刀具磨耗預測為智慧製造中重要的品質與維護問題。當材料、進給率或切深改變時，既有刀具磨耗預測模型常因資料分布差異而失效；然而新條件下可取得的標註資料通常有限，使重新訓練模型的成本與風險提高。
本研究提出一套基於封閉式地端協作代理人之遷移式刀具磨耗預測模型訓練方法，設計人類科學家與封閉式地端協作代理人間的人機協作模式及階段導向技術架構，並開發知識檢索、候選規劃、工具調用、審查控制與經驗回寫等實現技術。所建構之系統使代理人能在固定模型與有限訓練預算下，依據加工條件差異、少量目標域標籤、文獻規則與歷史實驗經驗，形成可執行且可稽核的遷移策略與參數決策。
本研究以 NASA Milling 資料集建立來源域至目標域之加工條件遷移任務，將材料、進給率與切深作為加工條件域的定義。預測模型固定為極端梯度提升（Extreme Gradient Boosting, XGBoost），代理人流程依序包含現況理解、目標需求分析、知識檢索與遷移策略選擇、候選與實驗計畫，以及執行、驗證與評估。為避免資料洩漏，目標域測試集在決策與調參期間均不可用；歷史知識亦排除目前目標域。代理人輸出採用結構化決策紀錄，記錄候選集合、限制、引用依據與決策結果，不保存完整自然語言推理過程。
實驗以二十個有向遷移組合及 0.1、0.3 兩種目標標籤比例形成四十組情境。完整代理人的平均絕對誤差（Mean Absolute Error, MAE）為 0.0820，低於僅用來源域資料基準的 0.1367 與僅用目標域資料基準的 0.1933；決定係數（Coefficient of Determination, R²）為 0.6372，高於兩項基準的 0.0252 與 -1.0312。在相同三次新增模型訓練預算與驗證規則下，完整代理人的四項平均指標亦優於固定規則法、無知識代理人與隨機候選法。受控組件替換中，替換 AutoGluon 參考後的 MAE 上升至 0.1019，為最明顯的平均退化。完整代理人於四十組情境中有三組依 MAE 判準發生負遷移，其中兩組為 MAE 略高但 RMSE 與 R² 仍改善的邊界情形。結果支持地端知識、跨情境歷史、參數參考與驗證選模的整體效益，但不保證每一遷移情境皆優於未調適來源模型。

Tool wear prediction is essential to machining quality and maintenance, but models trained under one material, feed rate, or depth of cut can lose accuracy after process conditions change. This study develops a closed on-premises collaborative agent for transfer learning under limited target-domain labels. The agent and a human scientist follow a five-stage workflow covering problem understanding, requirement analysis, knowledge retrieval and strategy selection, candidate planning, and execution, validation, and evaluation. The technical implementation combines domain-shift diagnostics, a Neo4j knowledge graph, document retrieval, cross-scenario experiment history, AutoGluon references, constrained candidate construction, CUDA XGBoost training, validation-based selection and ensemble construction, and auditable decision records. The NASA Milling Data Set is organized into machining-condition domains defined by material, feed rate, and depth of cut. Forty transfer scenarios are formed from twenty directed source-to-target pairs and target-label ratios of 0.1 and 0.3. Under the same Budget-3 constraint, the full agent achieves an average MAE of 0.0820, RMSE of 0.1015, R-squared of 0.6372, and nMAE of 0.1635. Its MAE is lower than the source-only, target-only, mixed-data, random-candidate, fixed-rule, and no-knowledge alternatives. Negative transfer occurs in three scenarios, showing that transfer remains condition dependent. The results indicate that knowledge- and experience-guided candidate formation can improve average prediction performance while preserving local deployment, test isolation, human review, and traceability.

Agogino, A., & Goebel, K. (2007). Milling data set. NASA Prognostics Data Repository, NASA Ames Research Center. https://catalog.data.gov/dataset/milling-wear.
Baratchi, M., Wang, C., Limmer, S., van Rijn, J. N., Hoos, H., Bäck, T., & Olhofer, M. (2024). Automated machine learning: past, present and future. Artificial Intelligence Review, 57, 122. https://doi.org/10.1007/s10462-024-10726-1.
Ben-David, S., Blitzer, J., Crammer, K., Kulesza, A., Pereira, F., & Vaughan, J. W. (2010). A theory of learning from different domains. Machine Learning, 79(1–2), 151–175. https://doi.org/10.1007/s10994-009-5152-4.
Bergstra, J., & Bengio, Y. (2012). Random search for hyper-parameter optimization. Journal of Machine Learning Research, 13, 281–305. https://www.jmlr.org/papers/v13/bergstra12a.html.
Byrne, G., Dornfeld, D., Inasaki, I., Ketteler, G., König, W., & Teti, R. (1995). Tool condition monitoring (TCM) — the status of research and industrial application. CIRP Annals, 44(2), 541–567. https://doi.org/10.1016/S0007-8506(07)60503-4.
Cai, W., Zhang, W., Hu, X., & Liu, Y. (2020). A hybrid information model based on long short-term memory network for tool condition monitoring. Journal of Intelligent Manufacturing, 31(6), 1497–1510. https://doi.org/10.1007/s10845-019-01526-4.
Cai, Z., Li, W., Song, J., Jin, H., & Fu, H. (2025). Tool wear state identification method with variable cutting parameters based on multi-source unsupervised domain adaptation. Sensors, 25(6), 1742. https://doi.org/10.3390/s25061742.
Caruana, R., Niculescu-Mizil, A., Crew, G., & Ksikes, A. (2004). Ensemble selection from libraries of models. In Proceedings of the Twenty-First International Conference on Machine Learning (p. 18). https://doi.org/10.1145/1015330.1015432.
Chen, T., & Guestrin, C. (2016). XGBoost: A scalable tree boosting system. In Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining (pp. 785–794). https://doi.org/10.1145/2939672.2939785.
Efron, B. (1979). Bootstrap methods: Another look at the jackknife. The Annals of Statistics, 7(1), 1–26. https://doi.org/10.1214/aos/1176344552.
Erickson, N., Mueller, J., Shirkov, A., Zhang, H., Larroy, P., Li, M., & Smola, A. (2020). AutoGluon-Tabular: Robust and accurate AutoML for structured data. arXiv. https://doi.org/10.48550/arXiv.2003.06505.
Ganin, Y., Ustinova, E., Ajakan, H., Germain, P., Larochelle, H., Laviolette, F., Marchand, M., & Lempitsky, V. (2016). Domain-adversarial training of neural networks. Journal of Machine Learning Research, 17(59), 1–35. https://www.jmlr.org/papers/v17/15-239.html.
Gerganov, G., & llama.cpp contributors. (2026). llama.cpp: LLM inference in C/C++. GitHub. https://github.com/ggml-org/llama.cpp.
ggml-org. (2026). Gemma 4 E4B instruction-tuned GGUF model. Hugging Face. https://huggingface.co/ggml-org/gemma-4-E4B-it-GGUF.
Goebel, K., & Agogino, A. (2007). Documentation for mill data set. NASA Ames Research Center; UC Berkeley BEST Lab. https://phm-datasets.s3.amazonaws.com/NASA/3.+Milling.zip.
Google DeepMind. (2026). Gemma 4 model card. Google AI for Developers. https://ai.google.dev/gemma/docs/core/model_card_4.
Gretton, A., Borgwardt, K. M., Rasch, M. J., Schölkopf, B., & Smola, A. (2012). A kernel two-sample test. Journal of Machine Learning Research, 13, 723–773. https://www.jmlr.org/papers/v13/gretton12a.html.
He, J., Sun, Y., Yin, C., He, Y., & Wang, Y. (2023). Cross-domain adaptation network based on attention mechanism for tool wear prediction. Journal of Intelligent Manufacturing, 34(8), 3365–3387. https://doi.org/10.1007/s10845-022-02005-z.
Hou, W., Guo, H., Luo, L., & Jin, M. (2022). Tool wear prediction based on domain adversarial adaptation and channel attention multiscale convolutional long short-term memory network. Journal of Manufacturing Processes, 84, 1339–1361. https://doi.org/10.1016/j.jmapro.2022.11.017.
Huang, Z., Li, W., Zhu, J., & Wang, L. (2024). Cross-domain tool wear condition monitoring via residual attention hybrid adaptation network. Journal of Manufacturing Systems, 72, 406–423. https://doi.org/10.1016/j.jmsy.2023.12.003.
Jiang, W., & Hu, F. (2025). Artificial intelligence agent-enabled predictive maintenance: Conceptual proposal and basic framework. Computers, 14(8), 329. https://doi.org/10.3390/computers14080329.
Kaufman, S., Rosset, S., Perlich, C., & Stitelman, O. (2012). Leakage in data mining: Formulation, detection, and avoidance. ACM Transactions on Knowledge Discovery from Data, 6(4), 1–21. https://doi.org/10.1145/2382577.2382579.
Kvålseth, T. O. (1985). Cautionary note about R². The American Statistician, 39(4), 279–285. https://doi.org/10.1080/00031305.1985.10479448.
Lewis, P., Perez, E., Piktus, A., Petroni, F., Karpukhin, V., Goyal, N., Küttler, H., Lewis, M., Yih, W.-T., Rocktäschel, T., Riedel, S., & Kiela, D. (2020). Retrieval-augmented generation for knowledge-intensive NLP tasks. In Advances in Neural Information Processing Systems (Vol. 33, pp. 9459–9474). https://papers.neurips.cc/paper/2020/hash/6b493230205f780e1bc26945df7481e5-Abstract.html.
Li, J., Lu, J., Chen, C., Ma, J., & Liao, X. (2021). Tool wear state prediction based on feature-based transfer learning. The International Journal of Advanced Manufacturing Technology, 113(11–12), 3283–3301. https://doi.org/10.1007/s00170-021-06780-6.
Li, K., Chen, M., Lin, Y., Li, Z., Jia, X., & Li, B. (2022). A novel adversarial domain adaptation transfer learning method for tool wear state prediction. Knowledge-Based Systems, 254, 109537. https://doi.org/10.1016/j.knosys.2022.109537.
Liu, C., Li, Y., Li, J., & Hua, J. (2022). A meta-invariant feature space method for accurate tool wear prediction under cross conditions. IEEE Transactions on Industrial Informatics, 18(2), 922–931. https://doi.org/10.1109/TII.2021.3070109.
Liu, X., Yu, H., Zhang, H., Xu, Y., Lei, X., Lai, H., Gu, Y., Ding, H., Men, K., Yang, K., Zhang, S., Deng, X., Zeng, A., Du, Z., Zhang, C., Shen, S., Zhang, T., Su, Y., Sun, H., … Tang, J. (2023). AgentBench: Evaluating LLMs as agents. arXiv. https://doi.org/10.48550/arXiv.2308.03688.
Marei, M., Zaatari, S. E., & Li, W. (2021). Transfer learning enabled convolutional neural networks for estimating health state of cutting tools. Robotics and Computer-Integrated Manufacturing, 71, 102145. https://doi.org/10.1016/j.rcim.2021.102145.
Mo, X., Hu, X., Sun, A., & Zhang, Y. (2025). A meta-learning method for smart manufacturing: Tool wear prediction using hybrid information under various operating conditions. Robotics and Computer-Integrated Manufacturing, 91, 102846. https://doi.org/10.1016/j.rcim.2024.102846.
Mohamed, A., Hassan, M., M’Saoubi, R., & Attia, H. (2022). Tool condition monitoring for high-performance machining systems—a review. Sensors, 22(6), 2206. https://doi.org/10.3390/s22062206.
Neo4j. (2026). Neo4j documentation. https://neo4j.com/docs/.
Pan, S. J., & Yang, Q. (2010). A survey on transfer learning. IEEE Transactions on Knowledge and Data Engineering, 22(10), 1345–1359. https://doi.org/10.1109/TKDE.2009.191.
Pardoe, D., & Stone, P. (2010). Boosting for regression transfer. In Proceedings of the 27th International Conference on Machine Learning (pp. 863–870). https://dl.acm.org/doi/10.5555/3104322.3104432.
Pech, M., Vrchota, J., & Bednář, J. (2021). Predictive maintenance and intelligent sensors in smart factory: Review. Sensors, 21(4), 1470. https://doi.org/10.3390/s21041470.
Pimenov, D. Y., Bustillo, A., Wojciechowski, S., Sharma, V. S., Gupta, M. K., & Kuntoğlu, M. (2023). Artificial intelligence systems for tool condition monitoring in machining: analysis and critical review. Journal of Intelligent Manufacturing, 34(5), 2079–2121. https://doi.org/10.1007/s10845-022-01923-2.
Schick, T., Dwivedi-Yu, J., Dessi, R., Raileanu, R., Lomeli, M., Zettlemoyer, L., Cancedda, N., & Scialom, T. (2023). Toolformer: Language models can teach themselves to use tools. arXiv. https://doi.org/10.48550/arXiv.2302.04761.
Shinn, N., Cassano, F., Gopinath, A., Narasimhan, K., & Yao, S. (2023). Reflexion: Language agents with verbal reinforcement learning. In Advances in Neural Information Processing Systems (Vol. 36, pp. 8634–8652). https://doi.org/10.52202/075280-0377.
Sugiyama, M., Krauledat, M., & Müller, K.-R. (2007). Covariate shift adaptation by importance weighted cross validation. Journal of Machine Learning Research, 8, 985–1005. https://www.jmlr.org/papers/v8/sugiyama07a.html.
Sun, B., Feng, J., & Saenko, K. (2016). Return of frustratingly easy domain adaptation. In Proceedings of the AAAI Conference on Artificial Intelligence (Vol. 30, pp. 2058–2065). https://doi.org/10.1609/aaai.v30i1.10306.
Sun, C., Ma, M., Zhao, Z., Tian, S., Yan, R., & Chen, X. (2019). Deep transfer learning based on sparse autoencoder for remaining useful life prediction of tool in manufacturing. IEEE Transactions on Industrial Informatics, 15(4), 2416–2425. https://doi.org/10.1109/TII.2018.2881543.
Teti, R., Jemielniak, K., O’Donnell, G., & Dornfeld, D. (2010). Advanced monitoring of machining operations. CIRP Annals, 59(2), 717–739. https://doi.org/10.1016/j.cirp.2010.05.010.
Wang, L., Ma, C., Feng, X., Zhang, Z., Yang, H., Zhang, J., Chen, Z., Tang, J., Chen, X., Lin, Y., Zhao, W. X., Wei, Z., & Wen, J. (2024). A survey on large language model based autonomous agents. Frontiers of Computer Science, 18, 186345. https://doi.org/10.1007/s11704-024-40231-1.
Wang, M., Zhou, J., Gao, J., Li, Z., & Li, E. (2020). Milling tool wear prediction method based on deep learning under variable working conditions. IEEE Access, 8, 140726–140735. https://doi.org/10.1109/ACCESS.2020.3010378.
Wang, P. (Edward), & Russell, M. (2020). Domain adversarial transfer learning for generalized tool wear prediction. Annual Conference of the PHM Society, 12(1), 8. https://doi.org/10.36001/phmconf.2020.v12i1.1137.
Wang, Y., Gao, J., Wang, W., Du, J., & Yang, X. (2024). A novel method based on deep transfer learning for tool wear state prediction under cross-dataset. The International Journal of Advanced Manufacturing Technology, 131(1–2), 171–182. https://doi.org/10.1007/s00170-024-13055-3.
Wei, J., Wang, X., Schuurmans, D., Bosma, M., Ichter, B., Xia, F., Chi, E. H., Le, Q. V., & Zhou, D. (2022). Chain-of-thought prompting elicits reasoning in large language models. In Advances in Neural Information Processing Systems (Vol. 35, pp. 24824–24837). https://papers.neurips.cc/paper_files/paper/2022/hash/9d5609613524ecf4f15af0f7b31abca4-Abstract-Conference.html.
Wu, Q., Bansal, G., Zhang, J., Wu, Y., Li, B., Zhu, E., Jiang, L., Zhang, X., Zhang, S., Liu, J., Awadallah, A. H., White, R. W., Burger, D., & Wang, C. (2023). AutoGen: Enabling next-gen LLM applications via multi-agent conversation. arXiv. https://doi.org/10.48550/arXiv.2308.08155.
Xie, R., & Wu, D. (2021). Optimal transport-based transfer learning for smart manufacturing: Tool wear prediction using out-of-domain data. Manufacturing Letters, 29, 104–107. https://doi.org/10.1016/j.mfglet.2021.08.004.
Yao, S., Zhao, J., Yu, D., Du, N., Shafran, I., Narasimhan, K., & Cao, Y. (2023). ReAct: Synergizing reasoning and acting in language models. In International Conference on Learning Representations. https://openreview.net/forum?id=WE_vluYUL-X.
Zhang, P., Gao, D., Hong, D., Lu, Y., Wang, Z., & Liao, Z. (2023). Intelligent tool wear monitoring based on multi-channel hybrid information and deep transfer learning. Journal of Manufacturing Systems, 69, 31–47. https://doi.org/10.1016/j.jmsy.2023.06.004.
Zhou, Y., Liu, C., Yu, X., Liu, B., & Quan, Y. (2022). Tool wear mechanism, monitoring and remaining useful life (RUL) technology based on big data: a review. SN Applied Sciences, 4, 232. https://doi.org/10.1007/s42452-022-05114-9.
Zhu, Y., Zi, Y., Xu, J., & Li, J. (2022). An unsupervised dual-regression domain adversarial adaption network for tool wear prediction in multi-working conditions. Measurement, 200, 111644. https://doi.org/10.1016/j.measurement.2022.111644.