What is the difference between anomaly detection and statistical process control (SPC)?
SPC is a traditional statistical approach. Anomaly detection is a machine learning approach.
Anomaly Detection
Anomaly detection implementation services for pharmaceutical manufacturers. The engagement covers the use case identification, the data preparation, the model development, the model deployment, the integration with the existing systems, and the validation.
Process anomaly detection
Process anomaly detection covers the identification of the unusual process patterns, the deviation from the normal operating range, and the early warning of the process issues.
Quality anomaly detection
Quality anomaly detection covers the identification of the quality issues, the deviation from the quality specification, and the early warning of the quality problems.
Model development
The anomaly detection model development covers the data preparation, the feature engineering, the model selection, the training, the validation, the testing, and the documentation.
How to use this page
Use this Anomaly Detection page as a planning checkpoint before vendor selection, architecture review, validation scoping or implementation sequencing. The strongest next step is to compare the guidance with your current SOPs, system inventory, batch records, data flows and QA review routines so the discussion starts from evidence instead of assumptions.
Evidence to prepare
For Anomaly Detection, prepare the records, owners, risks and decision criteria linked to process anomaly detection, quality anomaly detection, model development. Useful evidence includes current process maps, interface lists, audit trail expectations, exception workflows, data retention rules and the business reason for changing the current operating model.
Frequently asked questions
How is the anomaly detection model validated?
The anomaly detection model is validated per the GAMP 5 framework for AI/ML systems.