SkyVision On-Premise Deployment Reduces Assembly Omission Defects

DaoAI SkyVision 0-code video surveillance AI platform (on-site hourly training of proprietary models, behavior/event recognition, edge box real-time alerting, 100% localized data, DaoAI World semantic understanding) leveraged its robust on-premise deployment capability to significantly reduce bolt omission defect escape rate from a baseline of 2.8% to below 0.3% for a mid-sized automotive parts manufacturer's assembly process, effectively enhancing product quality and compliance. In today's industrial landscape, where data sovereignty and privacy are paramount, ensuring sensitive production data remains in-house has become a primary consideration for enterprises selecting smart manufacturing solutions.

In the realm of Industrial SOP / Work Instructions, ensuring every step in the production process strictly adheres to standards is crucial for guaranteeing product quality and production efficiency. Especially in complex assembly procedures, even minor omissions or errors can lead to decreased product performance, increased rework costs, and even safety hazards. Traditional quality inspection methods, such as manual visual inspection or sampling, are often limited by human eye fatigue, subjective judgment, and the inherent limitations of sampling, making it difficult to achieve 100% real-time monitoring of all critical steps. As the effectiveness of AI surveillance cameras in assisting law enforcement in urban public safety becomes increasingly prominent, the industrial sector is also actively exploring their immense potential in production line compliance checks, particularly how to meet enterprises' stringent requirements for data security and privacy through on-premise deployment. A mid-sized automotive parts manufacturer faced challenges with bolt omission in the assembly of door lock modules, directly impacting the reliability of the locks and overall vehicle safety. DaoAI SkyVision platform provided an end-to-end solution for this client, effectively addressing this pain point.

Pain Points: Why This Challenge Was Difficult to Overcome

For this manufacturer, the escape rate of bolt omissions in the assembly of door lock modules consistently hovered between 2.5% and 3.0%, an unacceptable level in the automotive parts industry. Traditional manual visual inspection, due to fast production cycles and prolonged repetitive tasks for workers, led to visual fatigue and attention lapses, resulting in persistently high escape rates. Furthermore, each time a bolt omission defect was found, manual intervention for rework was required, averaging about 15 minutes per rework incident, significantly dragging down overall production line efficiency. More critically, if defective products entered the market, they could trigger serious quality incidents and recall risks, causing immeasurable damage to brand reputation and economic benefits. From a process perspective, the bolts' color, size, and surrounding components were similar, and their assembly positions were relatively concealed, increasing the difficulty of visual identification. Existing traditional rule-based AOI systems, lacking semantic understanding of complex scenarios, struggled to differentiate subtle differences between installed and uninstalled bolts, leading to high false positive rates and numerous unnecessary line stoppages and manual re-inspections, further exacerbating the production burden. These issues compelled the enterprise to seek an intelligent solution that offered both precise detection and assured data security.

Delving into the root causes, it primarily stemmed from the limitations of human visual recognition and the shortcomings of traditional machine vision technology. On a high-speed assembly line, workers need to focus on multiple details simultaneously, and bolt omission is often a momentary oversight. Moreover, different batches of bolts might have slight variations in luster, or appear differently under varying lighting conditions, all posing challenges for human judgment. While traditional rule-based AOI systems can perform simple feature matching, their ability to determine whether a bolt is 'truly' tightened in place, or if it's obscured by other components in complex situations, is very limited, making them highly prone to false positives or negatives. This technological bottleneck, coupled with the company's strict policy that core production data must never leave the factory, made finding a solution that was both intelligent and secure particularly difficult.

Technical Principles

DaoAI SkyVision platform, through its core 0-code video surveillance AI technology, combined with edge box deployment and DaoAI World semantic understanding capabilities, provided a breakthrough solution for the automotive parts manufacturer. The platform first utilizes industrial cameras to capture production line video streams, which are then pre-processed and inferred in real-time by edge boxes. Its core lies in employing advanced deep learning algorithms, particularly models optimized for behavior and event recognition. Unlike traditional rule-based machine vision systems, DaoAI SkyVision can quickly train proprietary AI models for specific omission defects on-site within hours, using a small number of samples (even 1-20 good samples). This model learns and understands the visual characteristics of bolts in a correctly installed state, as well as the missing patterns in an omitted state, thereby achieving high-precision defect detection. Furthermore, the DaoAI World model provides powerful semantic understanding, allowing the system to not only identify 'whether a bolt is missing' but also comprehend deeper behavioral semantics such as 'whether a bolt installation action has occurred,' effectively filtering environmental interference and occasional false positives, enhancing detection robustness. Real-world data showed that DaoAI SkyVision reduced the bolt omission escape rate to below <0.3%, significantly outperforming traditional methods.

Compared to traditional manual visual inspection, DaoAI SkyVision achieves 24/7 uninterrupted 100% full inspection, completely eliminating errors caused by human eye fatigue and subjective judgment. Its detection accuracy and consistency far surpass human capabilities. Compared to traditional rule-based AOI systems, SkyVision's advantage lies in its '0-code' self-learning capability and adaptability to complex, variable scenarios. Traditional rule-based AOI requires engineers to write a large number of complex rules and thresholds, making it highly sensitive to changes in lighting, product deformation, etc. Any minor adjustment to the production line or product necessitates reprogramming, which is time-consuming and labor-intensive, often resulting in high false positive rates. In contrast, DaoAI SkyVision learns patterns from real data, maintaining high accuracy even under complex working conditions such as lighting variations or slight component movement. Crucially, its 100% localized data, never leaving the factory, completely alleviates enterprise concerns about data security and privacy, which is vital in industries like automotive parts manufacturing with extremely high demands for core technology and production data protection.

Typical Application Scenarios

• **Multi-component Assembly Sequence Detection:** On complex product assembly lines, ensuring components are installed in the correct sequence is critical to preventing functional defects. DaoAI SkyVision can monitor worker operation sequences in real-time, identifying any SOP violations (e.g., installing B before A) and issuing immediate alerts. The challenge lies in distinguishing subtle differences between components and the temporal aspects of operations.
• **Critical Fastener Omission/Misplacement Detection:** Beyond bolt omissions, in the assembly of core automotive components like engines and transmissions, omitted or misplaced critical fasteners such as nuts, washers, and clips can lead to catastrophic consequences. SkyVision can accurately identify the presence or absence of these tiny components, as well as their correct model (misplacement), effectively preventing quality issues caused by human error. The challenge lies in the small size, large number, and concealed assembly positions of these components.
• **Cable Harness Connector Insertion Detection:** In electronic products and automotive harness assembly, whether connectors are fully inserted directly impacts circuit continuity and signal transmission quality. DaoAI SkyVision can determine if connectors are correctly installed by identifying specific visual features in their fully inserted state (e.g., buckle closure, color mark alignment). Challenges include the wide variety of connectors, dynamic insertion processes, and varying visual features from different angles.
• **Label/Marking Application Compliance Detection:** On product packaging or component surfaces, the placement, orientation, flatness, and absence of bubbles on labels affect product appearance and information readability. SkyVision can detect label application quality to ensure compliance with design specifications. Challenges include reflective label materials, uneven surfaces, and high sensitivity requirements for tiny defects.

Case Study

A mid-sized automotive parts manufacturer, specializing in door lock module production, had a long-standing bolt omission escape rate of around 2.5% on its assembly line before adopting DaoAI SkyVision platform. This not only resulted in high rework costs but also posed potential quality recall risks. The client had extremely strict data security requirements, explicitly demanding that all production monitoring data be processed 100% locally, with no data uploaded to external cloud servers. Addressing this need, DaoAI provided an on-premise private deployment solution for SkyVision. In the initial phase, engineers deployed industrial-grade surveillance cameras on the production line, connected to local edge boxes. By collecting a small number of image samples of both installed and uninstalled bolts, and utilizing SkyVision's 0-code training interface, the first AI model for bolt omission detection was trained and deployed in just 3 hours. Production line data showed that after the system went live, the bolt omission escape rate significantly decreased from 2.8% to below 0.3%, and the false positive rate was reduced by approximately 75%, greatly easing the burden of manual re-inspection. Concurrently, with all data processed locally, the client successfully avoided data security risks, meeting its stringent compliance requirements. Through the deployment of DaoAI SkyVision, the client's production line efficiency improved by approximately 8%, and product quality stability was significantly assured.

DaoAI SkyVision's localized deployment not only solved our quality pain points but also completely eliminated our data security concerns, which was a key reason for our choice.

DaoAI Solutions and Products

DaoAI SkyVision 0-code video surveillance AI platform is the core of this solution. The platform's primary advantage lies in its '0-code' ease of use, enabling production line engineers, without specialized AI programming knowledge, to quickly train and deploy AI models on-site via an intuitive graphical interface. For assembly omission detection, SkyVision employs behavior/event recognition technology to accurately capture key actions and the final state during the bolt installation process. Specific implementation involves: first, installing industrial-grade HD cameras above critical workstations to collect real-time video streams; second, feeding these video streams into locally deployed edge boxes, where SkyVision runtime performs real-time inference. The modeling process is highly efficient; using DaoAI APDT positive/few-shot learning technology, only 1-20 good product images showing correct assembly are needed for the system to complete model training and automatic deployment within hours. If the production line undergoes changeover or minor process adjustments, only a few new samples are required to update the model within 5 minutes. DaoAI World model, serving as a unified foundation, provides SkyVision with powerful semantic understanding and cross-scenario generalization capabilities, ensuring high robustness in complex industrial environments. All data processing and model inference are completed on the client's local servers or edge devices, strictly adhering to the 100% on-premise private deployment principle, ensuring absolute security of production data and fully meeting the client's strict 'data never leaves the factory' requirement.

DaoAI SkyVision platform not only detects assembly omissions but also features real-time alerting. Upon detecting an anomaly, it can instantly notify relevant personnel via sound and light, email, or integration with the MES system. This real-time feedback mechanism significantly shortens defect response time, preventing defective products from flowing to the next process. In this case, the deployment of DaoAI SkyVision reduced the bolt omission escape rate from 2.8% to below 0.3%, and simultaneously cut manual re-inspection time by approximately 75%, greatly enhancing production line quality control and efficiency. Furthermore, by eliminating data leakage risks, this solution established a solid foundation of digital trust for the enterprise, allowing it to more confidently pursue intelligent manufacturing transformation. The solution's return on investment period is short, typically achieving cost recovery within 6-12 months, bringing significant business value to the enterprise.

Full solution for this scenario: SkyVision Video AI industry solutions

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.