Quantum Computing for Quality Control and Defect Detection

Workshop Description

For quality control leads, process engineers, and manufacturing R&D teams. Covers quantum machine learning approaches to automated visual inspection, sensor fusion for defect classification, and process anomaly detection on production lines. Includes honest assessments of where current NISQ hardware delivers benchmark-specific performance comparisons against classical deep learning baselines.

Modern manufacturing QC relies on classical convolutional neural networks (CNNs) for visual inspection and statistical process control for sensor-based anomaly detection. These approaches work well for high-volume, single-defect-type classification, but struggle with small training datasets, high-dimensional sensor fusion, and novel defect types that fall outside the training distribution. Quantum machine learning algorithms, specifically quantum kernel estimation (QKE) and quantum support vector machines (QSVM), offer a different approach: mapping inspection data into higher-dimensional quantum feature spaces where certain classification boundaries become more accessible. Published results from IBM Research, the University of Waterloo, and Fraunhofer show that for structured datasets with limited training examples (100-1,000 samples), quantum kernel classifiers can match or exceed classical SVMs, though they do not yet compete with large-scale CNNs trained on millions of images. This workshop maps where the crossover occurs for manufacturing QC workflows, examines the deployment architecture required, and evaluates which defect types and data characteristics are best suited to quantum approaches.

What participants cover

• Classical QC baselines: where CNN-based visual inspection and statistical process control reach their performance ceiling on novel defect types and small datasets
• Quantum kernel estimation (QKE) and QSVM for defect classification: how quantum feature maps encode inspection data for classification tasks
• Variational quantum classifiers (VQC) for multi-class defect categorisation: circuit design, training, and barren plateau mitigation
• Quantum-enhanced sensor fusion: qPCA for dimensionality reduction across vibration, thermal, and acoustic sensor arrays on production lines
• NISQ hardware limits: published benchmark-specific performance comparisons of quantum versus classical classifiers at current qubit counts and noise levels
• Deployment architecture: edge-cloud hybrid patterns for integrating quantum QC into production environments with line-speed inference requirements