High-Speed Camera System Ensures Sanitary Product Quality

Author(s):
Martin Balog - Datalan

Industry:
Consumer Goods

Products:
Software, FPGA Module, PXI/CompactPCI, LabVIEW, Machine Vision

The Challenge:
Developing a highly reliable inspection system for the production of sanitary products.

The Solution:
Using National Instruments LabVIEW software to develop TIViS, a flexible industrial machine vision system capable of rapid inspection and future development.

Sanitary products intended for personal consumption such as feminine hygiene products and nappies for children require both high-quality material and high-quality processing during manufacturing. Because several types of defects may lead to unhygienic products that, in critical cases, may result in injury, it is very important for producers to achieve the maximum possible quality of the final products.

Output inspection of such production processes has traditionally been performed through the random selection of products, which are manually examined in the laboratory. This is not enough to provide a 100 percent result. The use of automated inspection systems based on machine vision can achieve this goal.

Quality inspection of sanitary products imposes stringent requirements on the vision system. The first requirement involves product inspection at a high speed because the capacity of one assembly line is usually more than 1,000 pieces per minute. The second requirement is inspection at several points along an assembly line because the products are relatively complex and problems may occur within internal layers. The last requirement involves the materials used. These mostly include white nonwoven textiles and foils, which complicates considerably the ability to recognize all the necessary details of a product and differentiate errors.

System Design

The design of our feminine sanitary towel production inspection system was based on the customer’s demanding system requirements, including a high rate of inspection and  reliability. The objective was to determine the effectiveness of the identification system in detecting defective products while preserving an absolute minimum of false alerts, which increase the ratio of waste in production and represent a direct financial loss for the customer. Crucial for the customer were the requirements for data acquisition and statistical processing, evaluation of undesirable events, and reporting options. On the whole, we based the system on an open architecture to incorporate potential future requirements and expand application possibilities.

To develop a system that met the defined requirements, we pursued a PC-based, machine vision system. We chose NI LabVIEW to develop the application because of its real openness and options to apply a wide range of functions from other areas including data acquisition, graphical presentation, statistical processing, and reporting. The possibilities for designing a graphical user interface were practically unlimited, which enabled the design of a comfortable, transparent, and fast application.

We based the application’s internal structure on the framework we implemented within the majority of our complex machine-vision applications. The framework guaranteed basic functions for data flows, image acquisition, synchronization, data acquisition, and statistical analysis; statistical process control; and data presentation in real time. The aforementioned framework provided the customer with a verified concept of very simple and transparent data control and processing.

The system also needed to acquire images of the inspected products at two different locations on the assembly line. At the first location, individual layers and cores that form final products were inspected prior to being cut out from a continuous strip of material. The second image acquisition point was located just before the packaging of products into foils. Behind the second camera, the assembly line removed defective products from the manufacturing process before their packaging by means of an air jet. To work out an exact real-time synchronization, we needed to develop a specialized solution for real-time tracking control.

LED panels, used to provide product illumination, required more work with respect to the design and selection of an accurate type of light source. After thorough analysis, we selected custom LED panels, specifically red LEDs. These provided maximum possible contrast with regard to the selected material and were one of the key elements to achieve highly reliable product quality assessment. Moreover, we designed both the construction of lights and type of LED diodes to meet exact assembly line conditions.

To synchronize the machine-vision system and main programmable logic controller (PLC) of the assembly line, we chose a National Instruments PCI-8254R image acquisition board. The board provided a number of important functions and characteristics that we needed for our system. A big advantage was the combination of transistor-transistor logic (TTL) signals, used as an external trigger for AVT digital cameras, and optically isolated digital inputs and outputs. We also used the isolated channels when communicating and synchronizing with a PLC, as required by the customer. The most substantial benefit of the NI PCI-8254R board was the built-in FPGA. This technology fully met the requirements for the management of image acquisition and real-time synchronization with PLCs.

Tracking Control and Synchronization

Tracking control, which played a key role in the system, provided exact monitoring of each inspected product starting from the moment when a product reached the first inspection location and ending with its final packaging or removal from the process. The tracking control had to react to smooth changes in the speed of the processing machine, its considerable maximum speed, as well as the required accuracy of the camera synchronization via an external trigger.  

Using the LabVIEW 8.20 FPGA Module, we implemented the TIViS tracking control system in an FPGA on the PCI-8254R. The basic input signal for the system was the information indicating that a new product to be inspected was at the position of the first camera. At this point, the product was recorded in the system under a unique ID. The tracking control system then used the setup delay for the image acquisition initiation of the first camera. The system also received a similar signal from the PLC in the second camera position. Because of the information about the distance between these two positions and the speed of the processing machine, we could determine when each product was located on the second camera position. The image acquisition was again synchronized in terms of hardware via the external trigger of the camera. The results from the inspection of each camera were subsequently combined and, with one unique ID, returned back into the program in the FPGA. The moment the tracking control system returned the result from the inspection to the main PLC of the assembly line, the system sent reject signals for those products that failed to meet the required quality criteria. Considering the high speeds the assembly line reaches, the accuracy of the reject signal timing was crucial to prevent false air-jet removal of nearby products that were defect-free.

The tracking control program ran on an FPGA in several parallel loops that enabled an immediate response to input signals and, thus, provided for a real-time flow. Because of the asynchronous communication with the main application, which was running on the Microsoft Windows XP Professional SP2 operating system, the tracking control program did not have to meet the parameters of a real-time application and could be optimized for overall performance.

Image Processing, Defect Detection, and Algorithm Optimization

We had to apply a specific approach in the development of algorithms for the evaluation of the required product parameters and the detection of individual defects. First, the process included very high-speed image processing (up to 2x20 images per second). More importantly, many evaluated parameters required complex image processing. It was necessary to design special optimized algorithms for certain areas because the results could not have been attained by using standard machine-vision tools.

The first step in our image processing concept was illumination optimization. With the correct LED lighting choice and design, we attained a sufficient contrast increase on key positions of the products, for example, wings and individual product cores. Without this optimization, we could not have procured both the evaluation of some product characteristics  and the identification of critical types of defects that we needed to eliminate from the final production.

For each camera in the system, we developed algorithms on an individual basis in several iterations. For the image processing general prototype, we used NI Vision Assistant 8.2.1, part of the NI Vision Development Module 8.2.1. We used standard tools to verify identification possibilities and evaluate required product characteristics. We modified some partial algorithms several times to increase evaluation reliability. 

We also optimized the prepared prototypes. With such a complex evaluation level for several parameters, the standard functions and operators from libraries failed to produce an adequate output. Additionally, we had to develop specialized algorithms for certain parameters such as the assessment of the bending of wings. In this respect, we used an adjusted version of Hough transform, which identifies the low-contrast edge of wings vis-à-vis background material and evaluates its location, length, and angle.  

We rewrote a major part of the algorithms  in the Visual C++ language into a DLL. In addition to the DLL source codes, we used the OpenCV and IPL 2.5 libraries in the development phase. With those, we implemented some necessary low-level algorithms optimized for Intel Pentium processors. We also gained several evaluation and image processing functions, in particular pattern matching and some gauging functions, from the Vision Development Module libraries.

The appropriate combination of the LabVIEW and C++ libraries led to the development of very reliable and high-performance algorithms that fully evaluate the quality of defined products.

User Interface and Other Functions of TIViS

Because the overall application runs in LabVIEW (except for some algorithms), programmers can develop user interfaces. The user interface mainly provides functions such as image visualization, visualization of computed values and statistical quantities, setting of inspection parameters for individual product types, and overall application system settings.

We designed our interface to be ergonomic, well-arranged, and easily manageable. The main screen was divided into several parts that addressed a certain area of usage – Online, List of Articles, and Setup of Articles. Moreover, we could enter the system settings of the application and FPGA. Several control components as well as basic application status information were always displayed in the upper part of the window.

Online included the image display of products as well as results and statistical values. They provided a comprehensive summary of the inspection and presence of defects. A very useful function was the switching between the display of each product and the display of images of the last defective product. Because there were not many production defects, the operating staff was able to continually monitor the images of defective pieces and, thus, acquire a better overview of the production status. The computing results were displayed in different ways, for example, a table, a graph of actual values for a certain period of time, a graph of time-based error rate, and so on. These results were very important for the operating staff in terms of the processing machine setup and the maintenance of production process stability, which improved the effectiveness and productivity of the facility.

With List of Articles, it was possible to establish a database of various setups for different types of products. Whether we were dealing with different products or just different setups and variants of the same product for different customers, TIViS helped us define an unlimited number of items. Following the initiation of the inspection, we could simply select the correct article (automatically or manually), and the operating staff did not have to laboriously set up each parameter. 

Setup of Articles defined the processing and evaluation setups of individual computed product parameters. The well-arranged and interactive setup could work in online and offline mode. In the online mode, the actual image from the assembly line was taken three times per second, and its parameters were calculated based on default values. Because of this, operators could change setups in real time without having any direct impact on live production. If operators were satisfied with the performed change of parameters, they could apply them, and all products would be inspected according to the modified configuration. If operators did not apply the changes, the initial settings remained effective. Offline mode enabled the setup of an arbitrary item with the use of images saved in the database on the hard disk of the computer. It was also possible to perform such a setup while the inspection system ran at full capacity in the background.

Benefits and Advantages of TIViS

TIViS was designed to fulfill the most demanding customer requirements and provide an added value in the form of various visualization tools and functions for data processing and archiving. It offers an optimized solution for all customers, in particular through accomplishments in special algorithm development for each specified type of product. It is this approach that makes it possible to provide maximum reliability as well as an adequate rate of data processing and evaluation.

Because of National Instruments tools and components, the system is open to further expansion and functionality adjustments for new types of products (nappies for children, other types of hygienic towels, and so on) as well as in manufacturing process management, SPC statistical processing, watchdog (monitoring and signaling of undesirable risks), and integration with other information systems.

Author Information:
For more information on this Case Study, contact:
Martin Balog
Datalan
Slovakia
Tel: 421 2 5026 7519
Fax: 421 2 5025 7700
martin_balog@datalan.sk

Browse All Case Studies »