Traditional aspherical lens inspection technology initially emerged and evolved primarily in cutting-edge fields such as defense, aerospace, and high-end semiconductor lithography systems. In these traditional applications, the industry didn't have extremely stringent requirements for inspection efficiency. The reason is simple: the production logic was "single piece" or "several pieces."
To pursue extremely high absolute precision, engineers were willing to sit on benches for hours or even days, meticulously refining, calibrating, and grinding the surface shape of a single lens. However, when facing large-scale aspherical lens production, this traditional "slow and meticulous" inspection method is no longer applicable.
With the explosive growth of commercial aerospace constellations (such as the mass demand for aspherical lenses in satellite laser communication) and the large-scale mass production of millions of precision molded glass lenses in fields like intelligent driving, mobile phone lenses, and handheld devices, the traditional "slow and meticulous" inspection method is becoming the "suffocating lock" hindering the energy efficiency of the entire optoelectronic industry.
Large manufacturers are experiencing an extreme struggle for efficiency in their outgoing inspection, incoming inspection, and trial assembly discussions. I. The "Original Sin of Speed" in Traditional Inspection and the Survival Crisis of Million-Level Mass Production Currently, the two most commonly used benchmark imported devices by major domestic optical manufacturers for process R&D and mass production full inspection of aspherical and freeform surfaces are the UA3P (contact-type 3D profilometer) and LUPHOScan (non-contact ultra-precision optical profilometer). While they are indeed versatile and accurate, their snail-paced speed is a major obstacle for production managers and finance directors when dealing with glass molded lenses shipped in the millions:
● UA3P: Employing probe-based contact scanning, even with an optimized scanning path, it can only test a maximum of about 20 pieces per hour. Most critically, it essentially obtains 2D surface shape (cross-sectional line data), failing to fully reconstruct the entire aspherical surface across the entire field of view. Further 3D surface shape measurement faces a further sharp decrease in efficiency. It is worth emphasizing that optical engineers always require 3D surface shapes for analysis, because 2D surface shapes suffer from significant information loss during analysis, especially the loss of non-rotational aberrations, leading to huge discrepancies between the analysis results and actual results.
● LUPHOScan: Although it achieves 3D surface shape measurement through multi-wavelength interferometric scanning, due to its point-by-point/circle-by-circle scanning physical mechanism, it can only test about 10 pieces per hour. In the high-end consumer electronics and intelligent automotive sectors, this inefficient testing is causing a chain reaction of business disasters. It is well known that glass molding processes, due to factors such as temperature control, mold wear, and thermal stress release, inherently have lower yield stability than traditional plastic injection molding. Furthermore, the extremely high cost of glass lenses means that the yield of components (single lenses) is exponentially amplified, directly determining the final lens yield, and having a huge impact on the company's final costs and net profit. If the inspection speed can't keep up, large manufacturers either choose "gambling-style sampling," bearing the risk of high yield losses, or choose to "increase the frequency of sampling," but the fixed asset investment and depreciation of inspection equipment are daunting for business owners.
II. A Qualitative Efficiency Change Leading to a "Three-Tier Revolutionary Functional Upgrade" If a revolutionary, highly efficient inspection method could increase the inspection speed of 3D full-surface shape by 10-20 times, it would bring large manufacturers far more than just saving the purchase cost of a few imported machines. It would represent a revolutionary upgrade for the entire enterprise in terms of production model and core competitiveness.
Specifically, it can help large manufacturers achieve a qualitative leap in three business and engineering levels:
1. First Level: Absolutely Improved Inspection Efficiency (The Fundamental Basis for Cost Reduction and Efficiency Improvement) The inspection time for a single item, originally calculated in "minutes" or even "half an hour," is directly compressed to the second level. Full-scale 3D surface shape inspection, light-speed in and light-speed out. This directly transforms the testing workshop from a "capacity bottleneck" into an "automated production line," completely unleashing the delivery throughput of large-volume orders.
2. Second Level: Big Data-Guided Trial Assembly Optimization (From "Trial and Error" to "Engineering Certainty") In high-end lens development, the most time-consuming stage is the "engineering trial assembly." Due to the lack of large-scale, full-category real 3D surface shape data, traditional trial assembly can only rely on a very small sample (e.g., sampling a few dozen pieces) to "try their luck and find the optimal solution," with debugging cycles often lasting several weeks. If testing is achieved at the second level, large manufacturers can possess tens of thousands of sets of real, full-scale 3D surface shape big data in the early stages of the trial assembly process. Engineers can use this big data to optimize engineering trial assembly combinations, optimally misaligning lenses with different surface shape error distributions, forcibly improving the overall assembly yield of the lens at the assembly and adjustment stage.
3. The Third Level: Feeding the AI Manufacturing Model (The Ultimate Moat of Dimensional Reduction) Currently, the world is watching the meteoric rise of Silicon Valley giants in the field of AI, and what we see are mostly the changes AI brings to consumer search, copywriting generation, or coding. However, a trend far exceeding expectations is that once AI matures further, the competition among manufacturers will no longer be based on traditional channels or first-mover advantage, but on a comprehensive upgrade of R&D capabilities and organizational management capabilities. And the core engine here is the comprehensive empowerment of AI.
Currently, when major optical companies communicate with end-user clients, their proudest assets remain design capabilities, mold manufacturing capabilities, or production capacity.
However, we boldly predict that once a major manufacturer achieves a breakthrough in intelligent manufacturing, the standards for future business communication will be completely rewritten. The competition will then shift to AI computing power, large-scale digital and intelligent model capabilities, and the resulting "ultimate efficiency," "ultimate cost," and "ultimate user experience."
The underlying logic supporting all of this remains rapid, comprehensive, and accurate data. Without massive amounts of high-frequency, nanometer-level accurate 3D surface data for self-training, the so-called AI manufacturing large-scale model is merely water without a source. The core value of high-efficiency 3D inspection lies in its ability to continuously and stably input large volumes of realistic and accurate 3D surface data into the manufacturer's AI model.
Based on this massive amount of data, AI can gain real-time insight into the nonlinear mapping relationships between mold deformation, temperature drift, material batches, and high-order aberrations of the final surface shape, thereby continuously optimizing the overall configuration and correcting front-end process parameters. This represents a radical upgrade in production competitiveness: a company's core competitive advantage will shift from relying on "trial and error with very few samples based on the experience of veteran engineers" to a globally intelligent closed loop guided by computing power and big data.
Whoever feeds their AI model first will be the first to achieve extreme cost and efficiency, thus securing the sole ticket to the future of optical manufacturing.
III. Zhixing Optics: CGH Combined with Automation, Aiming for an Efficiency Miracle of 200 Pieces/Hour. To help major manufacturers take on this game that will reshape the industry landscape, Ningbo Zhixing Optics Co., Ltd., with its team of PhDs possessing both scientific research expertise and industry experience, has revealed its trump card.
In the process of interferometric measurement of aspherical surfaces, the most time-consuming, most experience-intensive, and most inefficient process is the aspherical zero-position adjustment (Alignment). Even highly skilled engineers, when faced with complex aspherical surfaces, often need about 2 minutes to repeatedly adjust eccentricity, tilt, and find the return error. Zhixing Optics has leveraged its self-developed CGH (Computer-Generated Hologram) alignment technology, combined with a high-precision electromechanical automation system, to deliver a powerful combination of hardware and software solutions:
10-Second Limit Alignment: We've dramatically reduced the time-consuming aspherical zero-position adjustment process from approximately 2 minutes for skilled workers to under 10 seconds!
AI Misalignment Model: Traditional alignment methods rely on cumbersome "sensitivity matrices." When dealing with complex multidimensional aberrations of high-slope, high-order aspherical surfaces, linear fitting is prone to distortion, leading to inefficient iterations of continuous "calculation-fine-tuning-recalculation." Zhixing Optics has completely revolutionized this approach. By using deep AI training on the misalignment errors in the CGH design files, we have successfully constructed a high-dimensional nonlinear misalignment model. With a single image captured by the interferometer, the AI algorithm instantly and precisely locks onto the current eccentricity and tilt data in the global aberration space, achieving extremely high-speed computation and significantly improving the system's response efficiency and alignment accuracy.
• Target – 200 pieces/hour: Currently, with the addition of a mechanically automated loading and unloading mechanism exclusively customized by Zhixing and a subsequent AI-powered rapid interference fringe analysis algorithm, the system is expected to soon boost the 3D full-field inspection efficiency of aspherical surfaces to 200 pieces/hour. This is 10 to 20 times faster than traditional scanning equipment!
🤝 Seeking Partners: Collaborative Rapid Validation and Strategic Win-Win: Zhixing Optics currently boasts a core team of experts led by 3 PhDs in optics and 2 provincial-level high-level talents. We are deeply rooted in the ultra-precision metrology field, leveraging our strong technical foundation from top national optical and mechanical research institutes to build a robust technological moat and a high degree of self-sufficiency. We understand that such a disruptive industrial efficiency revolution cannot be achieved in isolation; it must be deeply rooted in the real-world, million-level mass production lines of large manufacturers. We welcome collaborations with manufacturers of glass molded lenses, automotive lenses, and handheld device lenses who agree with our anticipated development trends and look forward to joint development and verification efforts.
Zhixing Optics enables precision measurement to truly achieve "unity of knowledge and action".
