1. Introduction

Vegetable oils contain essential fatty acids (EFA) and vitamins that are very beneficial to human nutrition and health benefits [1].Waste frying oil is a poor quality oil, in which contains excessive levels of peroxides, carbonyl groups, and propylene glycol [2,3]. Long-term use of poor quality oils can cause great harm to human health [4]. The vegetable oils are frequent subjects of fraud in which a relatively expensive oil is substituted or adulterated with waste frying oil. After vegetable oils mixed with waste frying oil, it is difficult to differentiate through some physical or chemical indicators. Conventional methods for the detection of waste frying oil include gas chromatography [5], conductivity measurement [6], and electronic nose technology [7]. Gas chromatography possesses higher efficiency and sensitivity, and can measure multiple components. However, this method is loss-prevention detection, expensive and complicated [8]. Electronic nose method is non-destructive testing and simple operation, but it is mainly used for the detection of some volatile substances [9]. Fluorescence method has many advantages such as high sensitivity, simple operation, rapid analysis, etc., and it has been widely used in the analysis of biological macromolecules [10–12].

Fluorescence spectroscopy technique now has the potential to replace or at least complement these classical methods mentioned above, and a number of works have been reported for detecting vegetable oils adulteration. Sikorska et al. determined the two fluorescence peaks of vitamin C and chlorophylls in vegetable oils based on fluorescence spectroscopy [13]. Poulli et al. identified the adulteration of corn, soybean and olive oils by synchronous fluorescence spectroscopy [14]. Mabood et al. used synchronous fluorescence spectroscopy and chemometrics analysis technology to identify olive oil [15]. Milanez et al. used fluorescence spectroscopy combined with the multivariate model for detecting the adulteration of extra virgin olive oil (EVOO) with soybean oil [16]. Their results show that satisfactory prediction can be obtained for all the regression models with Root Mean Square Error (RMSE) of prediction values. Tanajura da Silva et al. identified the classification of edible vegetable oils by fluorimetry and artificial neural networks [17]. Betül Öztürk et al. used molecular fluorescence spectroscopy to detect olive oil adulterated with sunflower and corn oils [18]. Other more, related reports have also confirmed that fluorescence spectroscopy is effective in the detection of oils [19,20].

In this paper, laser-induced fluorescence (LIF) has been used to determine the vegetable oils adulteration with waste frying oil. A LIF system was assembled in the experiment, using 405 nm laser as the excitation light source. The fluorescence spectra of vegetable oils and adulterated oils were measured. The identification of several important vegetable oils and the adulterated concentration were achieved by employing the principal component analysis (PCA) and partial least squares (PLS) model. The prediction errors of adulterated concentration was less than 2%. The results can provide reference for the quality identification of vegetable oils.

2. Samples and methods

2.1 Samples

The vegetable oils used in this work are rapeseed, peanut, corn and olive oils, respectively. The vegetable oil samples were purchased as commercial products from a local supermarket. Waste frying oil was selected from the shortening oil after repeated heating, and then was adulterated into different vegetable oils as the experimental samples. In this experiment, the frying oil was heated repeatedly for five times in the heating temperature of ∼200 ^oC, and each time lasts for one hour. The adulterated concentrations were changed from 5% to 50% with the gradient of 5%. Five types of original oil samples (rapeseed, olive, peanut, corn and waste frying oils) were prepared. Ten sets samples were prepared for each type of original oil, and four sets samples were prepared for adulterated oils of each concentration. Total 210 sets samples were measured in this work. All adulterated oil samples were churned by ultrasonic waves for 30 mins to make it uniform. In order to maintain realistic testing conditions, all the oil samples were stored in the dark at the room temperature until the time of measurement.

2.2 Methods

Figure 1 shows the used LIF system. The light source is a semiconductor laser with the output wavelength of 405 nm. The output power of laser is 60 mW, the diameter of laser beam is 5 mm. The laser beams transmit through an optical fiber with a numerical apertures (NA) of 0.22, and then irradiate into the cuvette filled with oil sample. The fluorescence spectra of oil samples were induced by incident laser, and were collected by a spectrometer through an optical fiber in the vertical direction of incident light to reduce the influence of incident laser. The gathered fluorescence spectra of different oil samples were analyzed by a computer. The resolution of spectrometer is 0.4 nm, and the integration time is 200 ms.

 

Fig. 1. Schematic of the experimental setup. OS: oil samples, OF: optical fiber, SP: spectrometer

Download Full Size | PDF

There are two reasons for using the 405 nm wavelength laser as the excitation wavelength: (1) the optimal excitation wavelength of fluorescence peaks in vegetable oils is about 400 nm; (2) since the emission wavelengths of fluorescence peaks move to the direction of long-wavelength band, all the fluorescence signals can be collected by using the 405 nm light source for avoiding the loss of the fluorescence signals of vegetable oil.

The analysis of the position and intensity of the fluorescence peaks was mainly used to determine whether the waste frying oil was adulterated into the vegetable oils. Firstly, the spectra obtained from the experiment were preprocessed for filtering and denoising of the signal. The second step was to use PCA to reduce the data dimension and remove the white noise. The last step was the prediction of adulteration concentration by employing a PLS model. Before the principal component analysis of vegetable oil, the collected fluorescence spectral signals need to be pretreated. The smoothing filter was used as the experimental pretreatment method, which can improve the smoothness of the spectra and eliminate the interference of noise and external factors.

3. Results and discussion

3.1 LIF spectra

Figure 2 shows the LIF spectra of five types of original oil samples, in which are marked with A-E, respectively. It can be seen that the fluorescence spectra of different vegetable oils are different from each other, the map gradually changed from pure vegetable oil to frying oil with the increase of the adulterated concentration. The intensities of the fluorescence peak around 500 nm increase with the increase of the adulterated content of frying oil, while the intensities of the fluorescence peak around 670 nm decrease with the increase of the adulterated content of frying oil. The peaks at ∼670 nm are attributed to chlorophylls which are usually used to identify the adulteration of olive oil. The olive oil presents the highest fluorescence peak around 670 nm, it indicate that olive oil contains more chlorophyll components than other oil samples.

 

Fig. 2. The fluorescence spectra of different oil samples: rapeseed oil (A), olive oil (B), peanut oil (C), corn oil (D), and waste frying oil (E).

Download Full Size | PDF

3.2 Statistical analysis

Figure 3 shows the PCA of six types of oil samples, where A, B, C, D, E, and F represent olive, peanut, corn, rapeseed, and waste frying and blend oils, respectively. There are four types of pure vegetable oils and one type of waste frying oil in this work, and 7 sample sets are prepared for each type of oil. 10 sample with different adulterated concentrations are prepared for each type of pure vegetable oil for PCA analysis, and there are 3 sets for PCA analysis. Total 300 samples were measured in this work. It can be seen that the first three factors of Z scores indicated 99.83% of the total variance (83.97, 14.99, and 0.87%). The PCA method was used to reduce the dimensionality of the data, and the oil samples were classified successfully. It can be seen from Fig. 3 that PCA can classify vegetable oils from each other effectively. Each type of different vegetable oils is significantly different after passing through the PCA process, and it can be seen from the figure that the four types of vegetable oils adulterated with fried oil gradually approach the location of the waste fried oil.

 

Fig. 3. PCA of six types of oil samples: rapeseed oil (A), olive oil (B), peanut oil (C), corn oil (D), waste frying oil (E) and blend oil(F).

Download Full Size | PDF

3.3 Quantitative analysis of the adulteration concentration

Figure 4 shows the fluorescence spectra of four types of vegetable oils adulterated with waste frying oil range from 5% to 50%, and the concentration gradient of adulteration is 5%. The fluorescence peaks around 500 nm are produced by the superposition of wide spectrum peak caused by several fluorophores. When the frying oil is repeatedly heated, some of complex compounds will be decomposed by high temperature, resulting in the increase of C = O fluorophores. So it can be seen that the intensities of fluorescence peaks around 500 nm increase with the increase of adulteration concentration of waste frying oil, but the intensities of fluorescence peaks around 670 nm decrease with the increase of adulteration concentration. For the fluorescence peaks around 670 nm, for example, the olive, the adulteration decreases the content of chlorophylls which resulting in the decrease of fluorescence intensity.

 

Fig. 4. The fluorescence spectra of four types of vegetable oils adulterated with waste frying oil range from 5 to 50%: rapeseed oil (a), olive oil (b), peanut oil (c) and corn oil (d).

Download Full Size | PDF

Four sets samples were selected in this work, in which three sets were used as the experimental groups and one set as the test group. The test group was analyzed and the data of the test group was predicted using the PLSR model. The results about the predicted and actual concentrations are shown in Fig. 5. The R² values of prediction linearity greater than 0.995 were obtained when four types of vegetable oils adulterated with waste frying oil with a mean square error less than 2%. The predicted values are highly consistent with the actual values. The results indicate that the method proposed in this work is feasible for the detection and quantification of vegetable oils adulteration with waste frying oil.

 

Fig. 5. The calculated concentration obtained from PLSR model vs actual concentration: rapeseed oil (a), olive oil (b), peanut oil (c) and corn oil (d).

Download Full Size | PDF

4. Conclusions

In this work, we have demonstrated that laser-induced fluorescence spectroscopy in combination with the principal component analysis (PCA) method and partial least squares (PLS) model can be used as fast and nondestructive method for the identification and quantification of vegetable oils adulteration with waste frying oil. Several kinds of vegetable oils adulterated with different concentrations of waste frying oil were successfully classified using the PCA method. The vegetable oils adulterated with waste frying oil range from 5 to 50% were predicted using a PLS model, and the prediction error was less than 2%. The experimental results verify the feasibility of the method for detecting whether the vegetable oil is adulterated with waste frying oil. Owing to the simple setup and the intense signals resulting from the visible laser excitation, LIF spectroscopy could be a powerful tool for the online detection for quality control applications.