Process acceptance test certificates and works certificates with KI

In industry, accurate data capture from factory certificates and acceptance test certificates is the foundation for standard-compliant quality assurance and regulatory compliance. However, the automated extraction of critical information from declarations of conformity, welding documents, material specifications and operating instructions poses a significant challenge as each supplier uses proprietary document formats. Konfuzio optimizes this process through the use of advanced artificial intelligence.

The initial situation: incoming goods inspection as a critical process step

The Incoming goods inspection is one of the most critical control points in the industry. It must be systematically checked whether all required accompanying documents are available and whether the information contained therein complies with the specifications.

Typically, the following information must be extracted and validated:

• Serial number and Batch number for clear traceability
• Order number and order reference
• Material specifications according to EN 10027 (steel designation system)
• Chemical composition with precise information on alloying elements (C, Si, Mn, P, S, Cr, Ni, Mo, etc.)
• Mechanical characteristics (tensile strength Rm, yield strength Re, elongation at break A, impact energy, etc.)
• Test certificates according to EN 10204 (type 2.1, 2.2, 3.1, 3.2)
• Heat treatment conditions and thermal parameters

Companies are faced with a fundamental dilemma: individually configured extraction processes are required for documents from individual suppliers. This results in prohibitively high maintenance costs and suboptimal extraction accuracy.

The following are particularly critical:

• Rotated scans - Error-prone processing of twisted acceptance test certificates
• Multi-page factory certificates - Insufficient processing of collective certificates with multiple batches and items
• Heterogeneous layouts - Lack of standardization between suppliers and steelworks
• Multilingual documents - Declarations of conformity in various languages according to the country of delivery

The need for a low-maintenance, high-precision solution with a standardized manufacturing process across all suppliers is evident - Konfuzio systematically addresses these requirements.

Document processing in two steps

Step 1: Document capture and normalization

In the first processing phase, Konfuzio performs a format-independent digitization:

Document classificationAutomatic identification of the document type (inspection certificate 3.1/3.2, works certificate 2.1/2.2, declaration of conformity, welding documentation, material specification, operating instructions)

Rotation detection and correctionAutomatic detection and alignment of twisted documents using computer vision algorithms - critical for fax transmissions and scanned accompanying documents

Optical Character Recognition (OCR)Use of neural networks for text recognition in scanned documents with high error tolerance, even with poor scan quality in the incoming goods department

Document normalizationGeneralization of all input documents regardless of the source format to prepare the semantic analysis

This step ensures a standardized database for subsequent contextual processing and enables integration into ERP systems without media discontinuity.

Step 2: Semantic analysis and structured data extraction

The second phase involves intelligent interpretation and structured data collection in accordance with merchandise management and quality management requirements:

Identification of critical metadataContext-based extraction of batch and serial number, order number, supplier number and order reference from variable document layouts

Material identificationAutomatic recognition and normalization of steel designations according to EN 10027, ASTM, JIS and other international standards

Understanding tablesUse of specialized algorithms for the precise interpretation of complex tables of chemical steel composition and mechanical properties - a key aspect in the verification of declarations of conformity and material specifications

Test value validationAutomatic comparison of extracted values with target specifications from order and material standards

A central innovation is the Region of Interest (ROI) Prediction Konfuzio probabilistically identifies the most probable positions of critical data fields - in particular the tables with chemical composition and mechanical parameters. Through this focused analysis, the system achieves an extraction accuracy of 99,79% with simultaneous cost efficiency.

Technology for incoming goods inspection

Multimodal Large Language Models (LLMs)These models process text and layout information simultaneously, replicating the interpretation skills of experienced incoming goods inspectors and quality engineers.

Adaptive machine learning algorithms: Through continuous learning, Konfuzio processes diverse document formats and dynamically adapts to new suppliers and document variations without the need for time-consuming retraining.

Domain-specific trainingSpecialization in metallurgical terminology, standard references and industry-specific document structures.

Semantic validationPlausibility check of extracted values based on metallurgical laws (e.g. sum of alloying elements, carbon equivalent).

Advantages of AI-supported document processing in incoming goods

Extraction accuracyKonfuzio achieves an accuracy of 99.79% in recording critical parameters such as batch number, chemical composition and mechanical characteristics - minimizing process errors and enabling consistent quality control.

Acceleration of the incoming goods inspectionAutomation of manual processing steps reduces throughput times from several hours to just a few minutes, minimizing idle times in the incoming goods department.

Complete traceabilityAutomatic linking of serial number, batch number and order number enables complete traceability in accordance with ISO 9001 and industry-specific requirements.

ScalabilityA standardized processing procedure for all suppliers and steelworks simplifies system maintenance and enables seamless integration of new document formats without reconfiguration.

Compliance securityComplete documentation and traceability fulfill regulatory requirements, especially for declarations of conformity and welding documentation.

Cost reductionElimination of manual data entry and reduction of errors in downstream processes (production, quality assurance, Audit).

Implementation and integration into existing processes

Konfuzio integrates seamlessly into existing ERP systems (SAP MM, Oracle Supply Chain, Microsoft Dynamics) and quality management platforms (SAP QM, QSYS, CAQ systems).

The solution supports various integration scenarios:

• Automated processing incoming e-mail attachments with acceptance test certificates
• Batch processing scanned accompanying documents from the incoming goods department
• Real-time integration via REST-API for immediate availability in the ERP system
• EDI connection for structured data exchange with suppliers

Containerized deployment options (Docker, Kubernetes) allow for flexible On-premise or cloud installation data protection requirements for sensitive supplier information are taken into account.

Conclusion: end-to-end digitalization of incoming goods inspection

Konfuzio fundamentally transforms the processing of factory certificates, inspection certificates, declarations of conformity and welding documents by combining robust document capture with advanced semantic understanding of specialized metallurgical content.

The extraction accuracy of 99.79% achieved for critical parameters such as serial number, batch number and chemical composition marks a paradigm shift in industrial document processing and enables end-to-end digital incoming goods inspection.

The intelligent application of state-of-the-art AI technology outperforms traditional extraction methods in all relevant dimensions:

• Quality - Virtually error-free data extraction even with complex table structures
• Implementation speed - Reduction of time-to-value by up to 70%
• Maintenance effort - Minimal maintenance through self-adaptive algorithms
• Process integration - Seamless connection to existing incoming goods and QM processes

Konfuzio enables companies to automate their incoming goods inspection while ensuring the highest quality standards and complete traceability - a decisive competitive advantage in an increasingly digitalized and regulated industry.