On-Premise Paper Surface Micro-Defect Detection & Data Security

Addressing the stringent data security and privacy demands in the chemical/materials industry, DaoAI 2D ACI equipment, through 100% on-premise private deployment, has resolved the challenge of online full inspection for complex micron-level surface defects such as scratches, stains, and foreign matter in paper production. Following its deployment at a leading paper group, production line false positive rates decreased by 75%, while ensuring core process data remained in-house, effectively supporting a zero-defect manufacturing strategy.

DaoAI 2D ACI equipment (high-resolution 2D imaging + deep learning secondary judgment, for surface/print/OCR/assembly defects, high-speed online full inspection, micron-level, semantic false positive filtering), through 100% on-premise private deployment, reduced the false positive rate for paper surface micro-defect detection at a major paper group from approximately 4% to less than 1%, significantly improving inspection efficiency and data security. The chemical and materials industry, particularly paper manufacturing, demands extremely high product quality. During paper production, various surface defects such as fine scratches, stains, fiber clumps, bubbles, foreign matter, and creases can easily occur across multiple stages from pulping, papermaking, pressing, drying, calendering, to winding. Although these defects are minute, they severely impact the paper's final performance, printability, and end-product appearance, causing ripple effects for downstream packaging, printing, and special materials applications. Traditionally, the detection of such defects primarily relied on manual visual inspection or rule-based AOI equipment. However, as production line speeds increase and defect complexity grows, the limitations of traditional solutions are becoming increasingly evident.

Pain Points: Why This Hurdle Is So Difficult to Overcome

Detecting micro-defects on paper surfaces presents multiple challenges in paper production. First, the diverse types and varied forms of defects, ranging from micron-sized fiber clumps to millimeter-sized oil stains, make it difficult for manual visual inspection to achieve stable, efficient, and comprehensive coverage, especially on high-speed production lines. This results in persistently high missed detection rates; production line data from a paper group showed manual inspection's missed detection rate once exceeded 1.5%. Second, traditional rule-based AOI equipment is highly sensitive to lighting, paper texture, and background noise, struggling to adapt to minor fluctuations in paper raw materials and process parameters. This leads to high false positive rates; even after introducing rule-based AOI, the group's false positive rate remained around 4%, requiring extensive manual secondary review of 'false defects' and significant labor waste. Third, changeover downtime is lengthy. When paper types and specifications change frequently, traditional AOI requires several hours for reprogramming and calibration, severely impacting production efficiency. Finally, and most critically, for large chemical and materials enterprises, core data such as production data, process parameters, and even defect images are vital assets and competitive advantages, making data leakage absolutely unacceptable. While traditional cloud-based AI vision solutions offer superior performance, their data upload-to-cloud model conflicts with strict enterprise data security policies, putting companies in a dilemma when choosing advanced technologies.

The root cause of these difficulties lies in the inherent complexity of paper material itself: its heterogeneous fiber structure, variable surface texture, and environmental interference that is difficult to completely eliminate during production make it exceptionally challenging to distinguish defect signals from normal backgrounds. Traditional vision algorithms struggle to effectively model such highly variable backgrounds, while detection relying on human experience cannot meet the demands for high speed, high precision, and round-the-clock operation. Concurrently, data security is a lifeline for enterprises; any potential data leakage risk is enough to deter companies from adopting external solutions, even if they significantly improve production efficiency. This highlights why, in the process of AI vision inspection technology serving zero-defect manufacturing, it is crucial to deeply integrate with industry characteristics and enterprise security requirements, in addition to the technology itself.

Technical Principles

The core advantage of DaoAI 2D ACI equipment lies in its combination of high-resolution 2D imaging technology with advanced deep learning secondary judgment capabilities. At the imaging level, we utilize customized high-speed industrial cameras and a multi-spectral lighting system to capture micron-level details on the paper surface, effectively suppressing interference from paper texture and background noise. By optimizing lighting angles and wavelengths, DaoAI 2D ACI can enhance the contrast between defects and normal backgrounds, making even transparent scratches or low-contrast stains clearly visible. At the algorithm level, DaoAI 2D ACI is equipped with the DaoAI ACI OS operating system, whose built-in visual foundation model performs feature recognition. Through APDT positive/few-shot learning technology, it requires only 1–20 good samples to achieve 0-code automatic programming within 5 minutes, rapidly establishing a detection model. This model focuses on learning the normal feature distribution of 'good products'; any regions deviating from this distribution are initially marked as potential defects. Subsequently, a deep learning secondary judgment module intervenes, using semantic false positive filtering technology to re-analyze the initially marked defects, distinguishing true defects from common 'false defects' (such as water marks, slight fiber lifting, etc., which are normal product features) that occur during production, thereby significantly reducing the false positive rate. Unlike traditional rule-based AOI that relies on engineers manually writing complex rules, DaoAI 2D ACI's deep learning model can automatically learn and generalize from vast amounts of data, adapting to changes in paper types and production environments, significantly improving detection robustness and accuracy.

Compared to traditional manual visual inspection, DaoAI 2D ACI achieves 24/7, high-speed, objectively consistent detection, eliminating errors caused by human eye fatigue and subjective judgment. In contrast to traditional rule-based AOI, DaoAI 2D ACI's deep learning model possesses stronger generalization capabilities and anti-interference, especially when dealing with complex and varied paper surface defects. Its semantic false positive filtering function can effectively identify and ignore non-defect features, thus keeping the false positive rate at an extremely low level. Furthermore, DaoAI 2D ACI equipment supports 100% on-premise private deployment, meaning all data processing and model training are completed within the customer's factory, with no data leaving the premises. This fundamentally resolves data security and privacy protection issues, an unmatched advantage over traditional cloud-based solutions, and a key factor in its widespread adoption within the chemical/materials industry.

Typical Application Scenarios

• **Paper Surface Scratch Detection:** During calendering, winding, and other processes, fine scratches can easily appear on the paper surface. DaoAI 2D ACI equipment, through high-resolution imaging and deep learning, can identify scratches down to micron-level widths, distinguishing between actual damage and light refraction caused by surface texture. The challenge lies in distinguishing scratches from paper fiber directions.
• **Stain and Foreign Matter Detection:** Oil stains, pulp residues, ink spots, or tiny foreign objects may be introduced during production. DaoAI 2D ACI comprehensively judges these defects based on features like color, shape, and grayscale, accurately detecting them. The difficulty lies in low-contrast stains and foreign objects similar in color to the paper.
• **Fiber Clump and Bubble Detection:** Uneven fiber distribution in pulp or improper papermaking can lead to fiber clumps, and bubbles may form during drying. DaoAI 2D ACI utilizes its multi-scale feature extraction capabilities to identify the 2D image projections of these 3D morphological features. The challenge is distinguishing small clumps and bubbles from normal paper roughness.
• **Crease and Fold Detection:** Winding and slitting processes can cause creases or folds in the paper. DaoAI 2D ACI quickly locates these structural defects by analyzing local light and shadow changes and texture discontinuities. The difficulty is with slight and discontinuous creases.
• **Size and Position Deviation:** For special-purpose papers (e.g., printing paper), it may be necessary to inspect edge integrity, cutting dimensions, or position deviations of pre-printed patterns. DaoAI 2D ACI can integrate OCR/OCV functions for high-precision size measurement and pattern alignment detection. The challenge lies in real-time sub-millimeter precision measurement in high-speed online scenarios.

Case Study

A leading paper group, with production lines annually yielding millions of tons of various paper types, has stringent dual requirements for product quality and data security. Previously, the group primarily relied on manual visual inspection combined with some rule-based AOI for paper surface defect detection on high-speed production lines. However, with increasing order volumes and product diversity, the fatigue associated with manual inspection made it difficult to keep the missed detection rate below 1.5%, while rule-based AOI suffered from a false positive rate of around 4% for complex defects, requiring significant daily manual re-inspection and severely impacting production efficiency and costs. More importantly, strict internal group policies mandated that core production data and defect images must be processed locally and not uploaded to the cloud. To address these challenges, the group introduced DaoAI 2D ACI equipment, specifically opting for a 100% on-premise private deployment solution.

The DaoAI team conducted system integration and model training at the client's site for two weeks. Utilizing the APDT few-shot learning function of DaoAI ACI OS, an initial model was automatically programmed within 5 minutes, using fewer than 20 good samples. After deployment, DaoAI 2D ACI equipment continuously monitored the paper production line, performing high-speed, full inspection of micron-level surface defects. Production line data showed that the system consistently maintained the missed detection rate for paper surface micro-defects at <0.3%, significantly lower than the manual inspection level of over 1.5%. Concurrently, with the support of deep learning secondary judgment and semantic false positive filtering, the false positive rate drastically decreased from approximately 4% before deployment to less than 1%, demonstrating a measured 75% reduction in false positives, which significantly reduced the workload for manual re-inspection. Crucially, all inspection data, images, and model training processes were completed on the client's in-house servers, fully satisfying the group's data security and privacy protection requirements, providing a solid guarantee for the client's smart manufacturing upgrade.

“DaoAI 2D ACI's on-premise deployment solution not only resolved our production line's long-standing issues of high false positive rates and difficult-to-control missed detection rates, but more importantly, it completely eliminated our data security concerns, allowing us to confidently embrace cutting-edge AI vision technology.” — Production Manager, a leading paper group.

DaoAI Solution and Products

The DaoAI 2D ACI equipment, with its outstanding performance and flexible deployment options, is an ideal choice for paper defect detection in the chemical/materials industry. The core of this solution lies in the DaoAI ACI OS operating system, which enables non-specialized personnel to quickly build and optimize detection models using 0-code APDT positive/few-shot learning technology. In paper defect detection scenarios, by collecting a small number of good sample images (typically 1-20), the system automatically learns normal paper textures and features, and identifies abnormal areas. For deployment, DaoAI 2D ACI equipment supports various integration methods such as SDK/API/Docker, and offers 100% on-premise private deployment. This ensures all data processing is completed within the client's factory, without uploading to any external servers, maximizing data security. Furthermore, the semantic false positive filtering function of DaoAI 2D ACI equipment effectively distinguishes between normal paper texture variations and true defects, greatly enhancing detection accuracy and efficiency, and reducing the burden of manual re-inspection. Coupled with the unified foundation of DaoAI World model, future capabilities include cross-scenario generalization and continuous learning from production line feedback, further elevating the level of intelligent detection.

Through the deployment of DaoAI 2D ACI equipment, the client achieved a leap from traditional manual/rule-based AOI to intelligent AI vision inspection. This solution not only significantly improved the accuracy and efficiency of paper surface defect detection, reducing the missed detection rate to <0.3% and the false positive rate by 75%, but also greatly shortened changeover downtime. Production line data from a paper group showed that changeover time was reduced from several hours to within 5 minutes. More importantly, the DaoAI 2D ACI solution, through 100% on-premise private deployment, ensured the absolute security of the client's core production data, eliminating enterprise concerns when adopting advanced AI technology. This allowed the client to focus on product quality improvement and production efficiency optimization without worrying about data leakage risks, thereby achieving zero-defect manufacturing strategic goals and establishing a sustainable competitive advantage for the enterprise.

Full solution for this scenario: the full inspection solution for 2D ACI Equipment

This article was generated by AI. Customer cases are simulated scenarios based on real product capabilities and figures are illustrative; see product pages for official benchmarks.