【2020 Solutions】 Fans and Motors Detect Abnormalities through AI, ASUS AI Solutions Team Exhibits Exceptional Capabilities
"Fans" are small but crucial components found in computers and home appliances. Though simple and comprising only a few parts, these components are key to maintaining 'coolness' and thus ensuring the longevity of machines. But how do fan manufacturers detect the quality of fans? The answer lies in 'listening' to the sound produced during fan operation to judge their quality. Issues such as axle misalignment, bearing anomalies, or blade interference can produce unusual noises, indicating that the fan is of poor assembly quality. Similarly, assembly quality inspection for other 'moving parts' like motors, engines, compressors is often carried out in a similar manner (As depicted below, auscultation testing of Ford RS engine assembly quality).
▲Auscultation testing of Ford RS engine assembly quality (Image source: kknews.cc)
Historically, fan industry inspectors must complete 3-6 months of 'auditory recognition' training before officially starting work, which takes place in soundproof booths similar to phone booths to ensure no background noise interference. However, after about 6 months (varies by individual), human ears become fatigued by specific sounds and require reassignment to other tasks. The ASUS AI Solutions Team leverages the fact that 'human hearing acuity decreases over time, but AI does not,' thus completely solving the pain point of continually having to retrain inspection staff.
AI Identifies Problematic Motors, ASUS Helps Monitor Quality for Businesses
To address the aforementioned pain points, ASUS uses recorded sounds of properly functioning fans to create 'good fan sound waveforms' for AI to learn. The co-general manager of ASUS Smart IoT Business Group, Quan-De Zhang, stated: 'This is not about matching sounds, which would require exact replication. Instead, for fans, AI needs to learn what constitutes a good category and what is considered an anomaly.' Additionally, ASUS's technology facilitates the training and creation of AI models, enabling their own PE engineers (production engineers) to perform model-building tasks, keeping the models within the company as a core competitive strength. For each new fan model introduced into production, only 3 recordings of 30 seconds each of the approved fan sounds are required to complete the AI model, swiftly integrating it into production. Several domestic 3C fan manufacturers are progressively implementing the ASUS Smart Waveform Detection Solution, and during its integration, ASUS supports training and optimizes AI models, reaching an inspection accuracy equivalent to quality control inspectors. They are happy to report that AI inspection can: 1. work continuously without interruption, 2. produce consistent quality, and 3. establish a digital production record for the products, facilitating future product tracking and process analysis—a long-term goal for fan manufacturers.
In addition to familiar 3C fans, the ASUS AI Smart Waveform Solution's sound anomaly recognition applies to all 'moving parts' and can also assess the assembly quality of motors produced in factories. Once AI 'hears' an abnormal sound from a motor, it can identify defective motors. Apart from sound, the same artificial intelligence solution also has the capability to recognize anomalies in electric current, voltage, and vibration waveforms. In environments with significant background noise or complex sound collection, detection can flexibly switch to monitoring electric current, voltage, or vibration waveforms to ensure quality inspection of assembled products.
AI Recognizes Digital Waveforms, Preventive Maintenance for Equipment
AI can also 'listen' to detect abnormalities in plant equipment motors, determining when motors need maintenance or replacement. Typically, maintenance checks on motors are carried out on a scheduled basis through inspections. For industries like steel or chemicals, where continuous production is crucial, any motor failure can cause significant material losses, massively outweighing the cost of the motor itself. Although larger motors are not readily available and unexpected downtime can extend for days or weeks resulting in substantial losses, many businesses use scheduled maintenance to replace motors earlier than necessary, opting for high costs over wasting entire production lines of materials or experiencing shutdowns. With ASUS's Smart Waveform Anomaly Detection Solution, all operating motors within a plant are monitored for anomalies, setting alerts to facilitate timely maintenance or parts replacement, achieving the goal of preventive diagnosis and maintenance.
AI Detection is Convenient and Rapidly Deployable
The environments for production and quality inspection vary significantly, and AI application development is highly customized. If the ASUS Smart Solutions Team had to customize AI for every single client, dealing with up to 200-300 suppliers and a variety of dispersed products, it would consume vast resources and be inefficient. Thus, 'reducing the customization ratio and increasing standardization' is another goal of the ASUS Smart IoT Team. They aim to reduce the industry-wide AI application customization ratio from 60-70% to 20-30%. In the future, every new project that ASUS Smart IoT Solutions handles will only require a 20% customization adjustment, allowing for replication and scalable deployment of solutions.
「Translated content is generated by ChatGPT and is for reference only. Translation date:2024-05-19」