Interpretable Self-Supervised Learning for Fault Identification in Printed Circuit Board Assembly Testing

Fault identification in Printed Circuit Board Assembly (PCBA) testing is essential 1 for assuring product quality, nevertheless, conventional methods still have difficulties due 2 to the lack of labeled faulty data and the "black-box" nature of advanced models. This 3 study introduces a label-free, interpretable self-supervised framework that uses two pretext 4 tasks: (i) an autoencoder (reconstruction error and two latent features) and (ii) Isolation 5 Forest (faulty score) to form a four-dimensional representation of each test sequence. A 6 two-component Gaussian Mixture Model is used, and the samples are clustered into nor- 7 mal and fault groups. The decision is explained with cluster-mean differences, SHAP 8 (LinearSHAP or LinearExplainer on a logistic-regression surrogate), and a shallow decision 9 tree that generated if–then rules. On real PCBA data, internal indices showed compact and 10 well-separated clusters (silhouette 0.85, Calinski–Harabasz 50,344.19, Davies–Bouldin 0.39), 11 external metrics were high (ARI 0.72; NMI 0.59; Fowlkes–Mallows 0.98), and the clustered 12 result used as a fault predictor reached 0.98 accuracy, 0.98 precision, and 0.99 recall. Expla- 13 nations show that the IForest score and reconstruction error drive most decisions, causing 14 simple thresholds that can guide inspection. An ablation without the self-supervised tasks 15 results in degraded clustering quality. The proposed approach offers accurate, label-free 16 fault prediction with transparent reasoning and is suitable for deployment in industrial 17 test lines.