c\u6279\u91cc\u65e9\u65e9\u70b8\u6389\u7684\u90a3\u4e2a\u706f\u6ce1，\u5f97\u62ce\u51fa\u6765\u770b\u662f\u4e0d\u662f\u88c5\u9519\u673a\u5e8a\u4e86<\/p>\n
* \u5de5\u51b5\u5173\u7cfb\u7f51 = \u54ea\u79cd\u5200\u5728\u54ea\u79cd\u6750\u6599\u54ea\u53f0\u673a\u5e8a\u4e0a\u6700\u8d39<\/p>\n
\u00a0<\/p>\n
\u5de5\u4e1a\u5e94\u7528：<\/p>\n
\u00a0<\/p>\n
*\u00a0<\/p>\n
"pandas.groupby(["tool_grade","material"])" \u805a\u5408\u5bff\u547d\u4ef6\u6570<\/p>\n
*\u00a0<\/p>\n
"scipy.stats.weibull_min" \u62df\u5408\u5bff\u547d\u5206\u5e03，\u53d6\u5f62\u72b6\u53c2\u6570 k<\/p>\n
*\u00a0<\/p>\n
"scikit-learn" \u7528\u5bff\u547d+\u659c\u7387\u504f\u5dee\u505a\u65e9\u635f\u805a\u7c7b<\/p>\n
*\u00a0<\/p>\n
"networkx" \u5efa\u00a0<\/p>\n
"\u5200\u5177\u578b\u53f7\u2014\u6750\u6599\u2014\u673a\u5e8a" \u4e09\u5c42\u56fe，\u627e\u9ad8\u6d88\u8017\u8def\u5f84<\/p>\n
\u00a0<\/p>\n
3.2 \u4e1a\u52a1\u903b\u8f91 \u2192 \u4ee3\u7801\u6620\u5c04<\/p>\n
\u00a0<\/p>\n
\u5b9a\u4e49\u5200\u5177\u52a0\u5de5\u8bb0\u5f55\u6a21\u578b<\/p>\n
\u00a0 \u00a0\u2502<\/p>\n
\u00a0 \u00a0\u25bc ToolLogLoader (pandas)<\/p>\n
\u5bfc\u5165 CSV：<\/p>\n
\u00a0 tool_id, tool_grade, coating, material, machine,<\/p>\n
\u00a0 start_part, end_part, stop_reason<\/p>\n
\u00a0 \u6821\u9a8c\u4ef6\u53f7\u8fde\u7eed\u3001\u53bb\u91cd<\/p>\n
\u00a0 \u00a0\u2502<\/p>\n
\u00a0 \u00a0\u25bc ToolLifeCalculator (numpy\/pandas)<\/p>\n
\u5355\u5200\u5bff\u547d：<\/p>\n
\u00a0 life_parts = end_part – start_part + 1<\/p>\n
\u00a0 \u6309\u5200\u805a\u5408，\u6807\u8bb0\u65e9\u635f<\/p>\n
\u00a0 \u00a0\u2502<\/p>\n
\u00a0 \u00a0\u25bc GradeStatistics (pandas\/numpy)<\/p>\n
\u578b\u53f7\u7ea7\u7edf\u8ba1：<\/p>\n
\u00a0 \u6309 (\u724c\u53f7,\u6d82\u5c42,\u6750\u6599) \u5206\u7ec4<\/p>\n
\u00a0 \u5747\u503c\/\u4e2d\u4f4d\u6570\/\u6807\u51c6\u5dee\/P10\/P90\/\u7f6e\u4fe1\u533a\u95f4<\/p>\n
\u00a0 \u00a0\u2502<\/p>\n
\u00a0 \u00a0\u25bc WeibullAnalyzer (scipy)<\/p>\n
\u5bff\u547d\u5206\u5e03：<\/p>\n
\u00a0 weibull_min.fit -> \u5f62\u72b6k, \u5c3a\u5ea6\u03b7<\/p>\n
\u00a0 k>5 \u96c6\u4e2d, k<2 \u79bb\u6563\u5927<\/p>\n
\u00a0 \u8ba1\u7b97 B10 \u5bff\u547d(10%\u5931\u6548\u4ef6\u6570)<\/p>\n
\u00a0 \u00a0\u2502<\/p>\n
\u00a0 \u00a0\u25bc AnomalyDetector (scikit-learn)<\/p>\n
\u65e9\u635f\u8bc6\u522b：<\/p>\n
\u00a0 \u7279\u5f81=[\u5bff\u547d, \u504f\u79bb\u540c\u7ec4\u5747\u503c\u6bd4\u4f8b]<\/p>\n
\u00a0 IsolationForest \/ KMeans \u627e\u5f02\u5e38\u5200<\/p>\n
\u00a0 \u00a0\u2502<\/p>\n
\u00a0 \u00a0\u25bc ToolConditionGraph (networkx)<\/p>\n
\u5de5\u51b5\u5173\u7cfb\u7f51：<\/p>\n
\u00a0 \u5200\u5177\u578b\u53f7-\u6750\u6599-\u673a\u5e8a \u4e09\u5c42\u56fe<\/p>\n
\u00a0 \u8fb9\u6743=\u5e73\u5747\u6d88\u8017\u4ef6\u6570\u5012\u6570(\u8d8a\u8d39\u6743\u91cd\u8d8a\u5927)<\/p>\n
\u00a0 \u00a0\u2502<\/p>\n
\u00a0 \u00a0\u25bc ToolVisualizer (matplotlib)<\/p>\n
\u53ef\u89c6\u5316：<\/p>\n
\u00a0 1. \u5404\u578b\u53f7\u5bff\u547d\u7bb1\u7ebf\u56fe<\/p>\n
\u00a0 2. \u5a01\u5e03\u5c14\u6982\u7387\u56fe<\/p>\n
\u00a0 3. \u65e9\u635f\u5200\u6563\u70b9\u6807\u8bb0<\/p>\n
\u00a0 4. \u5de5\u51b5\u5173\u7cfb\u7f51\u56fe<\/p>\n
\u00a0 5. \u5bff\u547d\u76f4\u65b9\u56fe+\u6807\u79f0\u7ebf<\/p>\n
\u00a0 \u00a0\u2502<\/p>\n
\u00a0 \u00a0\u25bc SyntheticToolGenerator (numpy)<\/p>\n
\u5408\u6210\u6570\u636e：<\/p>\n
\u00a0 5\u7c7b\u786c\u8d28\u5408\u91d1\u5200，\u6bcf\u7c7b\u591a\u628a<\/p>\n
\u00a0 \u5e7245\u94a2\/\u4e0d\u9508\u94a2\/\u94f8\u94c1，\u542b\u6b63\u5e38\u78e8\u635f+\u5c11\u91cf\u65e9\u635f<\/p>\n
\u00a0<\/p>\n
3.3 \u4e3a\u4ec0\u4e48\u7528\u5a01\u5e03\u5c14\u800c\u4e0d\u662f\u6b63\u6001\u5206\u5e03？<\/p>\n
\u00a0<\/p>\n
* \u95ee\u9898：\u5bff\u547d\u6570\u636e\u6709\u4e0b\u9650 0\u3001\u53f3\u504f，\u6b63\u6001\u4f1a\u7b97\u51fa\u8d1f\u5bff\u547d，\u4e14\u5bf9\u65e9\u635f\u4e0d\u654f\u611f\u3002<\/p>\n
* \u5904\u7406\u7b56\u7565：\u5a01\u5e03\u5c14\u5206\u5e03\u662f\u53ef\u9760\u6027\u5de5\u7a0b\u6807\u51c6\u6a21\u578b，\u5f62\u72b6\u53c2\u6570\u76f4\u63a5\u53cd\u6620\u5931\u6548\u6a21\u5f0f\u3002<\/p>\n
* \u5de5\u7a0b\u5408\u7406\u6027：\u8ddf\u5200\u5177\u5382\u5bff\u547d\u6837\u672c\u62a5\u544a\u53e3\u5f84\u4e00\u81f4，B10 \u5bff\u547d\u53ef\u76f4\u63a5\u7528\u4e8e\u6362\u5200\u7b56\u7565\u3002<\/p>\n
\u00a0<\/p>\n
3.4 \u5206\u6790\u524d\u540e\u5bf9\u6bd4<\/p>\n
\u00a0<\/p>\n
\u7ef4\u5ea6 MES\u6362\u5200\u8bb0\u5f55 \u672c\u7a0b\u5e8f<\/p>\n
\u540c\u578b\u53f7\u5747\u503c \u624b\u7b97 \u81ea\u52a8\u5206\u7ec4\u805a\u5408<\/p>\n
\u5bff\u547d\u5206\u5e03 \u65e0 \u5a01\u5e03\u5c14\u62df\u5408 + B10<\/p>\n
\u65e9\u635f\u8bc6\u522b \u9760\u8089\u773c \u805a\u7c7b\u81ea\u52a8\u6807\u7ea2<\/p>\n
\u5de5\u51b5\u5f52\u56e0 \u65e0 networkx \u5173\u7cfb\u7f51<\/p>\n
\u6362\u5200\u9608\u503c \u62cd\u8111\u888b P90 + B10 \u53cc\u53c2\u8003<\/p>\n
\u00a0<\/p>\n
\u56db\u3001OOP \u4ee3\u7801\u5b9e\u73b0<\/p>\n
\u00a0<\/p>\n
4.1 \u9879\u76ee\u7ed3\u6784<\/p>\n
\u00a0<\/p>\n
tool_life_eval\/<\/p>\n
\u251c\u2500\u2500 tool_life_eval\/<\/p>\n
\u2502 \u251c\u2500\u2500 __init__.py<\/p>\n
\u2502 \u251c\u2500\u2500 log_loader.py # \u5200\u5177\u8bb0\u5f55\u52a0\u8f7d<\/p>\n
\u2502 \u251c\u2500\u2500 life_calculator.py # \u5355\u5200\u5bff\u547d\u8ba1\u7b97<\/p>\n
\u2502 \u251c\u2500\u2500 grade_statistics.py # \u578b\u53f7\u7ea7\u7edf\u8ba1<\/p>\n
\u2502 \u251c\u2500\u2500 weibull_analyzer.py # \u5a01\u5e03\u5c14\u5206\u6790<\/p>\n
\u2502 \u251c\u2500\u2500 anomaly_detector.py # \u65e9\u635f\u5f02\u5e38\u8bc6\u522b<\/p>\n
\u2502 \u251c\u2500\u2500 condition_graph.py # \u5de5\u51b5\u5173\u7cfb\u7f51<\/p>\n
\u2502 \u251c\u2500\u2500 visualizer.py # \u53ef\u89c6\u5316<\/p>\n
\u2502 \u2514\u2500\u2500 synthetic_data.py # \u5408\u6210\u6570\u636e<\/p>\n
\u251c\u2500\u2500 tests\/<\/p>\n
\u2502 \u251c\u2500\u2500 __init__.py<\/p>\n
\u2502 \u2514\u2500\u2500 test_tool_life.py<\/p>\n
\u251c\u2500\u2500 results\/<\/p>\n
\u2502 \u251c\u2500\u2500 box_life_by_grade.png # \u7bb1\u7ebf\u56fe<\/p>\n
\u2502 \u251c\u2500\u2500 weibull_probplot.png # \u5a01\u5e03\u5c14\u6982\u7387\u56fe<\/p>\n
\u2502 \u251c\u2500\u2500 anomaly_scatter.png # \u65e9\u635f\u6563\u70b9<\/p>\n
\u2502 \u251c\u2500\u2500 condition_graph.png # \u5de5\u51b5\u5173\u7cfb\u7f51<\/p>\n
\u2502 \u251c\u2500\u2500 life_hist.png # \u5bff\u547d\u76f4\u65b9\u56fe<\/p>\n
\u2502 \u251c\u2500\u2500 grade_life_stats.csv # \u578b\u53f7\u7edf\u8ba1\u8868<\/p>\n
\u2502 \u251c\u2500\u2500 early_failure_list.csv # \u65e9\u635f\u6e05\u5355<\/p>\n
\u2502 \u2514\u2500\u2500 weibull_report.txt<\/p>\n
\u2514\u2500\u2500 run_tool_life.py<\/p>\n
\u00a0<\/p>\n
4.2 \u6838\u5fc3\u6e90\u7801<\/p>\n
\u00a0<\/p>\n
<details><\/p>\n
\u00a0<\/p>\n
<summary><\/summary><\/p>\n
\u00a0<\/p>\n
"""\u5200\u5177\u52a0\u5de5\u8bb0\u5f55\u52a0\u8f7d\u5668"""<\/p>\n
\u00a0<\/p>\n
import pandas as pd<\/p>\n
from pathlib import Path<\/p>\n
from typing import Optional<\/p>\n
\u00a0<\/p>\n
\u00a0<\/p>\n
class ToolLogLoader:<\/p>\n
\u00a0 \u00a0 """\u52a0\u8f7d\u5200\u5177\u52a0\u5de5\u6d41\u6c34\u8bb0\u5f55"""<\/p>\n
\u00a0<\/p>\n
\u00a0 \u00a0 STOP_REASONS = ("normal_wear", "chipping", "edge_break", "reach_life")<\/p>\n
\u00a0<\/p>\n
\u00a0 \u00a0 def __init__(self, filepath: str = "tool_logs.csv",<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0encoding: str = "utf-8"):<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 self.filepath = Path(filepath)<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 self.encoding = encoding<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 self._raw: Optional[pd.DataFrame] = None<\/p>\n
\u00a0<\/p>\n
\u00a0 \u00a0 def load(self) -> pd.DataFrame:<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 if not self.filepath.exists():<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 raise FileNotFoundError(f"\u6587\u4ef6\u4e0d\u5b58\u5728: {self.filepath}")<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 self._raw = pd.read_csv(self.filepath, encoding=self.encoding)<\/p>\n
\u00a0<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 rename = {}<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 for tgt, al in {<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 "tool_id": ["tool_id", "\u5200\u53f7", "id"],<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 "tool_grade": ["tool_grade", "\u724c\u53f7", "grade"],<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 "coating": ["coating", "\u6d82\u5c42", "coat"],<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 "material": ["material", "\u52a0\u5de5\u6750\u6599", "mat"],<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 "machine": ["machine", "\u673a\u5e8a", "mc"],<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 "start_part": ["start_part", "\u8d77\u59cb\u4ef6", "sp"],<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 "end_part": ["end_part", "\u7ed3\u675f\u4ef6", "ep"],<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 "stop_reason": ["stop_reason", "\u505c\u7528\u539f\u56e0", "reason"],<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 }.items():<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 if tgt not in self._raw.columns:<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 for a in al:<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 if a in self._raw.columns:<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 rename[a] = tgt<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 break<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 self._raw = self._raw.rename(columns=rename)<\/p>\n
\u00a0<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 req = ["tool_id", "tool_grade", "coating", "material",<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0"machine", "start_part", "end_part", "stop_reason"]<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 miss = [c for c in req if c not in self._raw.columns]<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 if miss:<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 raise ValueError(f"\u7f3a\u5c11\u5fc5\u8981\u5217: {miss}")<\/p>\n
\u00a0<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 for c in ["start_part", "end_part"]:<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 self._raw[c] = pd.to_numeric(self._raw[c], errors="coerce")<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 self._raw = self._raw.dropna(subset=["start_part", "end_part"])<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 self._raw = self._raw[(self._raw["end_part"] >= self._raw["start_part"])]<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 self._raw = self._raw.sort_values(["tool_grade", "tool_id"]).reset_index(drop=True)<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 return self._raw.copy()<\/p>\n
\u00a0<\/p>\n
<\/details><\/p>\n
\u00a0<\/p>\n
<details><\/p>\n
\u00a0<\/p>\n
<summary><\/summary><\/p>\n
\u00a0<\/p>\n
"""\u5355\u5200\u5bff\u547d\u8ba1\u7b97"""<\/p>\n
\u00a0<\/p>\n
import numpy as np<\/p>\n
import pandas as pd<\/p>\n
from typing import Optional<\/p>\n
\u00a0<\/p>\n
\u00a0<\/p>\n
class ToolLifeCalculator:<\/p>\n
\u00a0 \u00a0 """\u8ba1\u7b97\u6bcf\u628a\u5200\u53ef\u52a0\u5de5\u96f6\u4ef6\u6570"""<\/p>\n
\u00a0<\/p>\n
\u00a0 \u00a0 def __init__(self):<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 pass<\/p>\n
\u00a0<\/p>\n
\u00a0 \u00a0 def per_tool(self, df: pd.DataFrame) -> pd.DataFrame:<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 rows = []<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 for tool_id, g in df.groupby("tool_id"):<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 head = g.iloc[0]<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 # \u652f\u6301\u4e00\u628a\u5200\u5206\u6bb5\u52a0\u5de5\u65f6\u53d6\u5168\u5c40\u8d77\u6b62<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 sp = int(g["start_part"].min())<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 ep = int(g["end_part"].max())<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 life = ep – sp + 1<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 rows.append({<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 "tool_id": tool_id,<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 "tool_grade": head["tool_grade"],<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 "coating": head["coating"],<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 "material": head["material"],<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 "machine": head["machine"],<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 "start_part": sp,<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 "end_part": ep,<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 "life_parts": int(life),<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 "stop_reason": g["stop_reason"].mode().iloc[0],<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 })<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 out = pd.DataFrame(rows)<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 return out.sort_values("tool_id").reset_index(drop=True)<\/p>\n
\u00a0<\/p>\n
\u00a0 \u00a0 def tag_early_failure(self, life_df: pd.DataFrame,<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 group_keys=("tool_grade", "coating", "material"),<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 k_sigma: float = 2.0) -> pd.DataFrame:<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 """\u6309\u540c\u5de5\u51b5\u7ec4\u5747\u503c-2\u03c3\u6807\u8bb0\u65e9\u635f"""<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 df = life_df.copy()<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 stats = df.groupby(list(group_keys))["life_parts"].agg(["mean", "std"]).reset_index()<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 stats["std"] = stats["std"].fillna(0.0)<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 df = df.merge(stats, on=list(group_keys), how="left")<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 df["is_early"] = (df["life_parts"] < df["mean"] – k_sigma * df["std"]) | \\\\<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0(df["stop_reason"].isin(["chipping", "edge_break"]))<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 df["dev_from_mean"] = (df["life_parts"] – df["mean"]).round(2)<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 for c in ["mean", "std"]:<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 df[c] = df[c].round(2)<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 return df<\/p>\n
\u00a0<\/p>\n
<\/details><\/p>\n
\u00a0<\/p>\n
<details><\/p>\n
\u00a0<\/p>\n
<summary><\/summary><\/p>\n
\u00a0<\/p>\n
"""\u578b\u53f7\u7ea7\u8010\u7528\u5ea6\u7edf\u8ba1"""<\/p>\n
\u00a0<\/p>\n
import numpy as np<\/p>\n
import pandas as pd<\/p>\n
from scipy import stats<\/p>\n
from typing import Optional<\/p>\n
\u00a0<\/p>\n
\u00a0<\/p>\n
class GradeStatistics:<\/p>\n
\u00a0 \u00a0 """\u6309 (\u724c\u53f7,\u6d82\u5c42,\u6750\u6599) \u805a\u5408\u7edf\u8ba1"""<\/p>\n
\u00a0<\/p>\n
\u00a0 \u00a0 def __init__(self, conf: float = 0.95):<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 self.conf = conf<\/p>\n
\u00a0<\/p>\n
\u00a0 \u00a0 def summarize(self, life_df: pd.DataFrame) -> pd.DataFrame:<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 keys = ["tool_grade", "coating", "material"]<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 g = life_df.groupby(keys)["life_parts"]<\/p>\n
\u00a0<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 n = g.count()<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 mean = g.mean()<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 med = g.median()<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 std = g.std(ddof=1).fillna(0.0)<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 p10 = g.quantile(0.10)<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 p90 = g.quantile(0.90)<\/p>\n
\u00a0<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 # \u5747\u503c\u7f6e\u4fe1\u533a\u95f4<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 rows = []<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 for key_tuple in mean.index:<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 sub = life_df[(life_df["tool_grade"] == key_tuple[0]) &<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 (life_df["coating"] == key_tuple[1]) &<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 (life_df["material"] == key_tuple[2])]["life_parts"].values<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 if len(sub) > 1:<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 ci = stats.t.interval(self.conf, len(sub) – 1,<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 loc=sub.mean(), scale=stats.sem(sub))<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 ci_lo, ci_hi = round(ci[0], 1), round(ci[1], 1)<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 else:<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 ci_lo = ci_hi = round(float(sub.mean()), 1)<\/p>\n
\u00a0<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 rows.append({<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 "tool_grade": key_tuple[0],<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 "coating": key_tuple[1],<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 "material": key_tuple[2],<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 "n_tools": int(n[key_tuple]),<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 "mean_life": round(mean[key_tuple], 1),<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 "median_life": round(med[key_tuple], 1),<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 "std_life": round(std[key_tuple], 1),<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 "p10_life": round(p10[key_tuple], 1),<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 "p90_life": round(p90[key_tuple], 1),<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 "ci95_lo": ci_lo,<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 "ci95_hi": ci_hi,<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 "early_fail_cnt": int(((life_df[<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 (life_df["tool_grade"] == key_tuple[0]) &<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 (life_df["coating"] == key_tuple[1]) &<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 (life_df["material"] == key_tuple[2])<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 ]["is_early"] == True).sum()),<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 })<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 return pd.DataFrame(rows).sort_values("mean_life", ascending=False).reset_index(drop=True)<\/p>\n
\u00a0<\/p>\n
<\/details><\/p>\n
\u00a0<\/p>\n
<details><\/p>\n
\u00a0<\/p>\n
<summary><\/summary><\/p>\n
\u00a0<\/p>\n
"""\u5a01\u5e03\u5c14\u5bff\u547d\u5206\u6790 (scipy)"""<\/p>\n
\u00a0<\/p>\n
import numpy as np<\/p>\n
import pandas as pd<\/p>\n
from scipy import stats<\/p>\n
from typing import Dict, Optional<\/p>\n
\u00a0<\/p>\n
\u00a0<\/p>\n
class WeibullAnalyzer:<\/p>\n
\u00a0 \u00a0 """\u5bf9\u6bcf\u7ec4\u5bff\u547d\u505a\u5a01\u5e03\u5c14\u6700\u5c0f\u5206\u5e03\u62df\u5408"""<\/p>\n
\u00a0<\/p>\n
\u00a0 \u00a0 def __init__(self):<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 pass<\/p>\n
\u00a0<\/p>\n
\u00a0 \u00a0 def fit_group(self, life_values: np.ndarray) -> Dict:<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 x = np.asarray(life_values, dtype=float)<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 x = x[x > 0]<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 if len(x) < 3:<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 return {"k": np.nan, "eta": np.nan, "b10": np.nan}<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 # weibull_min(c, loc, scale), c\u5373\u5f62\u72b6k<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 c, loc, scale = stats.weibull_min.fit(x, floc=0)<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 k = float(c)<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 eta = float(scale)<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 b10 = float(eta * np.log(1 \/ 0.9) ** (1 \/ k)) if k > 0 else np.nan<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 return {"k": round(k, 3), "eta": round(eta, 1), "b10": round(b10, 1)}<\/p>\n
\u00a0<\/p>\n
\u00a0 \u00a0 def fit_all_groups(self, life_df: pd.DataFrame) -> pd.DataFrame:<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 keys = ["tool_grade", "coating", "material"]<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 rows = []<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 for key_tuple, g in life_df.groupby(keys):<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 res = self.fit_group(g["life_parts"].values)<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 rows.append({<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 "tool_grade": key_tuple[0],<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 "coating": key_tuple[1],<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 "material": key_tuple[2],<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 "weibull_k": res["k"],<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 "weibull_eta": res["eta"],<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 "b10_life": res["b10"],<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 "dispersion": "\u96c6\u4e2d" if (res["k"] and res["k"] > 5)<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 else ("\u79bb\u6563" if (res["k"] and res["k"] < 2) else "\u4e2d\u7b49"),<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 })<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 return pd.DataFrame(rows).sort_values("weibull_k", ascending=False).reset_index(drop=True)<\/p>\n
\u00a0<\/p>\n
\u00a0 \u00a0 def probplot_data(self, life_values: np.ndarray):<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 """\u8fd4\u56de\u5a01\u5e03\u5c14\u6982\u7387\u56fe\u5750\u6807"""<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 x = np.sort(np.asarray(life_values, dtype=float))<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 x = x[x > 0]<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 n = len(x)<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 if n < 2:<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 return x, np.array([])<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 f = (np.arange(1, n + 1) – 0.3) \/ (n + 0.4)<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 y = np.log(-np.log(1 – f))<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 return x, y<\/p>\n
\u00a0<\/p>\n
<\/details><\/p>\n
\u00a0<\/p>\n
<details><\/p>\n
\u00a0<\/p>\n
<summary><\/summary><\/p>\n
\u00a0<\/p>\n
"""\u65e9\u635f\u5f02\u5e38\u5200\u8bc6\u522b (scikit-learn)"""<\/p>\n
\u00a0<\/p>\n
import numpy as np<\/p>\n
import pandas as pd<\/p>\n
from sklearn.cluster import KMeans<\/p>\n
from sklearn.preprocessing import StandardScaler<\/p>\n
from typing import Optional<\/p>\n
\u00a0<\/p>\n
\u00a0<\/p>\n
class AnomalyDetector:<\/p>\n
\u00a0 \u00a0 """\u57fa\u4e8e\u540c\u5de5\u51b5\u504f\u79bb\u5ea6\u805a\u7c7b\u627e\u65e9\u635f\u5200"""<\/p>\n
\u00a0<\/p>\n
\u00a0 \u00a0 def __init__(self, n_clusters: int = 2, random_state: int = 42):<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 self.n_clusters = n_clusters<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 self.random_state = random_state<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 self.scaler = StandardScaler()<\/p>\n
\u00a0<\/p>\n
\u00a0 \u00a0 def detect(self, tagged_df: pd.DataFrame) -> pd.DataFrame:<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 df = tagged_df.copy()<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 # \u7279\u5f81：\u5bff\u547d + \u504f\u79bb\u5747\u503c\u6bd4\u4f8b<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 df["dev_ratio"] = df["dev_from_mean"] \/ (df["mean"] + 1e-9)<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 X = df[["life_parts", "dev_ratio"]].values<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 Xs = self.scaler.fit_transform(X)<\/p>\n
\u00a0<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 k = min(self.n_clusters, len(df))<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 km = KMeans(n_clusters=k, random_state=self.random_state, n_init=10)<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 df["cluster"] = km.fit_predict(Xs)<\/p>\n
\u00a0<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 # \u5bff\u547d\u4f4e\u7684\u7c07\u6807\u8bb0\u4e3a\u5f02\u5e38\u7c07<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 cmean = df.groupby("cluster")["life_parts"].mean()<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 low_cluster = cmean.idxmin()<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 df["cluster_is_early"] = df["cluster"] == low_cluster<\/p>\n
\u00a0<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 # \u7efc\u5408\u6807\u8bb0：\u89c4\u5219\u65e9\u635f OR \u805a\u7c7b\u5f02\u5e38\u7c07<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 df["final_early_flag"] = df["is_early"] | df["cluster_is_early"]<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 return df.sort_values("life_parts").reset_index(drop=True)<\/p>\n
\u00a0<\/p>\n
\u00a0 \u00a0 def summary(self, df: pd.DataFrame) -> pd.DataFrame:<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 sub = df[df["final_early_flag"]]<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 return sub[["tool_id", "tool_grade", "coating", "material",<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 "machine", "life_parts", "mean", "stop_reason",<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 "dev_from_mean"]].copy()<\/p>\n
\u00a0<\/p>\n
<\/details><\/p>\n
\u00a0<\/p>\n
<details><\/p>\n
\u00a0<\/p>\n
<summary><\/summary><\/p>\n
\u00a0<\/p>\n
"""\u5de5\u51b5\u5173\u7cfb\u7f51 (networkx)"""<\/p>\n
\u00a0<\/p>\n
import networkx as nx<\/p>\n
import pandas as pd<\/p>\n
from typing import Optional<\/p>\n
\u00a0<\/p>\n
\u00a0<\/p>\n
class ToolConditionGraph:<\/p>\n
\u00a0 \u00a0 """\u5200\u5177\u578b\u53f7-\u6750\u6599-\u673a\u5e8a \u4e09\u5c42\u5173\u7cfb\u7f51"""<\/p>\n
\u00a0<\/p>\n
\u00a0 \u00a0 def __init__(self):<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 self.G = nx.Graph()<\/p>\n
\u00a0<\/p>\n
\u00a0 \u00a0 def build(self, life_df: pd.DataFrame, stats_df: pd.DataFrame) -> nx.Graph:<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 self.G.clear()<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 # \u8282\u70b9<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 for _, r in stats_df.iterrows():<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 gid = f"{r['tool_grade']}|{r['coating']}"<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 self.G.add_node(gid, ntype="grade",<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 mean_life=r["mean_life"])<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 self.G.add_node(r["material"], ntype="material")<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 for m in life_df["machine"].unique():<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 self.G.add_node(m, ntype="machine")<\/p>\n
\u00a0<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 # \u8fb9：grade-material, \u6743=1\/\u5747\u503c\u5bff\u547d(\u8d8a\u8d39\u8d8a\u5927)<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 for _, r in stats_df.iterrows():<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 gid = f"{r['tool_grade']}|{r['coating']}"<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 w = round(1.0 \/ (r["mean_life"] + 1e-9), 5)<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 self.G.add_edge(gid, r["material"], weight=w,<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 mean_life=r["mean_life"])<\/p>\n
\u00a0<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 # \u673a\u5e8a\u7ea7\u5e73\u5747\u5bff\u547d<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 mc = life_df.groupby(["machine", "tool_grade", "coating"])["life_parts"].mean().reset_index()<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 for _, r in mc.iterrows():<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 gid = f"{r['tool_grade']}|{r['coating']}"<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 self.G.add_edge(gid, r["machine"],<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 weight=round(1.0 \/ (r["life_parts"] + 1e-9), 5),<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 mean_life=round(r["life_parts"], 1))<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 return self.G<\/p>\n
\u00a0<\/p>\n
\u00a0 \u00a0 def bottleneck_paths(self) -> pd.DataFrame:<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 """\u627e\u5e73\u5747\u5bff\u547d\u6700\u4f4e\u7684 grade-material-machine \u7ec4\u5408"""<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 rows = []<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 for u, v, d in self.G.edges(data=True):<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 if "mean_life" in d:<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 rows.append({"node_a": u, "node_b": v,<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0"mean_life": d["mean_life"],<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0"weight": d["weight"]})<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 df = pd.DataFrame(rows)<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 if df.empty:<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 return df<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 return df.sort_values("mean_life").reset_index(drop=True)<\/p>\n
\u00a0<\/p>\n
<\/details><\/p>\n
\u00a0<\/p>\n
<details><\/p>\n
\u00a0<\/p>\n
<summary><\/summary><\/p>\n
\u00a0<\/p>\n
"""\u53ef\u89c6\u5316"""<\/p>\n
\u00a0<\/p>\n
import numpy as np<\/p>\n
import pandas as pd<\/p>\n
import matplotlib.pyplot as plt<\/p>\n
import networkx as nx<\/p>\n
from pathlib import Path<\/p>\n
\u00a0<\/p>\n
plt.rcParams["font.sans-serif"] = ["SimHei", "DejaVu Sans"]<\/p>\n
plt.rcParams["axes.unicode_minus"] = False<\/p>\n
\u00a0<\/p>\n
\u00a0<\/p>\n
class ToolVisualizer:<\/p>\n
\u00a0 \u00a0 def __init__(self, results_dir: str = "results"):<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 self.results_dir = Path(results_dir)<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 self.results_dir.mkdir(exist_ok=True)<\/p>\n
\u00a0<\/p>\n
\u00a0 \u00a0 def box_life(self, life_df, stats_df):<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 fig, ax = plt.subplots(figsize=(14, 6))<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 labels = stats_df.apply(lambda r: f"{r['tool_grade']}\\\\n{r['coating']}\\\\n{r['material']}", axis=1)<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 data = []<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 for _, r in stats_df.iterrows():<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 sub = life_df[(life_df["tool_grade"] == r["tool_grade"]) &<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 (life_df["coating"] == r["coating"]) &<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 (life_df["material"] == r["material"])]["life_parts"].values<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 data.append(sub)<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 ax.boxplot(data, labels=labels, showmeans=True)<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 ax.set_ylabel("\u53ef\u52a0\u5de5\u4ef6\u6570", fontsize=12)<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 ax.set_title("\u5404\u578b\u53f7\u5200\u5177\u5bff\u547d\u7bb1\u7ebf\u56fe", fontsize=14, fontweight="bold")<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 plt.xticks(rotation=30, ha="right", fontsize=8)<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 ax.grid(axis="y", alpha=0.3)<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 plt.tight_layout()<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 plt.savefig(self.results_dir \/ "box_life_by_grade.png",<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0dpi=150, bbox_inches="tight")<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 plt.close()<\/p>\n
\u00a0<\/p>\n
\u00a0 \u00a0 def weibull_probplot(self, life_df, weibull_df):<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 fig, ax = plt.subplots(figsize=(12, 7))<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 for _, w in weibull_df.iterrows():<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 sub = life_df[(life_df["tool_grade"] == w["tool_grade"]) &<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 (life_df["coating"] == w["coating"]) &<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 (life_df["material"] == w["material"])]["life_parts"].values<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 x, y = WeibullProbHelper().probplot_data(sub)<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 if len(x) < 2:<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 continue<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 ax.plot(x, y, ".-", label=f"{w['tool_grade']}|{w['coating']}|{w['material']} k={w['weibull_k']}")<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 ax.set_xscale("log")<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 ax.set_xlabel("\u5bff\u547d\u4ef6\u6570 (log)", fontsize=12)<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 ax.set_ylabel("ln(-ln(1-F))", fontsize=12)<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 ax.set_title("\u5a01\u5e03\u5c14\u6982\u7387\u56fe", fontsize=14, fontweight="bold")<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 ax.legend(fontsize=7)<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 ax.grid(alpha=0.3)<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 plt.tight_layout()<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 plt.savefig(self.results_dir \/ "weibull_probplot.png",<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0dpi=150, bbox_inches="tight")<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 plt.close()<\/p>\n
\u00a0<\/p>\n
\u00a0 \u00a0 def anomaly_scatter(self, det_df):<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 fig, ax = plt.subplots(figsize=(12, 6))<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 normal = det_df[~det_df["final_early_flag"]]<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 early = det_df[det_df["final_early_flag"]]<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 ax.scatter(normal["mean"], normal["life_parts"], c="#3498DB",<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0label="\u6b63\u5e38\u5200", s=30, alpha=0.7)<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 ax.scatter(early["mean"], early["life_parts"], c="#E74C3C",<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0label="\u65e9\u635f\u5200", s=60, edgecolor="k")<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 # \u53c2\u8003\u7ebf y=mean<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 allx = det_df["mean"].values<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 ax.plot(allx, allx, "g–", alpha=0.5, label="\u540c\u7ec4\u5747\u503c\u7ebf")<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 for _, r in early.iterrows():<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 ax.annotate(r["tool_id"], (r["mean"], r["life_parts"]),<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 fontsize=7, color="red")<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 ax.set_xlabel("\u540c\u5de5\u51b5\u7ec4\u5e73\u5747\u5bff\u547d(\u4ef6)", fontsize=12)<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 ax.set_ylabel("\u672c\u5200\u5bff\u547d(\u4ef6)", fontsize=12)<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 ax.set_title("\u65e9\u635f\u5200\u5177\u8bc6\u522b\u6563\u70b9\u56fe", fontsize=14, fontweight="bold")<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 ax.legend()<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 ax.grid(alpha=0.3)<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 plt.tight_layout()<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 plt.savefig(self.results_dir \/ "anomaly_scatter.png",<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0dpi=150, bbox_inches="tight")<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 plt.close()<\/p>\n
\u00a0<\/p>\n
\u00a0 \u00a0 def condition_graph(self, G):<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 fig, ax = plt.subplots(figsize=(14, 9))<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 pos = nx.spring_layout(G, seed=42, k=0.6)<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 ncolors = []<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 for n, d in G.nodes(data=True):<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 t = d.get("ntype", "grade")<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 ncolors.append({"grade": "#E74C3C", "material": "#2ECC71",<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 "machine": "#3498DB"}.get(t, "#999"))<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 sizes = [300 + (G.nodes[n].get("mean_life", 50) * 3 if G.nodes[n].get("ntype")=="grade" else 300) for n in G.nodes()]<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 nx.draw_networkx_nodes(G, pos, node_color=ncolors,<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0node_size=sizes, ax=ax, alpha=0.9)<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 edges = [(u, v) for u, v, d in G.edges(data=True)]<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 ws = [G.edges[u, v]["weight"] * 8000 for u, v in edges]<\/p>\n
\u00a0 \u00a0 \u00a0 \u00a0 nx.draw_networkx_edg<\/p>\n
\u5229\u7528AI\u89e3\u51b3\u5b9e\u9645\u95ee\u9898，\u5982\u679c\u4f60\u89c9\u5f97\u8fd9\u4e2a\u5de5\u5177\u597d\u7528，\u6b22\u8fce\u5173\u6ce8\u957f\u5b89\u7267\u7b1b！<\/p>\n","protected":false},"excerpt":{"rendered":"
\u5468\u4e09\u591c\u73ed，\u673a\u52a0\u8f66\u95f4\u3002\\”\u8fd9\u6279\u4ef6\u5e72\u5230 180 \u4ef6，\u5200\u53c8\u5d29\u4e86，\\”\u5200\u5177\u7ba1\u7406\u5458\u8001\u5468\u628a\u51e0\u6839\u786c\u8d28\u5408\u91d1\u5200\u6446\u684c\u4e0a，\\”\u7cfb\u7edf\u91cc\u8bb0\u7684\u662f\u6362\u5200\u4e8b\u4ef6，\u53ef\u6211\u60f3\u77e5\u9053：\u540c\u724c\u53f7\u7684\u5200，\u5e72\u8fd9\u4e2a\u6750\u6599\u5230\u5e95\u80fd\u6491\u591a\u5c11\u4ef6？\u662f 150 \u4ef6\u5c31\u5230\u5934，\u8fd8\u662f\u80fd\u5230 220？\u73b0\u5728\u5168\u9760\u8001\u5e08\u5085\u62cd\u8111\u888b，\u6362\u65e9\u4e86\u6d6a\u8d39\u5200，\u6362\u665a\u4e86\u51fa\u5e9f\u54c1\u3002\\”\u6211\u6253\u5f00 MES \u7684\u5bfc\u51fa\u8868\u3002\\”\u8fd9\u8868\u91cc\u6709\u4ec0\u4e48ÿ<\/p>\n","protected":false},"author":2,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[1],"tags":[2468,4278,81,737,2606],"topic":[],"class_list":["post-112153","post","type-post","status-publish","format-standard","hentry","category-server","tag-matplotlib","tag-pandas","tag-python","tag-scikit-learn","tag-2606"],"yoast_head":"\n