Quality control is something many companies struggle with. For one company that packages sunflower seeds, almonds and peanuts, the need to offer a high quality product for a lower cost was imperative. Could automation help?
INO, a technological design and development firm for optic and photonic solutions, partnered with the firm to help them apply more stringent quality controls when sunflower seeds, almonds, or peanuts were removed from drying ovens.
INO’s partner, a company that asked not to be named, wished to reduce to a minimum the amount of foreign residue of various types that could make its way into product packages. The partner’s primary motivation was to offer a higher quality, less expensive product. The detection rate of the system previously used by the firm was generally around 30 per cent. Could the company meet a new objective of 70 per cent?
INO helped its partner by creating a hyperspectral sorting system capable of detecting maximum amounts of foreign material in real time during the processing of sunflower seeds, almonds and peanuts.
Using infrared spectroscopy technology
Infrared spectroscopy is an established technique widely used in various fields such as agrifood and pharmaceuticals. A number of laboratory research studies are looking at how visible and near-infrared spectroscopy can be used to detect defects or foreign materials in objects.
The core challenge called for INO to research observed values that discriminate various groups of different objects, using discriminant analysis and classification techniques.
INO developed a sensor in the range of visible and near-infrared wavelengths. The spectral analysis principle is implemented in the following manner: white light is projected on a sample, the reflected light is collected and split into its component wavelengths by a prism type optical element, and then a series of detectors measures the reflectivity at different wavelengths. The physical principle behind this technique is that the spectral signature of a type of material differs from that of another.
INO designed a camera to obtain real-time processing speeds that would be fully efficient in plant applications. The team also developed a special high-performance lighting system as well as powerful electronics to allow image processing at extremely high speeds. In addition, because the system must operate in a dust-filled environment, detection quality could not be affected by dust.
To inspect a complete object, a large number of points need to be spectrally analyzed. For this reason, the company used a spectral imaging linescan type sensor. Data analysis is performed using the FPGA/Power PC processing capabilities. By doing it this way, the sensor is capable of analyzing up to 15 different spectral bands simultaneously between 450 and 900 nm and to make real time decisions for the presence or absence of a specific feature. The hyperspectral imaging system can treat up to 2,000 profiles of 128 points per second. The system also has its own operating software, which INO developed.
Plant test results
After a series of preliminary trials was conducted in the lab, INO and its partner tested the system in the plant.
The system works as follows: first, seeds are directed from the container to a conveyer belt. An operator conducts an initial visual inspection for foreign bodies. The seeds are then sent down a chute to another conveyor belt. As they pass through this chute, the hyperspectral system detects contaminants present in the seeds: this is the optimum point for detection.
In-plant results show 90 per cent detection rate
To assess the sensor’s ability to correctly detect contaminants during production, the team measured detection and false alarm rates.
Those tests showed the sensor could achieve a real-time contaminated product detection rate of nearly 90 per cent without slowing plant production rates. The false alarm detection rate is relatively low at 0.3 per cent.
The advantage of such a system is its easy adaptation to various modes of production. By making a few modifications, the system could be used to detect foreign bodies in other materials, such as the presence of bones in fish or poultry, for example. Basically, sensor resolution is such that the slightest variation in product color is clearly visible.
Lastly, with high performance detection systems, users could purchase raw materials that have gone through less preliminary screening, thereby reducing production costs.
INO is a technological design and development firm for optic and photonic solutions for SMEs and large corporations. INO offers a complete range of integrated services in the fields of optics/photonics to clients of all descriptions in every field of industrial activity. Learn more at www.ino.ca.