The essential requirements for successful manufacturing are producing and shipping defect-free products. This sounds simple, but products must often meet a demanding array of technical, performance and cosmetic specifications.

Continuous manufacturing lines typically run in large campaigns at the highest possible speeds. Huge strides can be made with continuous process improvement, but the techniques, resources and sampling acumen of the inspection processes are the ultimate gates to production confidence and avoiding potentially expensive manufacturing upsets.

Films and Other Web-Based Products

A large number of industries such as textile, paper, semiconductor, plastic, engineered film, foil and other thin gauge metals and electronic materials supply web-based products that frequently have master roll format widths of 100cm or greater. Visual quality inspection of the large product areas in a continuously moving roll presents unique and difficult challenges. Demand for improved uniformity has fostered a growing need for and adoption of 100% visual inspection techniques to identify and cull a wide variety of random-variable defects.

(Learn more about machine vision systems with the IEEE Engineering360 guide.)

Why all the concern? Consider the specialized dielectric film produced to build multi-layer electronic circuits and insulate two or more circuit paths. Thinner and more complex circuitry is driving dielectric insulator sizes down to 12 microns (.013mm). With these fine circuit designs requiring parallel conductors at similar spacing, even a single metallic inclusion or foreign particle could result in short- or open-circuit failures.

Typical concerns include environmental inclusions, gel slugs, cracks, pits, scratches, discolorations and other product-critical flaws. Inspection systems typically have been a variation of these common approaches:

Increasing the number of inspectors/inspection stations and upgrading station lighting and magnifications.

Although common in many global industries with lower labor costs or narrower web profiles such as consumer electronics, the reliance on variable human vision and potential impacts from high line speeds, lighting and fatigue often results in critical misses of defects.

More rigorous sampling protocols for statistical validation. In theory an excellent approach, this method consists of long term data-based standard inspection procedures to lower sampling frequencies. But for a continuous production line, the drawback is the lack of real-time analysis. This can allow additional defective production while awaiting analysis, and also presents the problem of taking periodic samples from moving lines without impacting production.

Machine Vision Inspection Solutions

The smallest distance the human eye can discern is around 0.1mm, the diameter of an ultra-sharp pencil point or the average thickness of standard-grade copier paper. Many other factors must be considered such as visible light spectrum and diffraction distance, so this distance often increases. Production floors are typically more difficult for manual visual inspection then controlled lab environments. Even with enhanced inspection optics, the operator is challenged to identify and catch critical product defects, particularly with films moving at high speeds.

To address these manual visual inspection issues, manufacturers increasingly turn to automated machine vision systems to inspect materials produced in a continuous roll or web format. Camera systems and sensor configurations can be used to design a customized inspection, data capture, alarm and line control system specific to the output product, line speeds and critical defect parameters.

One of the core technologies driving machine vision solutions is line scan systems. At the technical level, line scan systems configure a line of one-dimensional sensor pixels to define a two-dimensional image. The second dimension results from the motion of the object being scanned. As the material moves perpendicularly across the line of pixels, two-dimensional images are generated and captured in a succession of single-line scans. According to machine vision suppliers, the advantages of line scan systems are:

• More cost effective on a dollar/pixel basis for large spatial image capture applications
• Better processing efficiency due to the elimination of frame overlaps
• Smear-free optical images without the need to install high cost strobe or specialty camera shuttering technology

Primary considerations for line scan systems are the need for uniform and high intensity illumination of the product scan area to compensate for short line scan integration times.

With machine vision performance requirements continuing to increase for higher resolutions at lower light levels, dual line scan camera architectures have been introduced into difficult applications. This technology combines improvements in pixel design to boost sensitivity along with rapid amplification of outputs. It also adds advanced photon management techniques to minimize the effect of photon shot noise, which is a limiting factor for accuracy in high-speed and low-light machine vision systems.

Achieving Machine Vision Design Success

Lighting: Key to creating contrast for the camera system, which serves as the main sensor input to the visual inspection system. Lighting must be constant and regulated to insure that light source changes do not feed false data. The inspection material must be the only variable in lighting changes. Poor lighting is a major cause of failure.

Choice of Optics: This is the heart of any vision system and includes one or more image sensors. Like any camera, each sensor requires a lens to gather reflected or transmitted light from the product being inspected. Knowing the precise field of view (FOV) as well as the working distance to the FOV is required for proper lens selection. Other variables include focal points and optical distortion allowances.

Sensor Conversion and Resolution Vision system optics convert light from the lens into electrical signals that are digitized into value arrays, also known as pixels. As noted, FOV and the number of physical pixels are primary design considerations and must be customized for each application. Pixel shape and appropriate high speed shutter exposures are necessary to “freeze” the camera images on a moving line, and to allow the system to provide instant analysis of the desired image capture.

Application Feasibility Study and System Specifications: Each application is unique in terms of lighting and desired performance response, so it is imperative that a fully comprehensive evaluation of the manufacturing variables be done up front with the system vendors and sub-contractors. A feasibility study can be a substantial undertaking, but it is a critical factor in identifying unknown variables and potential risks.

A statistical evaluation plan will typically be created to gather a large suite of images from an even larger sample base. The more samples the better in this critical design phase to make sure all possible inputs are tested with the actual system hardware and software. During this phase of the design, critical calibrations will be carried out specific to the anticipated line speeds, light levels and other key variables.

Another critical input response from the design team to the inspection system vendor is risk identification. Clear discussions at the start of the project on topics such as data gathering, facility upgrade planning, testing protocols and other existing and potential issues will minimize the risk of failure and downstream expense.

Automated inspection vision systems are an important tool for any web-based manufacturer looking to improve product quality, reduce costs and automate production. Such systems are 24/7 computers with eyes and can identify, inspect and communicate critical information that can be used to eliminate costly errors and improve line productivity.

Editor’s note: Art Creidler has most recently been a Product Development Specialist and Senior Product Manager for E.I. DuPont for their global flexible circuit materials business. He has been a leader in various DuPont product launches and commercialization growth in the military, aerospace and technology industries. He holds a Bachelor of Science in Mechanical Engineering degree from the University of Delaware.