Crimp Force Curve Dataset

The "Crimp Force Curve Dataset" is a comprehensive collection of univariate time series data representing crimp force curves recorded during the manufacturing process of crimp connections. This dataset has been designed to support a variety of applications, including anomaly detection, fault diagnosis, and research in data-driven quality assurance. A salient feature of this dataset is the presence of high-quality labels. Each crimp force curve is annotated both by a state-of-the-art crimp force monitoring system - capable of binary anomaly detection - and by domain experts who manually classified the curves into detailed quality classes. The expert annotations provide a valuable ground truth for training and benchmarking machine learning models beyond anomaly detection. The dataset is particularly well-suited for tasks involving time series analysis, such as training and evaluating of machine learning algorithms for quality control and fault detection. It provides a substantial foundation for the development of generalisable, yet domain-specific (crimping), data-driven quality control systems. The data is stored in a Python pickle file crimp_force_curves.pkl, which is a binary format used to serialize and deserialize Python objects. It can be conveniently loaded into a pandas DataFrame for exploration and analysis using the following command: df = pd.read_pickle("crimp_force_curves.pkl") The DataFrame consists of 2,439 rows (force curves) and 11 columns: CrimpID: Unique identifier assigned by the crimp force monitoring system as integer. Wire_cross-section_conductor: Wire cross-section of the conductor as float [mm²]. Force_curve_raw: Raw force curve with 3,566 datapoints as NumPy array with integer values [Force sensor value]. Force_curve_baseline: Baseline curve for comparability as NumPy array with integer values [Force sensor value]. Force_curve_RoI: Region of interest curve for crimp quality evaluation with 500 datapoints as NumPy array with integer values [Force sensor value]. Main_label_string: Quality class (OK, Missing Strands, or Crimped Insulation) manually assigned by the authors based on the preparation steps, as strings. Main_label_encoded: Encoded quality classes (OK, Missing Strands, or Crimped Insulation) as integers from 0 to 2. Sub_label_string: Fine-grained quality class (OK, One Missing Strand, Two Missing Strands, Three Missing Strands, or Crimped Insulation) manually assigned by the authors based on the preparation steps, as strings. Sub_label_encoded: Encoded sub-classes (OK, One Missing Strand, Two Missing Strands, Three Missing Strands, Crimped Insulation) as integers from 0 to 4. Binary_label_encoded: Binary encoded quality classes (OK, NOK) for anomaly detection, manually assigned by the authors as integers from 0 to 1. CFM_label_encoded: Binary encoded quality classes (OK, NOK) assigned by the crimp force monitoring system as integers from 0 to 1. This dataset is a valuable resource for researchers and practitioners in manufacturing engineering, computer science, and data science who are working at the intersection of quality control in manufacturing and machine learning.