Success Story
Global Renowned Precision Manufacturer Improves Product Quality with HOP AI Poka-Yoke
Background
A global leader in the semiconductor industry with over 25 years of extensive manufacturing history, this company specializes in precision machining and provides critical components for the high-tech sector, including reticle box, wafer handing, substrate shipping box, and similar products. The manufacturer is dedicated to integrating smart manufacturing into its global production network. Through substantial investments in automation equipment, data analytics tools, and artificial intelligence technologies, they have transformed traditional production lines into more efficient, flexible, and intelligent manufacturing systems.
For the reticle pod assembly workstation, the company has implemented the PowerArena HOP (human operation platform) to help review manual processing steps and further optimize production processes.
How to Solve SOP Optimization Challenges?
| Engineer from the Engineering Department: “We overlooked the blind spots in SOP design.“
Industrial Engineers (IE) have discovered that when the SOP is executed on the production line, there is often a gap between the planned and actual performance. This discrepancy arises from new employees being unfamiliar with the operations and lacking mature skills, as well as seasoned workers having ingrained habits. Even if the IE provides the same SOP, the performance of each operator can vary significantly, leading to inconsistencies in assembly quality and difficulties in management.
Introducing AI Vision, Advancing AI SOP
Before the project began, PowerArena engaged in multiple discussions with the plant manager and IE engineers to establish a more detailed and comprehensive SOP. Standardizing the operational procedures once again was not only for training and recognizing the AI models but also to help the production line reevaluate the processes from a human factors engineering perspective, thereby increasing the potential for accelerated production. The more detailed SOP includes specifications for the placement of materials and tools, the distances between them, and considerations for the different operating habits of left-handed and right-handed workers.
Is Production Data Always Insufficient for IE?
| Engineering Department Manager: “Why was it possible before, but not now?”
The IE shared that in the past, when supervisors questioned the legitimacy of improvement projects, the IE lacked data and video evidence to support their case. When abnormalities occurred, operators typically “made up for the errors with time,” but since these extra efforts were not recorded continuously, the additional time spent could not be traced.
AI Data + Video: Evidence for Production Line Optimization
Cameras are deployed at workstations to enable real-time AI visual analysis. HOP accurately identifies cycle time (CT) and records all 20 steps of assembling reticle pod covers, collecting production time data continuously. This replaces time-consuming manual measurements, eliminates human errors, and frees up management resources, allowing IE to handle other tasks without being burdened by these essential measurements.
With video records, IE possesses concrete evidence of “waste” occurrences, invaluable when demonstrating anomalies to line managers, production engineers, or operators. The combination of data and video allows IE to address supervisors’ concerns and justify the need for changes.
During rework or improvement processes, video records ensure progress and effectiveness are traceable, and the causes of anomalies are always identifiable. IE can pinpoint error sources precisely and propose concrete solutions, ensuring the improvement plans are executed effectively.
How to Overcome Veteran Habits and Newcomer Errors?
At this assembly station, managers found that experienced operators tend to rely on old habits, while new employees often make mistakes or forget steps. Variations in manual operations are hard to prevent, but deviating from the SOP can lead to defects.
AI Real-Time Alerts for Immediate Corrections
HOP provides real-time AI alerts to serve as the first line of defense. If an operator pauses too long, an alert sound reminds them, and an anomaly message is sent to managers for immediate attention.
HOP can also filter problematic video clips, reducing the need for manual review of extensive production footage, saving valuable time. By marking problem areas with AI alerts, IE can easily trace and understand the root causes of errors, allowing them to focus on addressing the underlying issues leading to anomalies.
High Costs of Operator Training and Allocation?
Operators at this station sometimes need to temporarily support other production lines. When they return, they may forget previous steps. The time spent re-learning becomes a hidden cost, slowing overall production efficiency.
Digital Prompts for Enhanced Operations
If operators forget the steps after pausing, they can check a nearby screen with HOP’s digital prompts to confirm the next action, ensuring correct continuation of assembly.
If the station’s monthly output increases and operators from other lines are brought in to help, the digital SOP aids in quickly familiarizing them with assembly tasks.
This is especially useful for expanding new lines and training new employees. Previously, assembly training took about half a day. With HOP’s assistance, training time can be reduced to under two hours, allowing new employees to integrate into the production line faster, improving overall quality and efficiency.
Results
19% Increase in UPH
In six months, the average assembly cycle time (CT) at this station improved from 3.5 minutes to 2.8 minutes, with the fastest time reaching 1 minute. Overall, UPH increased by 19%.
AI data-driven process optimization helped achieve this efficiency boost. HOP collected and analyzed extensive operation times, integrating key production metrics such as product serial numbers, machine numbers, OK/NG status, errors and reasons. This aids managers in planning and decision-making. Additionally, AI-tagged problem videos help accurately identify bottlenecks, improving employee training and issue resolution, thereby increasing UPH.
Maintaining a Yield Rate of 95% and First Pass Yield of 97.6%
For precision manufacturing industries that demand high-quality standards, maintaining yield rates is challenging due to complex processes and delicate craftsmanship. Even minor defects can render a product non-compliant, with little room for error. Any lapse in detail control can affect brand reputation.
With HOP, confidence and ease in maintaining yield rates significantly increase for line managers. Quality control manpower is released, allowing time to be spent on more valuable strategic planning. This station consistently maintains a yield rate of 95% and a first pass rate of 97.6%. Assembly errors, such as missing parts, are virtually eliminated. AI vision assists in monitoring the production line 24/7, instantly alerting anomalies and providing real-time access to images and messages, enabling managers to understand production status, correct errors promptly, reduce rework and waste, and optimize work without being restricted by data gaps. They can focus on anomaly resolution without worries, letting IE optimize work seamlessly.
Fully visualizing the production line paves the way for complete automation and expansion.
The factory is transitioning towards fully automated production. Machines handle tasks such as cleaning, packaging, and measurement, but assembly, relying on precision and flexibility, still requires manpower. HOP provides detailed insight into assembly stations, serving as a crucial tool before full automation. It helps managers better understand production conditions.
After creating a model production line with HOP, manufacturers can swiftly replicate standardized production environments. From station spacing to tool placement and facility layout, HOP ensures rapid setup of new production lines in new factories, maintaining portability and seamless production assistance. Even in unalterable environmental differences like ceiling height or lighting color, HOP adapts by incorporating new data and retraining models, further enhancing system adaptability.