Real-time detection of plastic shards in cheese using microwave-sensing technique

: Recently, Lidl had to set a recall action due to dangerous pieces of plastic found in the cheese products. The plastic shards, if swallowed, can cut the oral cavity or obstruct breathing. Current inspection techniques in the cheese industry are for the detection of metals using X-ray that does not offer a complete solution as many foreign bodies can go undetected. This paper demonstrates the use of a portable real-time microwave sensing technique for the non-destructive detection of plastic in cheese. The electromagnetic (EM) patch antenna was designed and tested on five Cheddar cheese samples. Different sizes of plastic shards, 1x10mm, 2x15mm and 5x20mm, were inserted into the samples and measurements were taken with and without the foreign objects. The initial results demonstrated that the patch antenna at 4GHz was able to detect and classify Polyvinyl Chloride (PVC) shards with an R 2 = 0.95. The initial results are promising and further investigation will be undertaken to detect different shapes and types of foreign objects in food products.


Introduction
The detection of foreign objects during the food inspection process is critical to ensuring the quality of food products [1]. Recently, Lidl had a recall action on Gouda Cheese products because of the risk of injury due to plastic shards in the cheese [2]. At present, foreign substances in food are detected mainly by using mechanical and optical methods as well as ultrasonic techniques. These techniques detect a large portion of the foreign substances based on their mass (Mechanical Sieving), their color (optical method) and their surface density (ultrasonic detection) [3]. Moreover, X-ray systems can reveal items made of hard non-metallic materials, namely stone, glass, bone, rubber, and plastic when embedded in food products. However, for soft materials (often organic materials) inside food products, accurate identification by X-ray systems is known to be difficult [4]. X-Ray method is an expensive and complex method, which is followed by complicated post-image processing procedures [5]. Cho and Irudayaraj [5] proposed the noncontact air instability compensation ultrasound transducers, which improved the velocity and thickness measurements of the calibration standard in varying temperatures. However, the recognition of differences between fragments and internal disorders in foods are still difficult.
Electromagnetic sensors are widely used and investigated in various applications and industries. Authors [6] used microwave spectroscopy for online non-destructive monitoring of meat drying and water holding capacity [7]. The electromagnetic sensors are also used for monitoring of the excess of moisture content in building fabrics [8], [9] as well as in public health for monitoring of insecticide level in developing countries [10]. The aim of this study is to use of a portable real-time microwave sensing technique for the non-destructive detection of plastic in cheese.

Experimental Setup and Methodology
The purpose of this investigation is to detect plastic shards in cheese using microwave technology, namely via utilizing a microstrip antenna. Five packages of 400g Mature Cheddar cheese [see Figure  1(a)] were purchased from a local Tesco shop for the investigation. Here is a list of equipment/items used for this experimental work: 1) Vector Network Analyser (VNA) 2) Plastic shards 3) 2.45 GHz patch antenna 4) Drill (to make holes for the plastic shards), Figure 2.

Samples and palstic shards preparation
PVC plastic was used as shards for this study. The plastic was cut into three different sizes, namely 2x15 mm, 1x10 mm and 5x20 mm, shown in Figure 1 The microwave measurements with the cheese samples were taken prior to inserting the plastic shards into them. Once the measurements were completed, the plastics shards were inserted into the cheese samples starting with the smallest size. A detailed measurement procedure is provided in sub-section 2.3.

Microwave Patch Antenna
A rectangular patch antenna was modelled and fabricated for this investigation. The antenna is presented in figure that was designed to resonate at 2.45 GHz frequency [11], which is an ISM

Experimental Setup and Data Acquisition
The patch antenna is connected to the VNA [see Figure 4(a)] to record a reflection coefficient. The calibration setting for this investigation is as follows: range 1GHz-6GHz, sweep points -4000, measurements -S11. The five cheese samples are unpacked and measured one by one by placing the sample on the patch area of the antenna [see Figure 4(b)]. Sample monitoring was implemented using Vector Network Analyzer (VNA) for attenuation of the electromagnetic wave propagation in the cheese. All the experimental results were recorded using Graphical User Interface designed in LabVIEW -the measurement time of one sample was less than 5 seconds. Five repeatability test for each sample was performed.

Results and Discussions
The microwave measurements were repeated five times for each cheese sample and the response then was averaged and analyzed in order to determine the plastic shards in the cheese samples. The full captured frequency spectrum (1-6 GHz, 4000 sweep points) from the sensor was scanned across using a bespoke LabVIEW program, namely to identify a linear relationship between the sensor response (S11) and the plastic shards (mm 3 ). The raw data presented in Figure 5, shows that the resonant frequency is shifted closer 1 GHz owing to the direct contact with the cheese sample. The noticeable change of the sensor response (S11) was between 3.5-4.5 GHz and the signal above 5 GHz was noisy, which is caused by a coaxial cable. The linear fit between S11 change at the frequency range of 3-4 GHz and plastic shards was more consistent and stronger compared to the rest of the spectrum (see Figure 6). The largest span (dB) can be seen in the frequency range of 3.7-4.3 GHz (the span above 5 GHz is not considered owing to the noise level), therefore the linear correlation at 4 GHz is presented in Figure 7, with R 2 = 0.95. The amplitude of the sensor response prior to inserting the plastic shards was -5.9 dB. Once the plastic shards were inserted, namely 1x10 mm, 2x15 mm and 5x20 mm, the amplitude decreased to -6.6 dB, -6.9 dB and -8.2 dB, respectively. The changes in the amplitude of the signal is thought to be caused by the change of the dielectric properties of the measured area of the cheese samples, i.e. the properties were altered by inserting various sizes of plastic shards, which have different chemical and dielectric properties compared to cheese samples.

Conclusion
To conclude, a microwave patch antenna resonating at 2.45 GHz was designed, constructed and utilized as a sensor to determine plastic shards in cheese samples. The microwave measurements were provided by S-parameters (S11, reflection coefficient) and repeated five times for each cheese sample prior to inserting plastics shards and with three different sizes of the shards inserted. The linear relationship between various plastic shards and an amplitude change of the reflected signal form the microwave sensor across the full captured frequency spectrum (1-6 GHz) was investigated. The data analysis demonstrated a strong linear correlation at 4 GHz, with R 2 = 0.95. This study presented a potential use of microwave sensing as a new technique to determine plastic shards in cheese products for health and safety purposes in the food industry.