2K Robotic 3D Concrete Printing Sensor Dataset for Process Anomaly Detection

This dataset contains synchronized sensor data collected from a dual-material (“2K”) robotic 3D concrete printing system at the University of Michigan’s Robotics Department. The data support the study “A Framework for Anomaly Detection in 3D Concrete Printing”. All signals were sampled at 2 Hz during full-scale printing experiments designed to characterize process dynamics under healthy, clogging, and segregation conditions. The file mmri2024.csv includes raw time-series measurements from multiple sensors—water temperature, mortar temperature, air temperature, pump pressure, and motor power—along with a label column that specifies the experimental test identifier (e.g., test1, test9). Because these identifiers do not represent class labels, two companion files are provided: mmri2024_label_map.csv — maps each test ID to its condition (healthy, clog, segregation) and subsystem (pump, head, full system). mmri2024_manual_labels.json — records original manual annotations indicating whether a test was considered “healthy” and/or part of the validation set. Note: In the original manual labels, test 4 was marked as healthy, but the anomaly-detection framework later identified it as anomalous, and post-experiment review confirmed this reassessment. The manual label file retains the original annotation for transparency. The dataset enables reproducible research in sensor-based monitoring, anomaly detection, and control for extrusion-based additive manufacturing. It is released under a Creative Commons Attribution 4.0 International (CC BY 4.0) license.