We propose using a frozen Chronos-2 time-series foundation model as a feature extractor paired with a lightweight MLP regression head for predicting Remaining Useful Life of industrial equipment, outperforming recurrent, convolutional, transformer, and gradient-boosting baselines on real-world sensor data.</p>\n","updatedAt":"2026-06-11T08:41:49.996Z","author":{"_id":"63de9b4310678851bb086b5e","avatarUrl":"/avatars/a4b2b57d81dfbff0cf1893f93a4d63d7.svg","fullname":"Amir El-Ghoussani","name":"AmirMaEl","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"isUserFollowing":false}},"numEdits":0,"identifiedLanguage":{"language":"en","probability":0.8172107338905334},"editors":["AmirMaEl"],"editorAvatarUrls":["/avatars/a4b2b57d81dfbff0cf1893f93a4d63d7.svg"],"reactions":[],"isReport":false}}],"primaryEmailConfirmed":false,"paper":{"id":"2606.11990","authors":[{"_id":"6a2a7326eb926374a3821961","user":{"_id":"63de9b4310678851bb086b5e","avatarUrl":"/avatars/a4b2b57d81dfbff0cf1893f93a4d63d7.svg","isPro":false,"fullname":"Amir El-Ghoussani","user":"AmirMaEl","type":"user","name":"AmirMaEl"},"name":"Amir El-Ghoussani","status":"claimed_verified","statusLastChangedAt":"2026-06-11T08:36:50.534Z","hidden":false},{"_id":"6a2a7326eb926374a3821962","name":"Michele De Vita","hidden":false},{"_id":"6a2a7326eb926374a3821963","name":"Ronald Naumann","hidden":false},{"_id":"6a2a7326eb926374a3821964","name":"Valiseios Belagiannis","hidden":false}],"publishedAt":"2026-06-10T00:00:00.000Z","submittedOnDailyAt":"2026-06-11T00:00:00.000Z","title":"Time-Series Foundation Model Embeddings for Remaining Useful Life Estimation","submittedOnDailyBy":{"_id":"63de9b4310678851bb086b5e","avatarUrl":"/avatars/a4b2b57d81dfbff0cf1893f93a4d63d7.svg","isPro":false,"fullname":"Amir El-Ghoussani","user":"AmirMaEl","type":"user","name":"AmirMaEl"},"summary":"Remaining Useful Life (RUL) prediction is essential for industrial predictive maintenance, yet many learning-based approaches rely on extensive feature engineering or large labeled datasets to train task-specific sequence models. In this work, we introduce a lightweight learning approach, in which we leverage a frozen pretrained time-series foundation model (TSFM) and combine it with a small regression head for RUL estimation from multivariate sensor streams. More specifically, we use Chronos-2 as a frozen backbone to extract context window features and train a lightweight regression neural network for RUL prediction. Experiments on real-world industrial sensor data from two device types show that Chronos-2 features consistently improve over recurrent, convolutional, Transformer-based, and gradient-boosting baselines under the same preprocessing and evaluation protocol. We further analyze the impact of context length and find that performance improves significantly with longer histories, indicating that TSFM representation offer a practical and data-efficient alternative for RUL estimation in industrial settings.","upvotes":3,"discussionId":"6a2a7327eb926374a3821965","ai_summary":"A lightweight approach combining a frozen pretrained time-series foundation model with a simple regression head achieves superior RUL prediction performance compared to various baseline methods on industrial sensor data.","ai_keywords":["time-series foundation model","Chronos-2","frozen backbone","regression neural network","RUL estimation","multivariate sensor streams","context window features","industrial predictive maintenance"],"ai_summary_model":"Qwen/Qwen2.5-Coder-32B-Instruct","organization":{"_id":"636d1e96dddcc01da6560fa6","name":"unierlangen","fullname":"friedrich alexander university","avatar":"https://www.gravatar.com/avatar/365eebb04ee97eb71542d5113a1b4cd5?d=retro&size=100"}},"canReadDatabase":false,"canManagePapers":false,"canSubmit":false,"hasHfLevelAccess":false,"upvoted":false,"upvoters":[{"_id":"641597d6486c7c9a5d14247f","avatarUrl":"/avatars/73a0934fb8975730ed093e6518d5ce5c.svg","isPro":false,"fullname":"Vasileios Belagiannis","user":"bazilas","type":"user"},{"_id":"6a2ae6c2e36bc84d91b6e7cc","avatarUrl":"/avatars/abf4b4c0020f9332b6827952cc53163e.svg","isPro":false,"fullname":"mmgood","user":"mmgood","type":"user"},{"_id":"63de9b4310678851bb086b5e","avatarUrl":"/avatars/a4b2b57d81dfbff0cf1893f93a4d63d7.svg","isPro":false,"fullname":"Amir El-Ghoussani","user":"AmirMaEl","type":"user"}],"acceptLanguages":["en"],"dailyPaperRank":0,"organization":{"_id":"636d1e96dddcc01da6560fa6","name":"unierlangen","fullname":"friedrich alexander university","avatar":"https://www.gravatar.com/avatar/365eebb04ee97eb71542d5113a1b4cd5?d=retro&size=100"},"markdownContentUrl":"https://huggingface.co/buckets/huggingchat/papers-content/resolve/2606/2606.11990.md"}">
Time-Series Foundation Model Embeddings for Remaining Useful Life Estimation
Abstract
A lightweight approach combining a frozen pretrained time-series foundation model with a simple regression head achieves superior RUL prediction performance compared to various baseline methods on industrial sensor data.
Remaining Useful Life (RUL) prediction is essential for industrial predictive maintenance, yet many learning-based approaches rely on extensive feature engineering or large labeled datasets to train task-specific sequence models. In this work, we introduce a lightweight learning approach, in which we leverage a frozen pretrained time-series foundation model (TSFM) and combine it with a small regression head for RUL estimation from multivariate sensor streams. More specifically, we use Chronos-2 as a frozen backbone to extract context window features and train a lightweight regression neural network for RUL prediction. Experiments on real-world industrial sensor data from two device types show that Chronos-2 features consistently improve over recurrent, convolutional, Transformer-based, and gradient-boosting baselines under the same preprocessing and evaluation protocol. We further analyze the impact of context length and find that performance improves significantly with longer histories, indicating that TSFM representation offer a practical and data-efficient alternative for RUL estimation in industrial settings.
Community
We propose using a frozen Chronos-2 time-series foundation model as a feature extractor paired with a lightweight MLP regression head for predicting Remaining Useful Life of industrial equipment, outperforming recurrent, convolutional, transformer, and gradient-boosting baselines on real-world sensor data.
Upload images, audio, and videos by dragging in the text input, pasting, or clicking here.
Tap or paste here to upload images
Cite arxiv.org/abs/2606.11990 in a model README.md to link it from this page.
Cite arxiv.org/abs/2606.11990 in a dataset README.md to link it from this page.
Cite arxiv.org/abs/2606.11990 in a Space README.md to link it from this page.
Discussion (0)
Sign in to join the discussion. Free account, 30 seconds — email code or GitHub.
Sign in →No comments yet. Sign in and be the first to say something.