Special Cases

21
2021.1
【2021 Application Example】 AI Complements the Disruption of Traditional Industry Experience: Production Forecast Analysis in Plastic Recycling Process

As the number of veteran craftsmen in traditional industries diminishes In Taiwan, SMEs have always played a central role in Taiwan's industry, accompanying Taiwan through various 'economic miracles' periods But as time progresses, the old masters gradually became elderly craftsmen Coupled with the phenomenon of fewer children and changes in the overall industrial structure, fewer and fewer of the new generation are willing to enter traditional industries Now, it can be observed that the main combination on most SMEs' operational fields is formed by 'elder craftsmen' together with 'foreign workers' These experienced craftsmen, who act as living dictionaries of field experience, suffer from a lack of successors to continue the tradition, leading to a growing difficulty in sustaining on-site experience transfer in traditional industries The limits of traditional hands-on process optimization are in sight Located in Tainan Baoan Industrial District, 'Tangxian Company' was established in 1972, initially manufacturing high-quality weaving equipment It possesses the capability to manufacture machinery, and in recent years it has actively developed environmentally friendly plastic recycling equipment in response to international green energy, recycling, and environmental protection demands Ultimately, they have successfully developed low energy consumption, low waste, high purity, and high output recycling granulators with a sleek and efficient machine design supplemented by advanced intelligent control technology Tangxian Company's self-developed plastic recycling granulator equipment However, in the production process of plastic recycling, when faced with hundreds of material types and dozens of process temperatures, speed settings, what is faced is thousands of possible parameter combinations Previously, the adjustment of various production process conditions was reliant on the on-site staff the experience of the craftsmen Thus, during the transition of production of different incoming materials such as PET, PP, PE, a significant amount of raw materials would be wasted during the trial phase The professional information gap in traditional industries Tangxian Company recognizes the importance of data In the past, although process parameters were recorded, due to a lack of data capabilities at the time, it was primarily in paper form, manually written down by the operating staff, accumulating a large amount of paper data However, this also meant a lack of scientifically accurate and detailed information available for real-time reference and adjustment Process parameters logbook, records the state of about a dozen machines and production figures hourly In quality control as well, due to a lack of control over the quality of output and monitoring and feedback mechanisms for unit time production, it becomes difficult to predict the profit conditions of each batch Production management can only estimate and average cost and productivity changes over the process from the outcomes, without being able to objectively and timely restore the production conditions to reasonableness or make clearer adjustments when facing quality abnormalities Site reality left image shows recycled scraps right image shows pellet production Taiwanese manufacturers possess strong machinery manufacturing capabilities, and many modern machines now have data capabilities, recording real-time status and information via IoT But is the infrastructure of the factory's on-site and information systems ready yet When the Old Master Meets AI With government referral, Tangxian Company partnered with a Taiwanese data science company, working together to integrate AI services and optimize internal processes using AI They started with a medium-sized plastic recycling production line within the factory as a trial field After establishing a successful benchmark, this model was expanded to larger plastic recycling machinery within the factory to continue verification and application Initially, both parties converted the past handwritten paper data into digital format using OCR supplemented by manual correction Tangxian Company also worked with the supplier of the human-machine interface of the machinery to integrate the control panel and parameter data into the factory's database, allowing real-time monitoring of machine status and eliminating the complexities and potential errors of manual transcription Panel of plastic recycling granulation machine, showing current process temperatures, speeds, and power usage Meanwhile, the Taiwanese data science company further modeled dozens of parameter data through AI, conducting scenario analysis to simulate various production possibilities under environmental parameters and material inputs, identifying key characteristic parameters and providing parameter adjustment recommendations to decrease costs during the trial phase Applying data analysis to traditional industry machinery processes After the old master receives the raw materials, they only need to enter the relevant material characteristic parameters, and the system automatically generates recommended process parameters After small adjustments by the old master, they proceed with the trial production of the material, effectively reducing the waste of materials, water, electricity, and manpower caused by incorrect attempts Moreover, Tangxian Company has proactively deployed the concept of 'production pedigree' in the plastic recycling process, allowing the batch's raw materials and process parameters to be accessed by scanning a QRCode Production and sales pedigree of plastic recycling pelletizing Taiwan's SMEs have strong machinery capabilities, just waiting for the 'east wind' of data From industrial revolutions 20 to 30, even 40, many Taiwanese SMEs face challenges in transitions not just in upgrading machinery, but after investing in modern equipment and generating data, they do not know how to utilize it effectively It is impractical for these manufacturers to develop a specialized data analysis department on their own meanwhile, Taiwan also has many innovative teams with strong software capabilities in AI and data analysis, possessing the technology but lacking the field and data Therefore, if the traditional industries of Taiwan could be fully integrated with the innovative teams in AI and data analysis, it would not only address the current challenges of manpower and experience transfer faced by traditional industries but also advance Taiwan's development and application of AI significantly「Translated content is generated by ChatGPT and is for reference only Translation date:2024-05-19」

2021-01-21
【2021 Application Example】 Fongyu Uses AI Knowledge-based Fish Farming to Effectively Increase Aquatic Production by 10%

Fisheries is an important industry in an island economy However, the fish farming industry has faced severe challenges in recent years, including climate change, labor shortage, and rising costs In particular, nearly 110,000 workers in agriculture will retire due to old age over the next 10 years For this reason, the need for aquaculture to move towards smart farming is becoming increasingly urgent Founded in 2014, Fongyu Corp Ltd has developed a unique eco-friendly farming model based on its own fish farming It uses AI knowledge-based fish farming to effectively increase aquatic product production by 10, and reduced labor cost by 15 The word "Fongyu" has a profound meaning "Fong" represents good mountains and "Yu" represents good water, and is the hope that companies will allow Taiwan to always have good mountains and good waterIt is also a homophone for "having a full figure," expressing the hope that products will give consumers a full and healthy body and mind The founder of the company, Liu Chien-Shen, has been through the difficult entrepreneurial journey of becoming an apprentice in fish farming, raising funds, renting fish farms, establishing a fish farming company, building a brand, and expanding sales Labor shortage and aging workers are hidden worries in the fish farming industry Currently, fish farms in Taiwan are still mainly traditional fish farms, and farming techniques are still passed down through word-of-mouth In addition, the labor shortage and average age of workers exceeding 60 years old has made it impossible to effectively stably improve productivity and yield This farming method makes it difficult to prevent and control diseases, and greatly increases the possibility of excessive use of drugs, environmental pollution, and water quality and ecological damage, creating a vicious cycle that lowers the quality of fish farming In addition, 651 of workers in Taiwan's fish farming industry are inadequately skilled With limited support from IoT sensors, traditional fish farmers still mainly rely on their own experience and knowledge for water quality management, feeding, and disease detection Fish farming management relies heavily on the ability of individual fishermen Once experienced workers retire, the industry will not only face the issue of succession, but also the difficult of stably supplying a certain amount of harvest that meets quality standards This may cause a dilemma for the entire industry from fish farming to sales In order to improve the pain point of inability to pass on experience in fish farming, and at the same time create a "digital" foundation for fish farming, the top priority must be to collect farming behavior data and develop AI services as an important starting point Fishery digital twin technology helps fishermen transition to smart farming With the assistance of the Institute for Information Technology III, Fongyu implemented the "fishery digital twin" technology to dynamically adjust the farming schedule In other words, the fish farming schedule is adjusted according to the species, habits, and variables of the fish The use of AI in fish farming not only effectively increase aquatic production by 10, but also reduced labor cost by 15 In terms of specific methods, we first digitalized the fish ponds, feed, and decision-making behavior for each species, such as sea bass and Taiwan tilapia, and recorded the seasonal temperature changes from releasing seedlings to harvesting, all of which were digitalized, gradually recording the experience and methods of experienced workers into a rich database Based on the recorded data, we analyzed the compound variables to find the best farming behavior and generate a dynamic farming schedule The records for each pool provide data on workers' experience However, fish farming behavior generally relies on rules of thumb Even experienced fish farmers cannot ensure that they will find the best solution Therefore, new methods are proposed to solve this issue That is, "to determine the best fish farming behavior by predicting the interaction with water quality and past data on feeding, and evaluating fish farming behavior based on water quality and fish farming," and provide fishermen with the most intuitive recommendations through daily schedules To continue optimizing the dynamic fish farming calendar on a rolling basis, iterations of the model will be developed through the three-step cycle 1 Input the current fish farming calendar into the model 2 The model predicts the future environment 3 Shortcomings of the fish farming calendar are corrected based on the future environment to obtain a new version of the fish farming calendar In the process, the experience of aquaculture experts is used to establish the causal relationship between fish farming behavior and the environment The establishment of a dynamic fish farming process and technology-based fish farming recommendation services provide a traceable and detailed fish farming process It is one of the few technologies that can digitalize fish farming Fishermen can quickly and easily record their daily behaviors to build knowledge without taking up too much time, but in the long run it can reduce labor cost by 15 and increase output and revenue by an average of 10 Smart fish farming has achieved outstanding results, reducing labor cost by 15 and increasing output by 10 At the same time, the fish farming calendar can also be extended to different aquatic species, such as white shrimp, milkfish, clams, and Taiwan tilapia, to produce fish farming schedules for ponds with different specifications, and the harvested aquatic species can be traced according to different specifications, establishing vertically integrated services for safe food products Fongyu's main products are divided into two categories One is aquaculture modules, including fry, feed, materials and probiotics, production planning and processes, and monitoring, which can be sold separately or exported as modules The high-quality aquatic products produced by Fongyu have repeatedly won awards Figure Fongyursquos official website The other category is high-quality aquatic products, including seabass fillets, seabass balls, oil-free seabass balls, seabass dumplings, and seabass soup The products have won various awards, including the top ten souvenirs in Pingtung in 2017, "Barramundi Fillet" won the 2017 Eatender of the Council of Agriculture COA, "Oil-Free Barramundi Fillet" won the 2018 Eatender Gold Food Award of the COA, and "Dumplings of Barramundi" and "Barramundi Broth" won the 2019 Eatender of the COA The consecutive awards represent that the "quality" of Fongyursquos aquatic products can be seen and eaten with peace of mind In addition, Fongyu has exclusive fingerlings that meet international needs, such as Pure seawater cultured tilapia fingerlings and seawater Taiwan tilapia fingerlings from selective breeding FY-01 are items that aquaculture companies in many countries are looking forward to The company also has aquaculture modules, disease monitoring tools, and feeding materials designed in accordance with the environment, in order to provide customers with more stable income

2021-09-28
【2021 Application Example】 AI Analysis Cloud Service Platform for Remote Sensing Big Data Enables the Smooth Application of Satellite Remote Sensing Images

Although satellite remote sensing images can make all surface objects visible, it still requires a lot of time and manpower to be truly applied to the industry In order to effectively solve the problems that customers face in digesting huge amounts of image data and eliminate technical obstacles for cross-domain users to process satellite remote sensing images, ThinkTron has developed an "AI Analysis Cloud Service Platform for Remote Sensing Big Data" as a new beginning for cross-domain AI applications for spatial information In recent years, in response to the impact of industrial globalization, Taiwan's agriculture has continued to transition towards technology-based and higher quality, improving the yield and quality of crops by solving problems, such as microclimate impacts and pest and disease control The demand of agriculture on images has expanded endlessly to accurately grasp the growing environment of crops In the early years when UAVs unmanned aerial vehicles were not yet popular, manual field surveys were the most basic but most labor-intensive work With the emergence of UAV drones, aerial photography operations might not be difficult, but the range that can be photographed is limited Furthermore, surveying expertise is required to accurately capture spatial information At this time, the use of satellite remote sensing data may break away from the past imagination of using image data Taiwan Space Agency TASA ODC data warehouse services In the past ten years, with the breakthrough of modern satellite remote sensing application technology, Digital Earth has become a new trend in global data collection Countries have developed data cube image storage technology, and the development of smart agriculture has become one of the largest image users Determining planting distribution is the first step in understanding crop yields Free satellite remote sensing images, powerful data warehousing support, and the team's robust image recognition technology are important supports for accelerating agricultural transformation Using satellite remote sensing image data can accelerate the development of smart agriculture However, in the past, it was difficult to extract large-area crop distribution through satellite remote sensing images, not to mention the cost If you wanted to use free information, you had to visit all websites of international space agencies, look through the wide variety of satellite specifications, and carefully evaluate the sensor specifications, image resolution, and revisit cycle After finding suitable images, you had to look at them one by one to filter the ones you need Next is downloading dozens of images that are often several hundreds of Megabytes MB each, which might exceed the capacity of your computer Also, after overcoming image access and preparing data, you must then start to confirm the downloaded image products and which bands you want, because the image you see is not just an image file jpg or png, but rather complex multi-spectral information, attribute fields and coordinate information It takes a lot of effort just to confirm the correct information Facing GIS software packages with complex functions is the start of another trouble The complex image pre-processing process and the inflexible machine learning package greatly reduce the efficiency of analyzing data After finally getting the results of crop identification, you might find that the best time for using map information may have already passed The above-mentioned complex and time-consuming satellite image processing problems are precisely the pain points of the market ThinkTron expanded from traditional machine learning to modern deep learning applications, and developed an "AI Analysis Cloud Service Platform for Remote Sensing Big Data" under the GeoAI framework, breaking through the constraints of details in the spatial information for customers Differences between the process before and after introducing the AI analysis cloud service platform ThinkTron said that Taiwan's ODC Open Data Cube system has been completed and began providing services after years of efforts from the Taiwan Space Agency TASA, formally becoming aligned with international trends The powerful warehousing technology allows users to easily capture and use image data of a specific time and spatial range according to their needs The warehouse stores multiple satellite image resources from international space agencies, including the ESA's Sentinel-1 one image every 6 days, Sentinel-2 one image every 6 days, USGS's Landsat-7 one image every 16 days, Landsat-8 one image every 16 days, and the domestic Formosat-2 one image every day and Formosat-5 one image every 2 days ThinkTron develops satellite image recognition tools based on Python Breaking free from the limitations of GIS Geographic Information System software packages, ThinkTron integrated GDAL Geospatial Data Abstraction Library based on Python, and considered computing efficiency and parallel processing when developing all tools required for satellite image processing and image recognition modeling, including coordinate system and data format conversion, grid and vector data interaction, and data intra-difference and normalization All of the tools are designed with AI applications in mind, and some commonly used tools are packaged into an open source package under the name TronGisPy to benefit the technical community ThinkTron utilized the team's understanding of satellite remote sensing images and the collected tagged data crop distribution maps to preset the image recognition modeling process, the required training data specifications, and dataset definitions This is imported into the machine learning LightGBM or deep learning CNN framework that was completed in advance, and the entire training process to be performed in the Web GIS interface, providing users with partial flexibility to freely filter images, confirm spatial and temporal ranges, select models, and adjust hyperparameters In addition to the operation of training models, it also provides historical models to output identification results, and finally displays the identification results of crop distribution on the Web GIS map In fact, agriculture is not the only industry that needs satellite remote sensing applications AI applications of spatial information have also appeared in various fields as companies in different industries aim to enhance their global competitiveness For example, surveying and mapping companies that have a large amount of map data can use the AI analysis cloud service platform to store map data while also accelerating the efficiency of digital mapping Under the severe global climate change and the risk of strong earthquakes, there is a wide variety industrial insurance, agricultural insurance, financial insurance, or disaster insurance are all inseparable from spatial information The use of remote sensing image recognition to understand insurance targets has long been an international trend AI Analysis Cloud Service Architecture for Remote Sensing Big Data

2021-11-28
【2021 Application Example】 HRT Technology Improves Production Efficiency by 20% Through AOI Detection of Defects in VCSEL Packaging

In 2017, the launch of the iPhone X made 3D sensor technology used in Face ID highly popular, which drove the development of VCSEL, a core component in the 3D sensor module In the detection of defects in incoming packaged VCSEL, the use of AI inference models can solve the industry's issue with low yield and improve reliability to 95 VCSEL technology currently can be used in many applications and various end consumer markets, including robots, mobile devices, surveillance, drones, and ARVR VCSELs are a good solution in applications that require high-speed modulation capabilities, such as cameras and biometrics VCSEL technology has a wide range ofnbsp applications, including in drones Pictured Zoyi Technology's Agricultural Drone VCSEL technology has a wide range of applications, AI technology assists in defect detection HRT Technology stated that the packaged VCSEL market is also facing strong price competition from competitors, and needs to further reduce costs and enhance product competitiveness One of the key problems is the replacement of glass lens with epoxy resin lens The production of traditional glass lenses has high yield, but the cost is higher than that of epoxy resin lenses Due to the cutting process of epoxy resin, the side wall of cutting lines can easily have rough edges, causing it to be oversized The release of stress caused by heat during the mounting process will directly cause the optical lens to break HRT Technology pointed out that the incoming inspection of VCSEL epoxy resin lenses is very important Under the constraints of packaging space, the space for fitting the package and optical lens is limited Moreover, the optical lenses will be confined to a metal frame If the dimensional tolerances are properly controlled, stress release due to heat during mounting can easily cause the optical lens to break, resulting in a yield loss of up to 10 in the VCSEL package reliability verification, resulting in an increase in production costs In order to solve the problems above, HRT Technology hopes to use AI to monitor the size and appearance defects of epoxy resin components in the VCSEL epoxy resin lens incoming stage, verifying whether their dimensions meet specifications, whether the cutting edges are smooth, and whether there are any defects in their appearance Since traditional incoming material inspection requires a rough visual inspection by humans to distinguish the quality The problem of image collection needs to be solved first to successfully collect image data Therefore, HRT Technology first developed an Automated Optical Inspection AOI device, which includes X, Y, Z three-axis motion, high-resolution cameras, and related control software to automatically record images After collecting the image data, opencv aligns the test image and a normal image to determine differences between the two images, and then pixel mapping is used to compare the pixel area to complete initial screening Manual labeling is carried out according to the image classification above, including samples that are normal, have defects in appearance, or have different shape characteristics, and then algorithm training and verification is carried out Residual neural network ResNet or other related algorithms are used for deep learning to identify the quality of lenses Implementation of AOI inspection improves production efficiency by 20 and above Comparing the differences before and after the implementation of AI image inspection, the incoming VCSEL lens inspection before implementation only involved manual inspection of the appearance The lens is packaged on the VCSEL package that has completed die bonding After passing the general light up test, the final reliability test high temperature reflow is performed Failed samples go into the rework process However, after the implementation of AOI inspection, it can screen defective lenses sooner and reduce the cost of subsequent materials input, it can also reduce the need for rework due to failure, improving yield to 95 and above in the reliability verification This is expected to help companies reduce production costs by 10 and increase production efficiency by 20 and above The difference before and after implementing AI image detection HRT Technology pointed out that this technology is an AI application developed based on tiny images It uses deep learning algorithms to identify defects in the images The trained network automatically classifies image data to predetermined categories Defect categories can be determined through reference images, so cumbersome programming is not required In the industrial machine vision environment, deep learning is mainly used for classification tasks in applications, such as inspection of industrial products or identification of parts In the future, with the development of IoT wearable devices and the trend of energy saving, the size of optoelectronic components will continue to shrink This technology can be applied to the detection of defects in the appearance of other tiny optoelectronic components in the future

2021-12-05

Records of Application Example

【導入案例】緯霖華岩科技聯手研發以AI預測性維護機台,提升血液透析機使用率
【2020 Application Example】 Latitude X Huayan Technology Jointly Develop AI for Predictive Maintenance of Machinery, Improving Utilization Rates of Hemodialysis Machines

Taiwan has the highest rate of dialysis in the world Keeping dialysis machines functioning properly is the top priority for reducing risks According to the latest annual report released by the US Renal Data System USRDS, Taiwan has the highest dialysis rate in the world In 2018, acute and chronic kidney disease patients spent NT51378 billion on health insurance, and the number of dialysis patients in the country surged past 90,000 When kidneys function no longer, replacing kidney function either through transplantation or dialysis is necessary, with about 90 of patients choosing hemodialysis commonly referred to as 'dialysis' Patients generally require treatment three times a week for 4-5 hours per session at specific medical facilities hemodialysis centers, commonly known as 'dialysis centers', which is a high-risk medical procedure During hemodialysis at the centers, unexpected events directly affect patient medical safety and quality of treatment, consuming medical resources and manpower to resolve or correct Reducing these incidents during hemodialysis is a major requirement for these centers The two most common incidents involve dialysis equipment problems and patient complications, with most technical issues attributed to the hemodialysis machines Hemodialysis machines are structurally complex and prone to safety hazards The design of a hemodialysis machine is intricate and precise, featuring an integration of fluid mechanics, electronics, mechanics, and optics in its extracorporeal circulation system Due to long operating hours, the machine is susceptible to thermal and chemical corrosion, causing wear and tear and potentially impairing the entire dialysis system's operational performance, with multiple risks and safety hazards When a hemodialysis machine experiences an 'event,' whether minor or major, reactive maintenance is triggered Not only do patients have to switch to an alternate bed, but during the approximately 2 to 3 days of maintenance downtime, the affected beds become unavailable, thereby reducing the number of available beds and causing scheduling issues for already booked patients Any 'event' involving hemodialysis machines is a significant concern for centers, thus improving the equipment utilization rate of these machines is a pressing issue Using AI for Predictive Maintenance to Improve Utilization Rates of Hemodialysis Machines Development Workflow By utilizing big data and AI predictive framework to adopt a proactive 'predictive maintenance' approach instead of a reactive 'fix-on-failure' approach, it helps reduce the occurrence of irregular incidents, improving the availability of hemodialysis machines and thereby hoping to handle their malfunctions better, conserving medical resources, manpower, and time, while improving treatment quality and protecting patient life safety Through AI predictions, maintenance of hemodialysis machines can be categorized as 'Predictive Maintenance' and 'Real-Time Fault Diagnosis' 'Predictive Maintenance' refers to regular checks of the machine's status using big data and an AI prediction model during the daily pre-heating of the machines, delivering health status alerts if unhealthy trends in parameters are detected 'Real-Time Fault Diagnosis' involves analyzing data and equipment status during dialysis using the AI model to ascertain if predictive maintenance is necessary when an issue arises, it can be diagnosed and non-major events immediately resolved Solution Diagram With an innovative service mode, promoted across dialysis centers in Taiwan or the Asia region The AI predictive maintenance model can reduce abnormal events during dialysis, optimize on-site resources, increase available hemodialysis bed numbers, and consequently provide further safety for patients For 'patients,' it reduces the incidence of mishaps causing harm and discomfort for 'medical staff,' it enhances the ability to handle such events easily, improving job satisfaction and quality and for 'hospitals,' it fosters improved medical quality, patient satisfaction, and cost savings, while minimizing medical disputes 'Increasing the Availability of Hemodialysis Equipment' is crucial for dialysis centers AI predictive maintenance as an innovative service model can be promoted extensively among dialysis centers with large patient volumes across Taiwan or Asia, also integrating individual dialysis statuses, including backend maintenance, dispatching, and parts inventories, planning a new cloud service operation model「Translated content is generated by ChatGPT and is for reference only Translation date:2024-05-19」

【導入案例】「AI智慧辨色及成本最佳化控管系統」,自動辨色,突破傳統調色模式,大幅降低成本、提升良率
【2020 Application Example】 "AI Color Recognition and Cost Optimization Control System" automatically recognizes colors, breaks through the traditional color grading model, significantly reduces costs, and improves yield!

Mixing new colors relies on the experience of master craftsmen The so-called "computer color matching" in the paint industry is simply the selection of "existing colors" for mixing, but there is actually no way to mix paint for a ldquonew colorrdquo and it all relies on the experience of master craftsmen Hence, it is necessary to start from scratch when a new color is encountered, which consumes a lot of manpower and time Moreover, due to the different color mixing habits of each master craftsman, the cost can be significantly different despite producing the same result The trilogy when paint factories face the crises of transformation I Lack of color mixing standards Generally, when traditional paint factories produce new colors, they will use a "spectrophotometer" to measure the LAB value of the sample color, and then the paint mixer will mix the paint of that color based on past experience After color mixing is completed, the instrument will be used to test the LAB value and C and H wavelength This process does not have a complete system and database records, and there are not standards for color mixing II Production costs are difficult to control Paint factories produce many pigments with different materials and functions, and the cost of paint will vary depending on the "color masterbatch material" used Even if the color number of the masterpiece is the same, the cost will be different if the ratio of the color masterbatch is different Paint mixers do not have a set of color mixing standards when mixing paint, making it difficult to control production costs III The color grading process is lengthy and personnel training is difficult As instruments cannot replace manual color mixing, the training of a paint mixer requires years of experience in paint mixing, familiarity with chromatology, as well as basic understanding of hue, saturation, and brightness If there is no basic reference color values when mixing paint, the paint mixer must spend a lot of time repeatedly mixing colors, resulting in a loss from time cost Developing an "AI Color Recognition and Cost Optimization Control System" The paint factory engaged in industry-academia collaboration with the Department of Computer Science amp Information Engineering of Chaoyang University of Technology through CDIT Information Co Ltd, and utilized the university's AI research capabilities to jointly develop the "AI Color Identification and Cost Optimization Control System" It established a database of "paint color numbers" and "color masterbatch material cost," and analyzes the optimal color mixing and optimal cost formula through data mining methods The paint mixer can refer to the formula analyzed by the system for color mixing, and then input the formula into the system after paint mixing is completed The formula is fed back to the basic database and an "artificial neural network model" is used by the system for deep learning, establishing a color grading standardization system for cost control and data collection, so as to solve the current difficulties faced by paint factories In the early stages of system development, CDIT planned the system requirements of the paint factory, established the system architecture and system database, and then worked with Chaoyang University of Technology on the implementation of model functions for the application of data mining and artificial neural network After the system is completed, CDIT will assist the paint factory in system testing and correction The system will be introduced after correction and testing are completed, and training on system use will be provided to ensure the correct use of the system System Screen Differences before and after using the system Expand new markets for the paint industry to see the paint industry thrive The "AI Color Recognition and Cost Optimization Control System" collects the color mixing formulas of paint mixers, establishes a paint color masterbatch formula database, and records the cost of each color number The system's deep learning function is then used with a spectrophotometer to analyze the optimal color mixing formula for each data entry, so that the paint factory can control the cost of paint mixing The optimal color mixing formula recommended by the system increases the speed of paint mixing and increases output value Future benefits include The improvement in product yield reduces customer complaints and improves customer satisfaction The breakthrough in the traditional color mixing model improves corporate image Improves the efficiency of paint mixing, and allows the remaining time to be invested in training to enhance the professional capabilities of personnel It will also allow the joint expansion of new markets with the paint industry and learning of new application technologies, and promote them to other paint companies, enhancing the industry's overall competitiveness to see the paint industry thrive

 【導入案例】「凱比同學機器人」有個AI腦,不再答非所問
【2020 Application Example】 Kebi Student Robot now features an AI brain, solving relevancy issues in responses!

The unstoppable trend of smart homes In recent years, the rise of 'smart home devices' has not only led to the release of various products by major tech companies but also propelled the popularity of voice assistants, chatbots, and companion robots The 'voice shopping' market is set to become the next trend in retail According to a survey by Juniper Research, by 2023, the market size for transactions based on chatbots is expected to surge from 73 billion in 2018 to 112 billion A well-known domestic manufacturer of household robots offers its self-developed educational and companion service robots, with the 'Kebi Student Robot' as its flagship product However, due to its insufficient voice interaction capabilities with users, consumers often find the robot not smart enough and quickly lose interest, leading to abandonment This has a long-term negative impact on the purchasing decisions of other consumers Hello Kebi Student, can you understand what I'm saying Surveys have found that many users of 'Kebi Student Robot' especially enjoy talking or chatting with the robot, covering a wide range of topics However, using just Google or Microsoft's cloud platforms for developing voice chat conversations is not cheap With Google charging based on service volume, system operational costs are high, and these vary dynamically causing major difficulties in system cost management On the other hand, the robot manufacturing service provider has already invested significant resources in the development of hardware, software, and digital content for Kebi Student Robot Developing natural language dialogue and semantic understanding technologies in-house would require a significant amount of manpower and is slow With limited resources, there's an urgent need to seek third-party solutions to enhance the robot's conversational service capabilities and development efficiency Integration of the Web-AI developed iboai voice assistant brain platform with Kebi Student The key to the transformation from 'playing the lute to a cow' to 'I know you are upset' Web Intelligence Co, Ltd is a well-known AI natural language understanding technology service company in Taiwan Its products include a natural input method, TTS voice engine, and the iboai voice assistant brain platform This platform has been applied in smart speakers, high-speed train voice assistant apps, and even employee benefit systems for companies like China Airlines It quickly addresses the deficiencies in the robot manufacturer's AI service dialogue skills development, enriches the dialogue content and skills, provides contextually related conversational services, and makes Kebi Student appear smarter to users within a short period Service Architecture This case further applies the latest Principle-based semantic understanding engine technology from the Academia Sinica's Institute of Information Science, achieving deep natural language processing and understanding, and carrying out intent and entity analysis to generate interactive dialogue logic for continuous conversation Therefore, it also strengthens the basic social communication abilities of Kebi Student Robots, enhances the number of AI dialogue skills, and advances from simple question-and-answer to having capabilities for 'multi-turn dialogue' and 'contextual conversational' responses, making the robot's responses more human-like Additionally, the development time for Kebi Student Robots has been significantly reduced, effectively and significantly lowering and controlling cloud service maintenance and management costs 服務架構1 AI Kebi, becoming your ubiquitous companion Currently, many robots and smart speakers and other voice assistants on the market can only provide single-turn conversational services that end after one question and one reply A major difference with the iboai voice assistant brain platform used in this case is its capability for 'multi-turn conversational interaction with contextual understanding,' which is also the only platform in Taiwan supported on local or cloud services The iboai voice assistant brain platform can support various enterprises services, allowing businesses to design their own voice assistants or AI Chatbots for customer service swiftly, which can be applied across LINE, Facebook Messenger, websites, apps, and IoT devices, among others This case adopts the 'iboai inside' strategy, highlighting its role as an Enabler to upgrade corporate services to AI, hoping to also assist existing Chatbot manufacturers, app developers, commercial software vendors, information hardware dealers, system integrators, and IoT device merchants in upgrading their existing products and services to have AI natural language conversational capabilities, collaborating to provide a new generation of AI smart robot services for numerous businesses「Translated content is generated by ChatGPT and is for reference only Translation date:2024-05-19」

【導入案例】AI加值「香蕉契約合作管理作業系統」,有效提升香蕉外銷產值
【2020 Application Example】 AI-Enhanced 'Banana Contract Cooperation Management System', Effectively Increasing Banana Export Value!

Banana industry faces low-priced impacts from abroad In recent years, our country's banana industry has been impacted by low prices from the Philippines and Ecuador, with sales volumes decreasing annually, no longer seeing the golden era of Taiwan's banana exports to Japan The structure of banana pricing at the green stage doesn't vary much internationally, with similar inputs of fertilizers and harvested weights among countries However, international banana pricing for a box from the Philippines is around 11 USD, whereas Taiwanese bananas cost around 22 USD per box This is primarily due to the efficiency of investment capital and output at the 'collection centers' post-harvest The fragmented and scattered local farmlands substantially increase the costs of final products and thus restrict the export dynamics Furthermore, climate change affects the traditional southern export regions for Taiwanese bananas Warmer winters and altered summer rainfall patterns affect the physiognomies of the bananas produced, causing their size to rapidly exceed export standards and increasing the cost per unit of qualified goods during collection center processing or excessive water content which depreciates the taste historically associated with them, leading to a decrease in market prices These pressures from rising costs and dropping prices further squeeze the commercial value and viability of Taiwanese bananas Differences in planting environment affecting the stability of banana quality for export A fruit and vegetable cooperative in Yunlin County, originally a domestic banana collection center located in Yunlin, wasn't historically a part of Taiwan's banana export regions Since a field survey conducted in 2017 by TaiNong Co, Ltd, it was discovered that the quality of bananas produced in Yunlin has been comparatively stable against those from the southern regions The tighter organization of local farmers and crop rotation practices between rice and banana farms helped reduce incidences of Yellow Leaf Disease and effectively maintain production levels Banana export However, without prior experience in exports, TaiNong gradually introduced Japanese standards for exporting with the local farmers, defining the size and width of fruit fingers, stalk cutting, and boxing methods This aims to gradually establish a banana export hub in the central region Yet, the climate in Yunlin significantly differs from the southern regions Current practices in banana exporting are based on experiences from Kaohsiung and Pingtung and do not incorporate how the shift in production areas northwards affects banana growth Hence, there remain excessive rejections at the collection centers occasionally causing disputes among farmers Agricultural risk management data service, development of banana specification volume fluctuation prediction model 台農發股份有限公司既有之集貨場對契作香蕉農戶包裝分類品檢機制,收集之數據資料與悠由數據應用股份有限公司配合,運用資料科學研究方法,透過研究規劃、資料蒐集擷取、資料清洗、特徵萃取、資料融合、資料分析演算法建立、分析結果、模板開發、專家會議討論等步驟建立分析應用流程。 By integrating dataset including collection centers' incoming batch container numbers, origin, banana quantities, data on each box of fruit bunches, and data of defect sampling records, along with internal purchase prices and prices from various purchasers, through the Banana Contract Cooperation Management System linked with data decision analysis systems and APIs, it supports subsequent judgments by providing analysis data to the fruit and vegetable production cooperative 悠由數據擷取與蒐集香蕉契作戶產地之歷年氣象環境資料、公開批發市場的產地價格及香蕉生理模式等數據,結合台農發的分類品規數據,建立「香蕉品規量能波動預測」演算機制,並將分析預測結果回饋至香蕉契約合作作業管理機制。 Visualized harvest scheduling analysis By leveraging varied predictive analytic outcomes of banana specifications, collection centers can utilize this as an advance warning and risk management decision-making tool, further adjusting supply to tackle inconsistent production capacity and specifications faced during acquisition Fruit and vegetable cooperative X TaiNong X Youyou Data Applications collaborating closely, creating a win-win-win This successful alliance formed a close cooperative relationship between the place of production, TaiNong, and Youyou Data Previously, farmers often distrusted traders, and traders lacked control over farmers, leading to conflicts This alliance allows the requirements of the distribution side to reflect actual shipment specification fluctuations and present them digitally, enabling farmers to objectively understand their shipping quality and empathize with the difficulties of traders, thus fostering cooperation Innovative model of banana contract management TaiNong's cooperation with Youyou Data on the banana contract management system provides a platform that combines crop physiology with climate predictions to obtain foresight data For other products managed by TaiNong, such as pineapples, lettuce, carrots, and pineapple sugar apples, this has been greatly enlightening In the future, by facilitating farmers to participate in the Production History System and connecting land registry data with this contract system, the introduction of the Production History System will be aided This system is also considered by TaiNong for commercial acquisition moving forward「Translated content is generated by ChatGPT and is for reference only Translation date:2024-05-19」

【導入案例】【文鼎木刻思打造AI造字助手】傳統鑄字行文化傳承現曙光
【2020 Application Example】 Wending x Woodcut Thinking creates an AI calligraphy assistant, and the cultural heritage of the traditional calligraphy industry is now dawning

The only remaining calligraphy shop in Taiwan, cultural heritage exposed to crisis A certain traditional type casting shop in China is the only one "still operating" in Taiwan It has a sense of mission and hopes to pass on Taiwan's long-standing beautiful letterpress technology But even if you want to continue to cast characters, the existing molds have been repeatedly cast for more than 40 years, and the "copper molds" used to cast lead characters have been damaged The tall lettered walls in the store are facing the predicament of being eroded by time Each "copper mold" can be used to produce 10,000 typefaces, so it is called the "mother of typefaces" If the handwriting on the copper mold is blurry, the cast lead characters will also be blurry After printing, there will be incomplete radicals and uneven strokes In Taiwan from the 1950s to the 1970s, the "block regular script" copper molds used to cast characters played an important role in spreading civilization Because of the serious collapse of the copper molds, the owner of the calligraphy shop launched the "Character Bronze Mold Restoration Project" in 2008 Together with a group of enthusiastic volunteers, they first repaired the "Blockcase" copper mold fonts In the past three years, various discussions and workshops have been in full swing, with non-stop discussions every week, and it seems that the day of replicating the copper mold is just around the corner However, this optimistic prospect encountered an unexpected crisis and was eventually forced to pause because the characters restored by each person had very different personalities Although they were beautiful, they did not look like the same set of fonts It is not easy to develop the "common characteristics" of copper mold font restorers The long-running "Font Bronze Mold Restoration Project" After the failure in 2008, it was a huge blow to the type foundry Because they could not use these fonts, they felt that they were unworthy of the enthusiastic efforts of the volunteers and the most important copper molds continued to be damaged, especially the most important ones in the store The valuable "block script" copper mold was damaged a little more every time a character was cast, which made Rixing anxious The damage to the copper mold starts from the "missing corner", gradually breaks into pieces, and finally collapses In order to at least preserve the "appearance" of the font before the copper mold was completely destroyed, the calligraphy shop restarted the restoration project in 2016 With the assistance of several important volunteers and the Justfont font team, the most severely damaged "block script" type 1 and some "Song Script" type 1 and 2 were scanned and saved When resources are available, they can be "Scan the image file" to convert the "font file", and then use the computer to refine it After that, the 60-year-old boss slowly repaired more than 120,000 Japanese fonts at a rate of 5 characters a day In view of the fact that the pace of manual repair is far less than the rate of wear and tear of the copper molds, the type foundry has used more rigorous testing and selection to gather 3 to 4 talents who are willing to assist in long-term restoration In addition to re-carrying out font education and training, we also added "calligraphy" course training Most importantly, in order to develop a unified standard for calligraphy repair, these restorers must repair calligraphy simultaneously for months or years, and review the repair results every day in order to reduce errors and achieve consistency It is expected that three restorers will work together to perform long-term restoration of 5 characters a day with pre-training, it is expected to reconstruct a complete 4,500-character "block script" initial size font for traditional Chinese characters in 2 to 5 years How many days does it take for a calligraphy master to finish all the calligraphy Wending Technology assists, creating AI word-making assistant Wending uses the world's leading Chinese character creation technology and tools to assist calligraphy industries, and also promotes the AI value-added transformation plan of information service providers through the Industry Bureau's AI smart application service development environment promotion plan, and AI The award-winning new manufacturer Mukesi cooperated with RD to integrate AI technology to improve character creation productivity, thereby shortening development time and reducing costs In the early days, Wending required font designers to create each character from scratch, stroke by stroke It has evolved to the point where it can use existing character roots to compose characters and pre-assemble complete characters However, this preliminary pre-assembled character may have overlapping strokes and poor space and thickness It will also require the designer to spend a lot of time adjusting before producing a usable font product Through the AI value-added module, the system can learn some of the font styles that have been modified by the designer, and automatically adjust the structure, stroke thickness, etc of the remaining characters Finally, the designer can spend less time confirming the quality and With minor modifications, a usable font product can be completed, significantly reducing the time cost of creating characters Importing Wending’s value-added AI character creation system process-1 of 2 importing AI engineering technology With global font and cross-platform font technical services as its core, Wending Technology provides various font solutions to major manufacturers, system vendors, and government units around the world Taking the development of new fonts in the past as an example, we have completed a set of It takes a whole year to create a 10,000-word font The Industrial Bureau of the Ministry of Economic Affairs guided Wending Technology to cooperate with the AI startup Mukesi to learn the font style through AI It only needs to complete 5,000 words to automatically generate the other 5,000 The fonts are not yet created, and then the quality is confirmed and adjusted, allowing the designer to complete the entire set of fonts in less time, greatly increasing work efficiency by 50 In the future, we will continue to optimize the character creation module, allowing AI to complete more than 90 of font design, accelerating the production of new fonts Importing Wending’s value-added AI character creation system process-2 of 2 importing Wending’s character creation platform Wending Technology’s font innovation has been adopted by all walks of life For example, the 30th Golden Melody Awards used fonts for stage visual design, and Tsai Ing-wen’s presidential campaign team also used platform fonts for presidential election propaganda In 2019, it was added through AI Value-added transformation of the operating model resulted in revenue of NT15 million in the first year and is expected to increase revenue to over NT180 million within 5 years Smart font design service platform Using AI-assisted character creation to lower the threshold of font design, it can be transformed into a "smart font design service platform" in the future, providing designers with self-created fonts, and also serving corporate font design, helping designers achieve what is otherwise impossible The complete set of font development completed by an individual can also achieve the division of labor between design and development in the professional field of character creation, and become the first step to the success of font OEM, which will have a significant impact on the design and application of fonts The iFontCloud font library with AI added value has changed the original operating model, from being limited to font design by Wending Technology’s internal designers to breaking through the limitations of the original customer base and cooperating with external designers Establish and activate the ecosystem within the word-making industry circle Font products produced by AI value-added character creation process Wending cloud platform font management tool General Manager Wu Fusheng of Wending Technology said The Industrial Bureau has guided and participated in the AI value-added plan to demonstrate the empirical results It has continued to invest 6 million every year since 2019, and has invested a total of 30 million in AI technology research and development by 2023 Wending plans The next stage will be transformed into a "smart font design service platform" and the iFontCloud font library will be opened to all people who love words Everyone can create personal style fonts through the platform and can be applied in various fields It is expected that will create greater business opportunities The font products produced by the iFontCloud-AI value-added character creation process are sold on the Wending cloud platform「Translated content is generated by ChatGPT and is for reference only Translation date:2024-05-19」

這是一張圖片。 This is a picture.
【2020 Application Example】 AI Detection System Using Deep Learning, Detecting Irregular Polyhedral Defects in Just 0.5 Seconds!

Traditional manufacturing industries rely on manual visual inspection of products, lacking stability in quality yieldFor products made by traditional manufacturing industries, 'quality yield performance' is a critical issue and a decisive factor for customer business requirements Although many AOI vision inspection systems have been introduced in recent years, there are still numerous limitations that cannot be overcome when automating these inspection systemsFor example, the production of small quantities of diverse products, the inability to standardize irregular polygonal product dimensions, and the halo effect on glass or metal products from different lighting angles make it difficult to assist product yield filtering through AOI vision inspection, thus many traditional manufacturing industries still use manual visual inspection on their production linesManual inspection is labor-intensive and time-consuming, with expensive solutions from abroadA domestic model creation company often needs to manufacture products that are customized and diverse Although it uses imported high-grade mold equipment, product appearance quality testing is still largely done by manual visual inspection Testing standards vary by employee, and to adequately inspect the appearance of each product, the time each person spends cannot be easily controlled Often the same product needs to be examined repeatedly to meet quality standards, which is very labor-intensive and time-consuming and also sensitive to external environmental influencesAlthough the model company had evaluated adopting foreign AOI vision inspection equipment, a single set of equipment is expensive and only capable of inspecting certain types of product parameters, and lacks a learning feature to achieve diversified inspection goals, thus passive maintenance of the original plan is still necessaryCustomized solution significantly improves inspection efficiency and saves labor costsTo reduce the misjudgment rate of manual operations and operational costs, thus enhancing the competitiveness of the company's products, the model company sought assistance from 500HU Tech Ltd, hoping through customized service to leverage AI Deep Learning technology to improve the shortcomings of traditional AOI vision inspection systems, expanding the range of products that usable vision inspection systems can handle, and more accurately enhancing the accuracy of vision-inspected productsWith the support of the AI Innovation Research Center at National Central University, and based on the definition of five defect conditions provided by the model company, such as scratches, lint, white spots, damage cracks, and uneven baking paint, the initial step involved gathering a training dataset and manually replicating defect conditions on other parts and angles of the product, then using a program to generate defect images under different angles and lighting changes, followed by marking defectsThen, using software methods for training sets required by different algorithms, such as VGG, RestNet, Inception, DenseNet, Xception, SqueezeNet, target migration learning, classification problem Faster_Rcnn, SSD, Yolo, Mask_Rcnn, and other object recognition algorithms, after comprehensive consideration of accuracy and speed, SSD was chosen as the main core testing and inspection algorithmThen, the format of the training set required by the selected algorithm was produced, used as the comparative model then, using different AI frameworks, such as tensorflow, keras, practical verification tests were conducted, and verification test reports were produced Ultimately, optimal application parameters were adjusted for each product inspection, ensuring an average inspection accuracy rate of 95, with the inspection time reduced from 5 seconds to an average of 05 secondsOriginally, the model company's production process involved manual inspection followed by stamping a QC stamp on batches or sorting out defective products After introduction of this inspection system, the original process was maintained, but it sped up the manual judgment time, and during the process, recording for archival purposes took place, with defective items highlighted in red and recorded as photos, thus categorized into a 'defective-to-be-inspected' section Manual inspection would then determine if the product was qualified to move to the next inspection, significantly enhancing inspection efficiency and saving labor costsLow-cost, high-efficiency new AI inspection optionAs the technology of visual inspection by machines replaces human labor, it plays an increasingly vital role in the production of small, diverse orders, urgent orders, and situations where there is a labor shortage In contrast to expensive foreign inspection solutions, domestic providers can offer relatively cheap and customized solutions whether in terms of purchase costs or inspection efficiency, they are attracting more businesses ready to try, effectively enhancing the quality yield of manufacturers and thereby increasing competitiveness「Translated content is generated by ChatGPT and is for reference only Translation date:2024-05-19」

【導入案例】「AI指紋判讀模型」,利用AI將現場指紋進行數位化轉換與辨識,讓辦案追查更即時
【2020 Application Example】 AI Fingerprint Recognition Model, Using AI to Digitize and Recognize Fingerprints at the Scene, Making Case Investigation More Immediate!

Accurate and fast fingerprint identification, restoring innocence to the innocent 'Fingerprints' are one of the indispensable pieces of evidence at crime scenes At such scenes, numerous fingerprints are collected, including those of victims, related persons, and suspects After forensics collects 'suspicious fingerprints', it is crucial to exclude 'related persons' or 'victims' to prevent matching innocent individuals and thus, wasting forensic resources Initial fingerprint evaluations are labor-intensive and time-consuming According to a certain city's annual police statistics report for 2018, there were 43,558 criminal cases Automated Fingerprint Identification Systems are expensive to set up the NEC fingerprint recognition system currently used domestically can cost tens of millions As such, investing huge assets solely for fingerprint exclusion is not feasible Thus, forensic officers continue to manually compare fingerprints with the naked eye for exclusion, and only after exclusions are confirmed, the excluded items are logged into the 'Crime Scene Investigation and Evidence Room Management Information System' for future control before matching the fingerprints of 'suspected criminals' Based on current case data statistics, 90 of crime scenes involve 1 to 2 related persons and 1 to 5 suspicious fingerprints collected For a scenario with one related person and three suspicious fingerprints, it takes 15 to 3 hours to complete the exclusion process Considering the number of criminal cases in 2018, the exclusion process alone consumes a significant amount of time AI fingerprint reading leaves no place for criminals to hide The 'AI Fingerprint Recognition Model' developed jointly by Xinyang Technology Ltd and Glory Technology AI team imports all fingerprint evidence collected by forensics at the scene into the 'Crime Scene Investigation and Evidence Room Management Information System' Then, 'AI fingerprint comparison' is executed The AI fingerprint reading program automatically detects fingerprint areas and extracts features The system annotates the results based on the reading, confirming if the item can be 'related person excluded' With AI, identification can be completed in just 2 to 3 seconds per case, making the fingerprint matching process at the scene faster and more automated The process of excluding related persons accelerates the forensic timeline Integrating and establishing an electronic fingerprint database continues to optimize the AI fingerprint recognition model, enhancing case handling efficiency Through integrating and establishing an electronic fingerprint database and utilizing AI for fingerprint recognition, case handling efficiency can be significantly improved The part of 'Fingerprint Database Integration' usually involves managing cases within a city's jurisdiction To achieve horizontal linkage of fingerprints across all of Taiwan, it is necessary to integrate data from various municipalities, which can substantially improve the effectiveness of fingerprint technology in handling cases Additionally, 'Fingerprint Cards can be digitized' Currently, fingerprints are directly pressed onto paper, then scanned into digital files for subsequent processing If it were possible for individuals to directly press their fingerprints onto electronic collectors immediately, this would greatly enhance the timeliness of subsequent digitization The successes of this 'AI Fingerprint Recognition Model' are currently usable for police officers, but there are several aspects that continue to be optimized including 'Execution Speed,' especially when used across different cases, and 'Accuracy of Judgment,' since the current AI model provides a basis for the manual judgment of police officers Continuously fine-tuning the technology to ensure a consistent accuracy level could make it feasible to fully automate the exclusion process of related person's fingerprints「Translated content is generated by ChatGPT and is for reference only Translation date:2024-05-19」

【導入案例】AOI驗布員降低誤殺率,減少70複判篩檢量
【2020 Application Example】 AOI fabric inspector lowers the false negative rate, and reduced the re-inspection volume by 70%

Low detection rate, slow speed, difficult recruitment and high personnel costs The textile industry has always been a labor-intensive industry At present, almost all textile companies worldwide still inspect fabrics manually There are three major pain points in manual fabric inspection Low detection rate, slow speed, difficulty in recruiting workers, and high personnel costs On average, a fabric inspector can find up to 200 defects in one hour with a defect detection rate of about 70 However, inspectors are only able to maintain their concentration for 20 to 30 minutes at most, and their fabric inspection speed is generally limited to 20 to 30cms Fabric inspectors become fatigued if they exceed this time and speed Domestic and foreign AOI fabric inspection machines purchased by textile manufacturers have not yet been officially integrated into the production line At the beginning, 10,000 suspected defects could be detected in one roll of fabric The detection rate was high but the accuracy screening was low The number of suspected defects has been reduced to 7000, but is still not at the level of experienced inspectors High-speed cameras capture defects and record their locations The rule-based defect identification method currently used by manufacturers requires a lot of adjustment time about 1 to 3 months before the manufacturers site actually uses it, and there is currently no solution to automatically correct the identification model after use As a result, manufacturers need to spend extra time to adjust parameters Therefore, it requires considerable cost for both manufacturers and clients sites Current grew fabric inspection process of manufacturers The specific method used by the guidance team and cooperating manufacturers to implement AI identification technology and learning framework for model retraining into the defect inspection process is described below 1 AI-based defect identification model Utilizes the large amount of image data collected including fabrics with and without defects to construct the defect detection model through machine learning, such as SVM, or deep learning object detection methods, such as SSD or YOLOv3 This model is used to determine the condition of the surface of grey fabric and determine if it is a normal product or a defective product, thereby achieving defect identification 2 Identification model retraining framework If there is an error in the judgment of the visual inspector, the image will be marked and the data will be used in the dataset for re-training After a certain number of misjudged data is accumulated, the system will automatically start the identification model retraining function, and the new model that is generated will automatically replace the old recognition model, thereby achieving the purpose of model update Grey fabric defect inspection process after the implementation of this project Low false negative rate and solves the challenges of labor shortage and higher quality requirements in the industry This project uses a deep learning network architecture to reclassify defects that are detected, including real defects and false defects, and can further classify real defects and false defects to lower the false negative rate of traditional AOI solutions This is expected to reduce re-inspection volume by 70 and above for fabric inspectors, eliminate concerns about implementation in the current production line, accelerate the application of AI-based AOI solutions by textile manufacturers, and solve the challenges of labor shortage and higher quality requirements in the industry

【導入案例】「AI智能客服維修回覆系統」,用聊天就能即時解決客戶機台故障問題
【2020 Application Example】 AI Smart Customer Service Maintenance Response System, solving customer machinery fault issues instantly through chatting!

A tool machine manufacturer that markets successfully both domestically and internationally, but also faces challenges A domestic tool machine manufacturer specializing in CNC wire cut machines, CNC EDM machines, and CNC fine hole EDM machines, uses its strong core capability in electromechanical development to deliver high-precision, high-quality products It has successfully developed an aviation engine turbine ring wire cutting machine and specializes in designing and manufacturing super-large custom models, successfully marketing its products to over 30 countries worldwide Though capable of marketing high-quality products, the lack of standardized processes and methodologies for machine maintenance means that it often requires significant manpower and time to address machine failures, increasing maintenance costs No fast repair solutions, difficult personnel training, high maintenance time costs While the tool machine manufacturer can sell high-precision machinery globally, encountering maintenance situations always consumes a lot of manpower and money This is due to the lack of standardized troubleshooting processes for machine maintenance, mainly relying on the experience of maintenance technicians and the machine error codes Not all faults can be diagnosed through codes Technicians can only initially judge based on the error codes, then hypothesize the likely fault causes for further inspection and maintenance There is also no standard way to record the repair methods, making it difficult to quickly troubleshoot similar issues in the future In addition to 'lack of standardized fault troubleshooting process', there are also issues of 'difficult personnel training' and 'high maintenance time costs' Technicians need years of repair experience and must be familiar with mechanics, electronics, and mechanical engineering If error codes are not available during repair, it requires considerable time to identify the problem with the machine, causing significant time and cost losses Traditional way of addressing issues through email Implementing the 'AI Smart Customer Service Maintenance Response System' reduces costs for maintenance visits, shortens the duration of repairs, and simultaneously enhances the product's value Considering the pain points mentioned, the needs of the tool machine manufacturer are threefold firstly, establishing a 'fault troubleshooting AI image recognition maintenance knowledge base system' Then, collecting data on machine failures to establish a 'machine fault condition database' Lastly, integrating AI image recognition and deep learning functions to analyze photos taken at the time of the machine's failure in order to identify the most closely related fault issues and troubleshooting methods This 'AI Smart Customer Service Maintenance Response System' predominantly uses 'supervised learning' as its primary AI technique The 'AI model' part involves 'CNN' Convolutional Neural Networks, which is used for image recognition and obtaining extensive training data on machine malfunctions and recommended maintenance methods for effective AI predictions The 'data analysis' part uses 'DNN' Deep Neural Networks to acquire reference data related to fault conditions after training, providing answers that maintenance staff and clients desire for repairs, reducing the rate of maintenance visits and enhancing the product's added value Additionally, 'AlexNet' is used as a preliminary development tool its parameters can be set independently and executed automatically, ensuring that the AI model trained aligns closely with expected outcomes Currently, the tool machine manufacturer has around 10,000 graphic and text entries, predominantly 'image data' The system uses images for fault identification and text to assist in the diagnosis of abnormalities It employs '360-degree panoramic modeling' to archive graphic data and stores numerous image files internally Additionally, it gathers relevant data such as electrical currents, voltages, water pressures, and flow rates via sensors, utilizing them for associated decision-making processes The following pictorial representation shows the system service process AI Smart Response Customer Service System Service Process Chart This system gathers experiences from technical maintenance staff and information on machine faults to establish databases containing machine fault conditions, machine fault images, maintenance actions, and completions of machines It logs the comprehensive repair records, and leveraging AI image recognition and data analysis, it determines the most likely fault conditions Through accumulated maintenance experience, the machine is enabled to autonomously learn and decide, offering the most suitable solutions to technicians or clients, thus shortening the training and repair time for technicians, reducing clients' downtime and costs, and increasing the machine's additional product value Promoting the 'AI Smart Customer Service Maintenance Response System' across various industries for greater economic impact This 'AI Smart Customer Service Maintenance Response System' initially sets up a maintenance knowledge base, then employs Chatbot technology to integrate smart customer service, allowing clients to interact directly via chat to quickly resolve basic machine faults In the training of maintenance technicians, AI can also swiftly classify and inform of the likely fault causes and troubleshooting steps, thus lessening training and repair duration By effectively solving issues like the lack of quick repair solutions, difficulty in training personnel, and high maintenance time costs, it is poised to expand its applications to other industries for more significant economic outcomes in the future AI Intelligent Reply Customer Service System - Smart Image Recognition Customer Service Illustration「Translated content is generated by ChatGPT and is for reference only Translation date:2024-05-19」

rows
Rows:73, 9 pages