Improving automated inspection in manufacturing
Manufacturing quality control practices have long relied on visual inspection. Of course, visual inspection of products rolling off the production line is important. Visual inspections are also used for internal and external assessments of equipment including storage tanks, pressure vessels and piping.
Machine vision and artificial intelligence (AI) are making their way into production and manufacturing. Deep learning is providing faster, cheaper, superior automation for inspection practices. Most inspection processes take place at regular intervals, making automation ideal for the application. Areas where inspections are performed include:
- Automobile parts
- Electronic components
- Building materials
- Raw materials
- Medical supplies
Automated inspection vs. manual inspection
Automated inspection overcomes many of the limitations of manual inspection systems. In manufacturing, visual inspection errors take one of two forms. The first is the missing of an existing defect. The second is the incorrect identification of a defect. Misses lead to a loss in quality while incorrect identifications result in unnecessary production costs and overall waste. These errors are often traced back to the undependability of human vision alone, imprecision of eyesight, and cost of labor.
Automated inspection systems typically surpass the standard of manual inspection. Machine vision surpasses human vision in quality and quantity measurements because of its speed, accuracy, and repeatability. Machine vision systems can find object details too small to be detected by the human and inspect them with greater reliability.
Machine vision systems can also go beyond human visual acuity. Machine vision can view in the ultraviolet, x-ray, and infrared regions of the spectrum. On production lines, machine vision systems can inspect hundreds or thousands of components per minute.
Automated inspection and deep learning
With deep learning, machines learn by example. Automated inspection systems can recognize images, distinguish trends, and make intelligent decisions. Deep learning and machine vision enable a system to perform quality checks in great detail. Inspections are accomplished by means of image acquisition, preprocessing, and classification.
Deep learning uses thousands of layers within neural networks to distinguish anomalies, parts, and characters all while tolerating natural variations. Deep neural networks improve as they are exposed to new images, speech, and text.
Computer vision systems can be set up with some tolerances. But, systems without deep learning are limited. It’s the artificial intelligence that helps analyze complex surface and cosmetic defects, like scratches or dents on parts that are turned, brushed, or shiny.
Automated inspection system implementation
AI doesn’t require a lot of physical equipment. Hardware only requires a feeding system, an optical system, and a separation system. But, the software is robust. It requires advanced image analysis algorithms and heavy programming. The systems are often trained on thousands of images to detect meaningful deviations from the “standard” appearance.
AI and machine vision are taking over mundane and complex tasks like inspections. This lets humans focus on more sophisticated tasks. Costs of AI are forecasted to drop as efficiency increases. Machine vision and deep learning will also be an integral part of Industrie 4.0 as manufacturers look for new levels of efficiency and productivity.
This article originally appeared in Vision Online. AIA is a part of the Association for Advancing Automation (A3), a CFE Media content partner. Edited by Chris Vavra, production editor, CFE Media, email@example.com.