Abstract
Polycyclic Aromatic Hydrocarbons (PAHs) are complex chemical compounds that occur naturally in unprocessed food when it is exposed to contaminated air during transportation, natural emission such as volcano, forest fire and through pesticides spray. It is reported by different agencies that there are 16 types of PAHs in which BaP (Benzo[a] pyrene), BaA (Benz[a]anthracene), BbF(Benzo [b] fluoranthene), Chr (Chrysene) are considered to be carcinogenic and it can occur due to different processes. In processed food it occurs due to various processing methods like overheating, incomplete burning, drying etc. The presence of PAH in food is conventionally found through analytical, traditional, and semi-automatic methods. These methods are found to be valuable but expensive and time-consuming. Further, these methods are used only for the detection of PAHs and the toxicity level is measured or identified based on expert knowledge of researchers and the Standards. Therefore, in this research, a simple harmfulness index system has been developed using Fuzzy Logic System(FLS). The proposed system has been designed based on the PAH values of different food and food products. Hence to initiate the study and to determine the significance of the results, PAH data have been collected from different articles that have investigated food products experimentally. These PAH data were analyzed using statistical measures such as Min, Mean, Max, Standard Deviation, Variance and Kurtosis method. Based on the observations from the results, the fuzzy sets were designed with four membership functions for each PAH and the rules were framed. The strength output from the inference engine has been associated with harmfulness index such as normal, low risk, medium risk, and high risk. From the evaluation, it can be observed that 89.72% of the food samples were recognized along with their degree of harmfulness. Also it can be inferred that 11% of the misclassified samples showed clear metrics of their harmfulness with PAH variations.
Keywords
Introduction
Polycyclic Aromatic Hydrocarbons (PAHs) are a large class of organic compounds containing two or more fused aromatic rings made up of carbon and hydrogen atoms [1]. There are various researchers who have studied the PAHs of different materials/substances and categorized their effect based on the molecular weight [2]. PAHs are classified as low molecular weight compounds containing up to 4 benzene rings and high molecular weight PAHs consists of more than 4 benzene rings [3]. The general characteristics of PAHs are high melting and boiling points (therefore making them solid), low vapor pressure, and very low aqueous solubility, the latter two tending to decrease with increasing molecular weight. The Physico chemical characteristics of PAH have a major influence on how they are absorbed or accumulated in food. The chemical structure of PAH also influences their existence in particulate matters. The PAH which are heavier are largely found in particulate matter whereas the light PAH are found in the vapor phase or intermediate position. The adsorbed PAH accumulated on leafy and non-leafy vegetables get distributed in human and other organisms when consumed.
The United States Environmental Protection Agency (USEPA) have reported that there are 16 PAHs in which BaP (Benzo[a] pyrene) is carcinogenic to human and BaA (Benz[a]anthracene), BbF(Benzo [b] fluoranthene) and Chr (Chrysene) are considered to be possible carcinogens [4]. These PAH have a higher degree of toxicity and naturally found in our ecosystem such as fresh water and marine sediments, industrial smoke, automotive exhausts etc. Unfortunately, these PAHs are also widespread in our surroundings such as garbage [5] and in the preparation and processing of food [4], all of which have a negative impact on the quality of life and can cause a wide range of diseases including cancer [3]. In addition, PAHs also occur by human activities like stubble burning, disposing sewage into the agricultural fields, having motor ways/air ways close to farm land increases the deposition of PAH on the surface layer of the soil, setup of industries, open fire place for domestic purpose increase the level of atmospheric PAH, burning of coal for power generation, smoking tobacco pollutes the kitchen atmosphere, oil spills during transportation through land and water [6–8].
Scientific Committee on food [6], reported that 15 PAH are genotoxic and carcinogenic in nature. To detect the occurrence and effect of carcinogenicity in food the EFSA (European Food Safety Authority) and SCF recommended using BaP (Benzo(a) Pyrene) as a marker until 2008. Later, in 2011 European Union revised the existing regulation by including PAH4 which is the sum of BaA, Chr, BbF and BaP as another marker along with BaP [9], since the concentration of other carcinogenic PAHs were found to be in higher concentration when compared to BaP [2]. Appendix A.1: presents the overview of maximum levels for BaP and PAH4 in foods, established by Commission Regulation (EC) No 1881/2006 with amendments (EC 2006) as given in Appendix 1. Therefore, analyzing these PAHs using signal processing algorithms and AI techniques provide more valuable information and systems to protect the ecosystem.
From the literature, it can be identified that the engineering technology can be employed to measure the PAHs value through different sensors such as gas and temperature sensors. Recently, different methods and techniques have been developed to determine the presence of PAH in various products and substances. Some of the methods used for deciding the presence of PAH in food products are namely Analytical method, Traditional method, Semi-Automatic method [10–17] (Refer Appendix A.2). In analytical method, presence of PAH is found and quantified using different techniques such as Gas Chromatography (GC) with flame ionization detection (FID), coupled to Mass Spectrometry (MS), High-Performance Liquid Chromatography (HPLC) with ultraviolet, fluorescence detection and coupled to Mass Spectrometry [4, 5]. HPLC method has been used to determine the existence of PAH in food products like meat, fish, smoked fish, puddings, biscuits and cakes, sugar and sweets, breakfast cereals, bread, white bread, sunflower oil, soya oil, corn oil, vegetable oils, lard and dripping, margarine, cheese, butter, chocolate, beer, skimmed milk powder, dried fruit and pulses, desiccated coconut. Using GC/FID, the concentration of PAH in different varieties of biscuits [3], grilled meat, smoked herring, smoked fish, kale, olives, grapes [5] beef, stripe, pork, chicken fillet [4] were determined and concluded that the levels of PAH found in these samples need to be monitored for possible effects in consumers. GC/MS have been used for the determination of PAHs in fish, mollusks, and shrimp, extruded wheat flour, smoked fish, dry infant formula, sausage meat, freeze-dried mussels, edible oil, wheat flour and different types of chocolates [4, 17] where it is suggested that effective food processing strategies need to be applied to curb PAH contamination in food. Traditional method and semi-automatic methods were used in determination of PAH in edible oils [19] where the measure of benzo[a]pyrene was low (<0.2–0.8μg/kg) in all oils except for a sample of sunflower oil which is 11μg/kg and synthesized the impact of edible oils and their effect in food products. At present, the PAHs are determined using SPME followed by gas chromatography/mass spectrometry (GC/MS) analysis in Selected Ion Monitoring (SIM) mode. Either the use of an auto sampler, or a manual approach have been used to perform the SPME extraction and the subsequent injection of collected analyses into the GC/MS [20].
From the review, it has been observed that the methods used for detecting PAHs were time consuming and expensive. Further these methods require more sophisticated laboratories, and these techniques were used only for the detection of PAH and to measure its concentration. Through mathematical modelling techniques and statistical variations, the PAHs are described as harmful or harmless with reference to the limits prescribed by the European Union for different products [20]. According to the regulation, the BaP concentration in meat and meat products should be less than 2μg/kg whereas for the cereal based processed food the BaP concentration should be less than 1μg/kg.
There have been some researchers who have used intelligent system like Artificial Neural Network and fuzzy systems in their investigation such as S. Bonvicini et al., 1998 [21] evaluated the uncertainty in the analysis of risk associated with the transportation of hazardous material through road and pipeline using fuzzy logic. The uncertainties pertaining to fire hazard relating to chemical substances and installations were studied using fuzzy logic and indices for these substances and their consequence were found by A.N. Paralikas et al. [22] in 2004. Health risk in consumption of produced water which contains naturally occurring radioactive materials, was studied by Shakhawat et al. 2006 [23] using fuzzy logic and concluded that it had minimum cancer risk and was within regulatory guidelines. In 2007, Jianbing Li et al. [24] developed a fuzzy stochastic model for the risk assessment of petroleum contaminated ground water where the contaminant of interest was xylene. Using fuzzy rule base, ingestion of xylene was methodically investigated for environmental-guideline-based risk (ER) and health risk (HR) to attain the general risk levels. Andrew Larkin et al. 2012 designed FNN to predict the effect of PAH-mediated perturbations of dermal Cyp1b1 transcription in mice [25]. Meghdad Pirsaheb et al. 2020 [26] modelled the PAH formation in grilled meat products using Artificial Neural Network and determined the accurate level of PAH in processed meat product. It is concluded that the concentration of PAHs depends on the type of meat. For instance, the concentration in poultry meat is 5–91 ng/g, duck meat is 106 ng/g and red meat is 0.1–547.5 ng/g.
According to Huyen Thi Thu Do et al. [27] the root cause of the risk involved in companies dealing with hazardous chemical substances are assessed using fuzzy logic semi-quantitative risk assessment, composite indicator where 77 industrial establishments were tested among which 18 were classified as high HCA risk, 36 establishments as medium HCA risk and 23 establishments as low HCA risk. Air pollution Monitoring system modelled by Saja Sattar Hasanh et al. [28] uses fuzzy logic system along with Arduino and gas sensors to detect the air pollution in the environment. It is concluded that fuzzy logic artificial intelligence gas sensors were used to detect the concentration of CO and CO2 and decide if the air is polluted or unpolluted.
From the literature, it is evident that PAHs are analyzed based on different AI techniques and risk assessments are identified but there are no research papers that correlates to the percentage of harmfulness or the toxicity level. Further, it is also clearly identified that there are no automated systems in general to predict the degree of toxicity or level of harmfulness of different products. Therefore, to develop an intelligent system and to detect the level of harmfulness in various products, a simple fuzzy logic-based system has been used. Figure 1 illustrates the proposed fuzzy logic-based toxicity index system. The proposed system has been developed based on statistical measures of 16 different PAHs along with expert knowledge and regulations set by several government agencies related to Environment & Sustainability.
The proposed fuzzy logic-based toxicity index system.
The subsequent section of this article provides a detailed explanation on the proposed Fuzzy Toxicity Index (FTI) system. The data gathered for the analysis of PAH is described in section 2.1. The analyzed data and their results are discussed in section 2.2. The design of the toxicity index system is presented in section 2.3 and the results are reported in section 2.4.
The main objective of this research is to use Artificial Intelligence techniques to detect the degree of toxicity of the food products based on the 16 PAHs, and also to relate its effects towards the quality of life. The proposed FTI system has been designed to detect the toxicity index of food products based on PAH content in them. The level of toxicity is indicated by an index number which depends on the percentage of harmfulness of the product. Thus, the food products are categorized into harmful or harmless. Hence, the input to the fuzzy logic system has been formed based on the quantitative analysis done on the PAH concentration of various food samples. The PAHs data collection and the analysis done are described in the following section.
PAH data collection
PAH contamination in unprocessed food is caused by exposure of food to different environmental pollution. In processed food, contamination is due to different food processing techniques like smoking, drying and different cooking methods such as grilling, frying, roasting, dry, and moist baking, deep frying, microwave cooking etc [29]. PAH contamination differs based on the duration for which the food is processed, type of fuel used and cooking temperature. Thus, the PAH data relating to food products were only considered in this study such as Meat and Meat Products/Fish Fishery Products, Baby Food, and Infant Formula/Processed Cereal Based Food, [30] Oils and Fats, Cocoa Bean and Derived Products, Plant powders used in the preparation of drinks. Table 1 represents the PAH data of the collected food products from different articles [3, 31].
Different food and food products for PAH analysis
Different food and food products for PAH analysis
Statistical measures obtained for each food category
The above 467 food products were considered in this study and categorized according to EU (European Union) set limits. There are 97 food products under category 1 which includes meat, grilled meat, smoked meat and meat products, smoked fish, herring fish and shellfish, mussels, smoked poultry and game [1, 18]. 184 food products under category 2 which includes baby food, infant formula, malt, bread [29]. 162 food products under category 3 which includes vegetable oils and frying oils [31], soybean oil, canola, grape seed oil, Sunflower oil which are used at different temperatures [7], butter and cheese [14]. 12 food products under category 4 which includes chocolates, coffee bean, coffee instant mix [18, 20]. 12 food products under category 5 which includes vegetable drink, green tea bags, adlay powder [14].
There are 16 harmful PAHs reported in the collected data set. These PAHS are Acenaphthene (Ace), Acenaphthylene (Aceph), Anthracene(An), Benz[a]anthracene (BaA), Benzo[a]pyrene(BaP), Benzo[b]fluoranthene (BbF), Benzo[g,h,i]perylene (BgP), Benzo[k]fluoranthene (BkF), Chrysene (Chr), Dibenz [a,h] anthracene (DBA), Fluoranthene(Flu), Fluorene(Fl), Indeno[1,2,3-c,d, 1,2,3-c,d]pyrene(Inp), Phenanthrene(Ph), Pyrene (Pyr) and PAH4 (Sum of Bap, BaA, Chr, BbF). It is also observed that for some of the equivalent food products, PAH level varies due to the different cooking procedures (like grilling, smoking, etc.), cooking temperatures, cooking time, fuel used, oxygen accessibility, fat content, and the drying process. The collected data were analyzed using different statistical measures, which is explained in the following section.
Statistical analysis has been employed in this study to understand ambiguity of harmfulness in the food products and to correlate the level of uncertainty. In the analysis of PAH data, six different statistical measures were considered they are Minimum (min), Maximum (max), Mean (
These statistical measures have been determined for all PAHs under the five different categories. Based on the above analysis, the results have been collected and analyzed through graphical tools such as time series and bar charts. The results collected from the statistical measures are demonstrated from Figs. 2 to 6.
Statistical evaluation for Category 1.
Statistical evaluation for Category 2.
Statistical evaluation for Category 3.
Statistical evaluation for Category 4.
Statistical evaluation for Category 5.
From Fig. 2 and Table 3, it can be inferred for category 1, the minimum concentration is 0μg/kg for Ace, Aceph, An, BgP, BkF, DBA, Flu, Fl, InP, Ph, Pyr and with 0.1μg/kg concentration for the PAHs BaA, BaP, Chr, PAH4. BbF has the least maximum concentration of 15.4μg/kg and PAH4 have the greatest maximum concentration of 265μg/kg. The average concentration for BaP is 2.73μg/kg with standard deviation of 16.339 and extremely positive kurtosis indicates the double exponential distribution where most of the sample concentration are ranging at the tail of distribution instead of being close to mean. It can also be observed that eight food samples exceeded the EU limit of 12μg/kg.
Statistical analysis of food products from Category 1 to Category 5
From Fig. 3 and Table 3, it can be inferred for category 2, the minimum concentration is 0μg/kg for BaA, BbF, Chr, PAH4, An, BgP, DBA, Fl, InP, Ph, Pyr and with –0.079μg/kg concentration for BkF and 0.23μg/kg for Aceph and Ace. BbF have least maximum concentration of 8.93μg/kg and Flu have the greatest concentration of 130μg/kg. The average concentration of BaP is 1.4038μg/kg with standard deviation of 2.9μg/kg indicates the samples are not much deviated away from the average value. The high positive kurtosis with 15.193μg/kg shows most of the samples are above the standard limits. It is also observed that 96 samples exceeded the safe limit of 1μg/kg.
Considering Fig. 4 and Table 3, the oil products of category 3 has the minimum concentration of 0μg/kg for BaA, BbF, Chr, Ace, Aceph, An, BkF, DBA, Fl, Ph and 0.39μg/kg for BaP and 2.4μg/kg for Pyr. The least maximum concentration is 0μg/kg for Ace, Aceph and the next least maximum concentration is 0.3μg/kg for DBA and the highest maximum concentration 43.3μg/kg for Ph. The average concentration for BaP is 2.612μg/kg with standard deviation of 1.627μg/kg which shows that the data are all around the expected value. The kurtosis value of 3.50 represents the data sets are normally distributed and in turn represents more toxic than normal.
Considering the fat products of category 3 in Fig. 4 and Table 3, the minimum concentration is 0μg/kg for all PAHs except for PAH4 which is 0.042μg/kg. The least maximum concentration is 0μg/kg for Ace, Aceph, An, Flu, Fl, Ph, Pyr and the next least maximum concentration is 0.3μg/kg for DBA and PAH4 has the highest maximum concentration of 5.9μg/kg. The average value for BaP is 0.124 with standard deviation is 0.265μg/kg which shows most of the data are around the expected value but extremely high positive kurtosis indicates the double exponentially distributed data with the samples at the tail of the distribution. Hence from the 167 products, 68 samples are above 2μg/kg for BaP and 3 samples are above 10μg/kg for PAH4.
It can be inferred from Fig. 5 of category 4 and Table 3, the maximum concentration is 0μg/kg for Ace, Aceph, An, BgP,Flu, Fl,Ph,Pyr and the next least maximum concentration is 0.06μg/kg for DBA and the highest maximum concentration is 10μg/kg for PAH4. The average value for BaP is 0.3442μg/kg having standard deviation as 0.8378μg/kg. Since the maximum concentration of BaP is 3μg/kg, the fourth moment is 11.894μg/kg.
Considering Fig. 6 of category 5 and Table 3, the minimum concentration is 0μg/kg for BbF, Ace, Aceph,An, BgP,BkF, DBA, Flu, Fl, InP, Ph, Pyr and 0.075μg/kg for BaP, 0.225μg/kg for PAH4. The maximum concentration is 0μg/kg for Ace, Aceph,An, Flu, Fl,Ph, Pyr and 0.403μg/kg for BaA, 1.968μg/kg for DBA. The average concentration is found to be 0.4628μg/kg for BaP, 0.5597μg/kg as standard deviation with kurtosis –0.347 indicates the less harmfulness of the product.
From the analysis, it is evident that the maximum and minimum portfolio higher moments are the appropriate range for deciding the marginal level. This in turn enables us to set the limits for the membership functions in the fuzzification. Based on the above analysis the data can be used for identification of level of harmfulness using fuzzy logic system.
Fuzzy logic aids in analyzing the level of risk involved in vague circumstances [32]. Fuzzy logic system can be designed to interpret the uncertainty using membership functions and the strength values can be calculated using fuzzy rules. The well-established Type-1 Mamdani Fuzzy Inference System has been used in this analysis. Since it is more intuitive, the construction of the rules are based on human expert knowledge. The 16 PAH data were considered as inputs for the fuzzification process, and the range of each input has been designed based on the quantitative analysis of the PAH data relating to the products. In this regard, each PAH fuzzy set have been designed with four membership functions namely Low Risk, Medium Risk, Maximum Risk and Extreme Risk. The concentration of Ace, Aceph, An, BaA, BaP, BbF, BgP, BkF, Chr, DBA, Flu, Fl,Inp, Ph,Pyr and PAH4 across the 5 food categories range from 0.23 to 16.41μg/kg, 0.23 to 23.17μg/kg, 0 to 34.78μg/kg, 0 to 26.7μg/kg, 0 to 157μg/kg, 0 to 15.4μg/kg, 0 to 13μg/kg, –0.079 to 11μg/kg, 0 to 21.67μg/kg, 0 to 9.63μg/kg, 0.1 to 130μg/kg, 0 to 24μg/kg,0 to 18μg/kg, 0 to 94μg/kg, 0 to 47μg/kg, 0 to 265μg/kg respectively.
From the data collected it is observed that food products under category 2 were found to have more than five PAHs and the EU limit for BaP and PAH4 is same for this category i.e., 1μg/kg. Based on the analysis, the range of the first membership function has been built with [0, 0.5], the second and third membership functions were designed in between [0.25, 0.75] and [0.5, 1.5] respectively. The fourth membership function has been chosen from 1 to maximum concentration of the product for the PAHs Aceph, An, Ace, BgP, BkF, DBA, Flu, Fl, Inp, Ph, Pyr. Similarly, the range of the membership functions are set for other PAHs. The output of the fuzzy system has been associated with the degree of the harmfulness of PAH in food which is represented by the four membership functions Normal, Low risk, Medium Risk and High Risk. The range of the Normal Membership function is [0, 25], for Low Risk it is [12.5, 37.5], for Medium Risk it is set as [25, 50] and last membership function is set as [45, 100].The fuzzy membership functions are illustrated in Fig. 7.
Membership Functions of Fuzzy input and output.
In the fuzzification process, 16 different fuzzy sets were formed based on the 16 PAHs as shown in Equation 1
For example, a fuzzy input of category 2 food product is (0.22, 0.16, 0.14, 0.61, 1.13, 1.6, 6.1, 1.4, 4.2, 0.18, 1.6, 3.4, 11, 9.1, 2, 2.3) then the crisp value can be calculated as shown in Equations (2 to 6).
Similarly, the crisp values of other PAHs are calculated in the designed system.
In the validation process, it is observed that according to the EU limits, there are 295 samples from 467 samples that were classified under harmless category of food and 172 samples were classified under harmful category of food depicted in Fig. 8.
Statistics on harmful and harmless data as per EU limit.
Of 295 harmless categories, 89 samples are from category 1, 88 from category 2, 94 from category 3 and 12 each from category 4 and category 5. Out of 172 harmful samples, 8 are from category 1, 96 from category 2, 68 from category 3 and none were found in category 4 and 5. Therefore, there are either harmful or harmless categories of food, but we have designed the fuzzy logic toxicity index system to indicate the degree of harmfulness. So, 150 fuzzy rules are framed to map the 16 harmful PAH found in food with the degree of harmfulness of PAH. With reference to EU limits as well as bench markers which are established by EU, PAH4 and BaP were used in framing the initial 144 rules by comparing BaP with BaA, Chr, BbF, PAH4, and comparing PAH4 with BaA, BaP, Chr, BbF and by comparing Bap with PAH4. For example, BaP and PAH4 having 4 membership functions each, when compared resulted in 16 IF –THEN rules as shown in Table 4. Six fuzzy rules were also included to estimate the index level based on the PAHs level of the food products from category 2, 4, and 5.
Fuzzy inference rule
Therefore, the fuzzy inference engine has been developed with 150 if-then rules in total to determine the strength of the crisp output. The strength value is further defuzzied and associated with the level of harmfulness, the degree of harmfulness in the defuzzification process has been estimated through centroid method.
The fuzzy logic-based toxicity index system has been developed and the data collected from different sources were used for the evaluation. The 16 PAHs of 467 samples with 295 harmless samples and 172 harmful samples are passed through the fuzzification process and the fuzzified input is synthesized through the inference engine. The strength output from the fuzzy logic toxicity index system is defuzzied and the degree of harmfulness is related to a specific index level which ranges from 1 to 10 to decide on the percentage of harmlessness of the PAH.
The food products with index level 1 to 5 are considered harmless and the index level above 5 are considered as the harmful category. In the evaluation of fuzzy logic toxicity index system 467 samples were used in the fuzzy logic system in which 291 samples were classified as harmless from 295 actual harmless food products. From this it can be inferred that 98.64% of the harmless data were classified precisely. The fuzzy output ranges from 0% to 50% as indicated in Fig. 9.
Statistics of harmful and harmless data from fuzzy toxicity index system.
These samples also had their index level below 5. 122 samples were classified under harmful category of food products from 172 actual harmful food products. From this, it can be inferred that 70.93% of the harmful data were classified precisely with the output ranging above 50%. The fuzzy logic toxicity index system results are tabulated in Table 5 and are also illustrated in Figs. 10 to 13 for each category of food.
Confusion matrix of fuzzy logic toxicity index system output
The distributed fuzzy output range with the corresponding number of samples are given in Table 6.
Fuzzy output
Among the wrongly classified 54 samples, 50 samples are classified by fuzzy system as harmless as their index level indicates only 30% harmfulness for 36 samples, 40% harmfulness for 4 samples and 50% percent for 10 samples. The wrongly classified samples are reported in Table 7.
Fuzzy harmless index table
The remaining 4 samples which was classified by fuzzy system as harmful and their index level shows 50% harmfulness for 3 samples and 60% harmfulness for 1 sample as given in Table 8.
Fuzzy harmless index table
The range of PAH concentrations for the 36 samples in fuzzy harmless index table is as follows: BaA concentration ranging from 0–14.51, BaP ranging from 0.03–17.46, BbF ranging from 0–8.92, Chr ranging from 0.04–16.67, PAH4 ranging from 0.05–29.31, Ace ranging from 0.31–16.41, Aceph ranging from 1.08–23.17, An ranging from 0–34.78, BgP ranging from 0.68–2.07, BkF ranging from –0.079–5.62, DBA ranging from 0–9.6, Flu ranging from 1.18–49.89, Fl ranging from 0–21.25, Inp ranging from 0–11.27, Ph ranging from 0–51.95, Pyr ranging from 0.31–17.63.
Regarding 4 samples in fuzzy harmless index table, BaA concentration ranges from 7.07–14.54, BaP ranges from 2.11–10.22, BbF ranges from 0.47–8.93, Chr ranges from 1.5–2.3, PAH4 ranges from 2.34–27.53, Ace ranges from 10.01–16.41, Aceph ranges from 15.83–23.17, An ranges from 24.65–31.77, BgP ranges from 1.14–1.16, BkF ranges from 1.88–3.53, DBA ranges from 0.6–2.79, Flu ranges from 3.38–38.94, Fl ranges from 1.33–4, Inp ranges from 3.4–7.18, Ph ranges from 3.21–34.17, Pyr ranges from 1.53–17.63.
For the 10 samples whose fuzzy index is 5, the BaA concentration is from 0.087–26.7, BaP is from 0.1–17.46, BbF is from 0–9.63, Chr is from 0–15.54, PAH4 is from 0.33–53.92, Ace is from 3.05–12.97, Aceph is from 1.09–34.21, An is from 3.5–20.11, BgP is from 0.2–0.88, BkF is from 0–5.59, DBA is from 0–9.57, Flu is from 0–8.56, Fl is from 0–21.25, Inp is from 0.117–3.23, Ph is from 0–34.35, Pyr is from 0–3.48 as depicted in Figs. 10a and 10b.
% of harmless based on the defuzzification.
Index graph for harmless food products.
It can be also observed from Fig. 10a & 10b be that the harmless food products of all categories range from 12.5% to 52.7852% and index ranging from 1 to 5 respectively.
For the 4 samples in fuzzy harmful index table, BaA concentration ranges from 0.74–26.7, BaP ranges from 1.13–17.46, BbF ranges from 3.35–9.92, Chr ranges from 0.25–15.54, PAH4 ranging from 6.63–53.92, Ace ranging from 3.05–12.97, Aceph ranges from 1.09–34.17, An ranges from 3.5–34.78, BgP ranges from 0–0.88, BkF ranges from 0–5.59, DBA ranges from 0–9.6, Flu ranges from 1.18–49.89, Fl ranges from 0–21.25, Inp ranges from 0–11.27 as shown in Figs. 11a and 11b.
% of harmful based on the defuzzification.
Index graph for harmful food products.
It can be also observed from Figs. 11a & 11b that the harmful food products of all categories range from 56.5898% to 72.38% and index ranging from 6 to 8 respectively. The effect of these results is due to the unavailability of certain PAHs in the samples. The fuzzy logic toxicity index system has identified these data as harmful since the index level is above the threshold level of 5. Thus, it can also be considered as a low level of toxicity.
From the evaluation, it can be validated that uncertainties pertaining to determination of toxicity in various food products can be identified by the design of fuzzy logic toxicity indexing system and attained the maximum accuracy of 89.72%. The remaining 10.28% corresponds to 54 samples which was discussed in the previous paragraph.
The performance of fuzzy logic-based toxicity index system has been investigated in this study based on the statistical measures. The 467 samples collected based on five different products are categorized into harmless and harmful using the level of PAH prescribed by the European union (EU) standards. In comparison to the EU based harmless description of the food products, the proposed fuzzy toxicity index system attained 89.72% of accuracy and it is evident that system is robust regardless to the level of PAH, using the proposed methodology. Further, there are different methodologies [21, 28] that has been employed to analyze the PAHs using artificial intelligent techniques but detection of harmful index are very limited. In particular Huyen Thu Do et al. [27] analyzed hazardous chemical substances using fuzzy logic to formulate semi-quantitative risk assessment and Saja Sattar Hasanh et al. [28] designed an Air pollution Monitoring system using fuzzy systems. Therefore, research in the field of harmful detection using artificial intelligence systems is limited and the results obtained using the proposed methodology is distinct in this area of research. With regards to the objective of this research, the results obtained from the evaluation were found to be efficient and provide significant results for the further development of a generalized prototype model. Also, this study has given confidence to use the fuzzy logic system for the carcinogenic analysis and toxic nature of different products.
In future, it is proposed to develop a harmfulness detection system using sensor technology and signal processing methods. There are sensors to detect the basic properties such as temperature, vapor pressure, gas sensors. The parameters can be used through different analytic techniques and quantify the PAHs, instead of processing or analyzing the materials/food/water through gas chromatography-mass spectrometry (GC-MS) or high-performance liquid chromatography-ultraviolet (HPLC-UV). The quantified data can be characterized through our proposed fuzzy based toxicity index system.
