# Predicting Industrial Equipment’s Remaining Useful Life with AI AI is increasingly used for predictive maintenance (PdM) of industrial equipment. In settings such as semiconductor plants, the approach anticipates equipment anomalies so maintenance can be planned before failure. Source: https://www.iplexlaw.co.kr/en/blog/844148 HOME / NEWS & INSIGHTS NEWS & INSIGHTS Predicting Industrial Equipment’s Remaining Useful Life with AI AI is increasingly used for predictive maintenance (PdM) of industrial equipment. In settings such as semiconductor plants, the approach anticipates equipment anomalies so maintenance can be planned before failure. AI & Software 2023.05.15 published IPLEX 5 min read Artificial intelligence is used in many areas. In factories, it supports predictive maintenance (PdM) of equipment such as robotic arms. AI can support predictive maintenance by identifying signs of abnormal operation in industrial equipment, including semiconductor-plant machinery, so maintenance can be planned in advance. Unexpected equipment failures can interrupt production and cause substantial losses, making reliability important. Earlier approaches often relied on predefined maintenance rules. As sensor datasets grew, detecting abnormal behavior quickly and accurately became harder. AI-based analysis is one way to address that challenge. Time-series analysis can estimate equipment's remaining useful life (RUL). This article reviews a deep-learning paper on RUL prediction and considers technical features relevant to patent drafting. The approach is outlined below. This is the author's explanatory summary and may simplify aspects of the original paper. References Xiang Li & Qian Ding & Jian-Qiao Sun, 2017, Remaining Useful Life Estimation in Prognostics Using Deep Convolution Neural Networks, https://escholarship.org/content/qt5ns8r3fs/qt5ns8r3fs_noSplash_7d463f0d432194b63917f89e056f2644.pdf >> Approach to predict the remaining life of industrial equipment << First, let’s take a quick look at what approaches are available to predict the remaining life of industrial equipment. RUL prediction generally uses three approaches. The first is a model-based approach. It uses a model of degradation to estimate remaining useful life, for example through an Eyring model or a suitable particle-filter estimation framework. The second is a data-driven approach. It learns degradation patterns from historical observations. AI-based RUL estimation falls within this category. The third is a hybrid approach, which combines model-based and data-driven methods. Today’s paper relates to predicting RUL through a data-driven approach. >> Summary of article content << The paper estimates aircraft-engine remaining useful life from sensor data. Using DeepCNN to predict the final output associated with RUL, the structure of DeepCNN is as follows: The input is a window of sensor data arranged as a feature map, with sensor channels along one axis and time steps along the other. Once the data sampling is complete, enter it sequentially in four convolution layers with filters of the same size. And the last convolution layer has a filter of different size than the one used in the four convolution layers. Enter the last output of the four convolution layers in the last convolution layer. In this case, a high-level representation of the raw feature is output from the last convolution layer. The last convolution layer is then connected to a fully connected layer and finally the value for the RUL prediction is output. In the reported experiments, the proposed deep CNN outperformed the compared DNN, RNN and LSTM models for RUL estimation. These results relate to the paper's experimental setting. >> Extracting key features for patent applications << I will now explain to you, based on the previous article, what would be the main feature and what would be the patent application. I believe that using the DeepCNN structure to predict RUL is a major feature of this paper. If you have a DeepCNN structure, there are experimental results that show better performance than other models. Including relevant comparative experimental results in a specification can help explain the technical effect when responding to an inventive-step objection. On the other hand, DeepCNN’s detail structure can be the content that underpins key features. Specifically, the detailed structure of the model, in which a plurality of first convolution layers with a first filter of the same size are connected to a second convolution layer with a second filter of different size with the first filter, and the second convolution layer is connected to a fully connected layer, may support the main features of this paper. An incidental feature is the sampling of input data using the Time window and related to data normalization. A hypothetical claim structure for explaining this approach could include: 1. Inputting sensor data into a deep CNN with convolutional layers and a fully connected layer. 2. Estimating remaining useful life from the network's output. Subject to prior art and disclosure support, further claims could address: 1. Detail structure of DeepCNN (features for filters of multiple convolution layers, order of layer connections, etc.) 2. How to sample data input to DeepCNN 3. How to Normalize Data These features could provide additional claim limitations. Read the Korean source This article reflects the information available when it was published. Contact us to discuss your circumstances. Discuss this topic ↗ All articles TALK TO IPLEX Discuss your IP questions We consider your technology and business needs together. ↗ Contact us Newer Column: Yongduck Kim on Securing AI Patents ↗ Older Publication Announcement: Design Protection Law — Protect Your Designs ↗ Related insights AI & Software 2026.10.01 FiX: fine-grained forgetting in softmax attention Yongduck Kim examines FiX’s feature-wise gates, numerical implementation and paged cache, distinguishing reported gains from unresolved limitations. ↗ Read article AI & Software 2026.09.30 MHAR: Reading earlier layers through different feature subspaces Yongduck Kim examines Multi-Head Attention Residuals: depth routing, reported training results, implementation costs and the relationship between technical features and effects. ↗ Read article AI & Software 2026.09.28 Column: Claude Computer Use and the Data That Trains AI Agents Writing for AI Times, IPLEX Managing Partner Yongduck Kim examines the training data behind computer-operating AI agents through U.S. Patent No. 12,585,862. ↗ Read article - https://www.iplexlaw.co.kr/en - https://www.iplexlaw.co.kr/en/blog/category/ai - https://www.iplexlaw.co.kr/forum/view/844148 - https://www.iplexlaw.co.kr/en/contact - https://www.iplexlaw.co.kr/en/blog - https://www.iplexlaw.co.kr/en/contact - https://www.iplexlaw.co.kr/en/blog/845881 - https://www.iplexlaw.co.kr/en/blog/843161 - https://www.iplexlaw.co.kr/en/blog/fix-fine-grained-forgetting-attention - https://www.iplexlaw.co.kr/en/blog/fix-fine-grained-forgetting-attention - https://www.iplexlaw.co.kr/en/blog/mhar-multi-head-attention-residuals - https://www.iplexlaw.co.kr/en/blog/mhar-multi-head-attention-residuals - https://www.iplexlaw.co.kr/en/blog/claude-computer-use-agent-patent - https://www.iplexlaw.co.kr/en/blog/claude-computer-use-agent-patent