SkyVision: Screw Tightening Compliance, Enhanced Detection & Reduced Omissions

In the industrial SOP/operation compliance domain, DaoAI's SkyVision 0-code video surveillance AI platform, with its core capabilities like on-site hourly custom model training, behavior/event recognition, edge box real-time alerting, 100% on-premise data, and DaoAI World model semantic understanding, has successfully elevated screw tightening action compliance detection rates to over 99.5% and reduced omission rates to below 0.5%, providing robust quality control assurance for a major automotive component supplier's production line.

On industrial production lines, screw tightening is a critical assembly process, and its compliance directly impacts product performance, safety, and reliability. Particularly in high-value manufacturing sectors like automotive, aerospace, and precision machinery, any under-tightened, over-tightened, or missed screw can lead to severe quality incidents, even recalls. Traditional operation compliance monitoring primarily relies on manual patrols and spot checks, which are inefficient and susceptible to subjective factors, making real-time, comprehensive coverage of all critical workstations challenging. As IoT technology increasingly enables data collection, analysis, and traffic flow optimization strategies in smart city traffic management, the industrial sector urgently needs similar intelligent monitoring and data analysis methods to enhance the precision and efficiency of operation compliance. DaoAI's SkyVision 0-code video surveillance AI platform is designed to address these pain points. It performs real-time analysis of video streams using deep learning models, precisely identifying whether screw tightening actions comply with SOPs, thereby significantly boosting compliance detection rates and reducing omissions.

Pain Points: Why This Hurdle is Difficult to Overcome

In screw tightening operation compliance monitoring, enterprises commonly face multiple challenges. Firstly, high omission rates. Traditional manual patrols often only conduct spot checks, making it difficult to cover all production stages and time points, leading to a higher actual incidence of missed or incorrect tightening, with some production lines even experiencing omission rates of 1.5%–2%, severely impacting product quality. Secondly, substantial re-inspection and rework costs. If issues are discovered after products have moved downstream or even been delivered to customers, it not only requires significant human and material resources for recalls, inspection, and repair but also causes irreparable damage to brand reputation. Thirdly, high reliance on manual inspection leads to high labor costs and personnel fatigue, especially in multi-shift, high-beat production environments, continuous high-intensity visual inspection is prone to fatigue, further increasing the risk of omissions. Finally, data silo issues are prominent. Traditional solutions struggle to form structured data precipitation from the operation process, making quality issue traceability difficult and preventing effective data-driven process optimization.

The root cause of these pain points lies in the complexity and subtlety of screw tightening actions. The tightening action itself involves multiple dimensions such as tool selection, posture, force direction, and number of turns, and SOP requirements vary for different screw types and product parts. Traditional rule-based machine vision systems struggle to adapt to these varied and subtle behavior recognition needs, often requiring complex programming and debugging for each process, and are sensitive to lighting and angle changes, resulting in high false alarm and omission rates. Manual inspection, on the other hand, is limited by human physiological constraints and subjective judgment, making it difficult to achieve standardized, unbiased continuous monitoring. Similar to how smart city traffic management requires precise capture and prediction of rapidly changing vehicle behaviors, industrial SOP compliance monitoring needs an intelligent system capable of understanding “dynamic behavioral semantics.”

Technical Principles

DaoAI's SkyVision 0-code video surveillance AI platform achieves precise behavior recognition for screw tightening actions through its core DaoAI World semantic understanding capabilities, combined with advanced deep learning algorithms. The platform first uses high-resolution industrial cameras to collect production line video streams, which are then transmitted to edge boxes for real-time pre-processing. At the edge, DaoAI's self-developed lightweight behavior recognition model quickly analyzes video frames to identify key action nodes such as operator tool grasping, screw positioning, tightening start, and tightening end. Unlike traditional vision systems based on image feature matching, SkyVision's model, trained on vast amounts of industrial behavior data, understands the temporal and spatial relationships of actions, thereby distinguishing compliant from non-compliant tightening operations, such as whether the correct tool was used, whether tightening was completed, or if there were any missed tightenings. For example, the platform can identify abnormal behaviors like “wrench leaving before touching the screw” or “tightening action duration being too short,” classifying them as non-compliant operations.

Compared to traditional manual visual inspection and rule-based machine vision systems, DaoAI's SkyVision demonstrates significant advantages. Manual visual inspection is limited by eye fatigue and subjective judgment, making continuous high-precision 24/7 monitoring impossible and failing to provide quantitative data support. Rule-based machine vision systems, while automated, have poor robustness to environmental changes (e.g., lighting, occlusion, personnel position), and require extensive reprogramming and debugging whenever production SOPs or product models change, leading to high changeover costs. SkyVision platform's core advantages lie in its “0-code” and “on-site hourly custom model training” capabilities. This means users without professional AI development backgrounds can quickly train customized models for specific tightening SOPs on the production floor with minimal annotated data, achieving high adaptability and flexibility. Furthermore, the SkyVision platform's real-time alerting mechanism via edge boxes issues warnings immediately when non-compliant behavior occurs, nipping problems in the bud. The 100% on-premise data processing ensures enterprise data security and privacy.

Typical Application Scenarios

• **Under-tightening/Over-tightening Detection:** On assembly lines, DaoAI's SkyVision can analyze the trajectory, posture changes, and duration of the operator's tightening tool movements to determine if the screw has reached the specified torque range. The challenge lies in varying torque requirements for different workstations and subtle differences in tool trajectories due to operator habits. SkyVision generalizes to recognize these variations through multi-sample learning.
• **Missed Tightening Detection:** For workstations with multiple screws, the platform can identify whether the operator has completed the tightening actions for all specified screws. The difficulty arises when screws are numerous and densely located, making omissions easy. SkyVision ensures that a tightening action corresponds to each predefined screw position through area monitoring and target tracking technology.
• **Tool Usage Compliance Detection:** Certain special screws require specific tools (e.g., torque wrenches, electric screwdrivers). DaoAI's SkyVision can identify whether the operator used the correct tool for tightening. The challenge is the variety of tools and potential partial occlusion in the video. The platform uses DaoAI World model's semantic object recognition to accurately determine tool type even with partial occlusion.
• **Tightening Sequence Compliance Detection:** In some critical component assemblies, the screw tightening sequence is crucial (e.g., diagonal tightening). SkyVision can monitor whether the operator followed the preset sequence. The difficulty is real-time tracking of multiple targets and determining their operational timing. The platform precisely judges operation sequences through event chain analysis.
• **Operator Posture and Environmental Compliance:** Beyond tightening itself, DaoAI's SkyVision can also monitor whether operators are wearing necessary safety gear (e.g., gloves, goggles) or if there are foreign objects in the operating area, addressing environmental compliance issues. The challenge is that these auxiliary behaviors and environmental factors are more diverse and harder to quantify, but SkyVision's semantic understanding capabilities effectively identify them.

Case Study

A leading automotive component supplier, whose products demand extremely high fastener reliability, faced challenges. On their engine assembly line, manual patrols for critical screw tightening action compliance were inefficient and suffered from persistently high omission rates. On average, about 1.2% of product batches still had under-tightened or missed screws, leading to high rework rates and multiple customer complaints. To address this pain point, the supplier introduced DaoAI's SkyVision 0-code video surveillance AI platform. Before deployment, their team validated the platform's “on-site hourly training” capability, completing custom model training for their specific engine screw tightening SOP in just half a day. In the initial phase, SkyVision operated in parallel with manual patrols. The platform used edge boxes to real-time identify and alert non-compliant operations, such as “tool reset before contacting screw” or “tightening action not completed.” After a month of trial operation, DaoAI's SkyVision platform consistently elevated the screw tightening action compliance detection rate to over 99.5% and successfully reduced the omission rate to below 0.5%, significantly outperforming traditional manual inspection levels. Concurrently, the platform retained video evidence and timestamps for every non-compliant operation, greatly simplifying quality issue traceability and responsibility attribution, providing valuable data for subsequent process improvements.

DaoAI SkyVision transforms screw tightening compliance from a blind spot into a quantifiable, traceable intelligent monitoring object, significantly enhancing production quality and efficiency.

DaoAI Solution and Products

DaoAI's core solution for screw tightening operation compliance monitoring revolves around the SkyVision 0-code video surveillance AI platform. This platform deploys industrial-grade high-definition cameras at critical workstations along the production line to collect real-time video data. This data is then transmitted to locally deployed edge computing boxes, where AI algorithms driven by SkyVision's integrated DaoAI World model perform real-time analysis. Users do not need to write code; they can simply upload short video clips of compliant and non-compliant actions through the platform's intuitive interface to train highly customized behavior recognition models within hours. The trained models can accurately identify the compliance of various screw tightening actions, including but not limited to tool selection, tightening posture, tightening duration, tightening sequence, and whether any screws were missed. Upon detecting non-compliant behavior, the edge box immediately issues real-time alerts through various methods such as audible and visual alarms, MES system message pushes, and email notifications. All video data and analysis results are 100% stored and processed locally, ensuring data never leaves the facility, meeting stringent customer data security and privacy requirements. Furthermore, the platform provides detailed reports and analysis functions to help managers understand compliance trends, identify high-risk workstations and operators, and provide data support for continuous improvement.

Through DaoAI's SkyVision platform, customers achieve comprehensive, real-time, high-precision monitoring of screw tightening operations, increasing screw tightening action compliance detection rates to over 99.5% and reducing omission rates to <0.5%, significantly mitigating quality issues caused by human oversight. This not only lowers rework costs and customer complaint rates but also markedly improves overall production efficiency and product reliability. Concurrently, through data-driven management of the operation process, enterprises can better optimize SOPs, enhance employee training effectiveness, and ultimately achieve lean manufacturing goals. DaoAI's SkyVision offers a short deployment cycle and fast model training, enabling enterprises to achieve intelligent transformation at a lower cost and faster pace.

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.