ارزیابی خطرات بهداشتی مواجهه تنفسی با بنزن به روش شبکه عصبی و شبکه عصبی فازی بر روی ظرفیت های تنفسی و پارامترهای خونی در یک صنعت شیمیایی

Benzene is a aromatic colorless liquid is produced as a primary component of crude oil and gasoline as well as as a by-product of the coke and coal industry, which is used in various industries including chemical, pharmaceutical, polystyrene, synthetic rubber, nylon, detergents, Paints, polishes and solvents are used in laboratories.Complications of chronic exposure to benzene, decreased body hematopoiesis, impaired immune system as well as leukemia, anemia, respiratory disorders, delayed ossification of the human fetus, infertility, production of lymph node tumors and injury to the liver. Prolonged or repeated exposure causes damage to the lungs, kidneys, liver, spleen, blood, brain and endocrine glands. Prolonged exposure to benzene has destructive effects on the tissues that make up blood cells, especially bone marrow cells. Benzene vapors in high concentrations have a irritating effect on the mucous membranes of the eyes, nose and respiratory tract, and if they enter the lungs, they can cause severe swelling of the lungs and may cause death. Inhaling benzene vapors in the air is the main way to expouser with benzene. Benzene has been described as a useful and dangerous chemical by many reputable organizations, including NIOSH, OSHA, and ACGIH as a definitive carcinogen for humans. Therefore, due to the definite carcinogenicity of this substance in humans and widespread use in industries and destructive effects on various organs of the body, especially the hematopoietic and respiratory systems, as an innovative and practical research on the health hazard of respiratory exposure to benzene on respiratory capacity and parameters Blood in a chemical industry was evaluated by neural network method and fuzzy neural network method. This case study was conducted in 1398 on a statistical population of 50 people working in a chemical industry in four exposed groups (including personnel of loading, production, laboratory and quality control units) and a blank group (personnel of administrative unit) Took. Due to the fact that men are employed in these units, only the group of men was selected and one year of work experience in the mentioned units was selected as the minimum acceptable time to enter the research. The 12-hour staff shift and the type of ventilation system in the studied units are natural and using appropriate personal protective equipment by the staff. Pulmonary volumes and capacities are measured by a spirometer that can be analyzed to assess respiratory function. The most important main lung capacities that cause lung diseases (obstructive pattern, restrictive pattern and mixed pattern) include forced vital capacity (FVC), forced expiratory volume (FEV), forced vital capacity / forced expiratory volume (FEV / FVC), the average amount of strong expiratory airflow (FEF2575) as four variables related to respiratory capacity (FEV, FVC, FEV / FVC, FEF) was considered. One type of blood cancer is myeloid or bone marrow cancer. This type of leukemia affects bone marrow cells. Complete blood count (blood test) is one of the main tests to diagnose many diseases (cancer, anemia or leukemia). Blood parameters including red blood cells (RBC), white blood cells (WBC), platelet (PLT) and mean corpuscular volume (MCV) as four variables related to blood parameters in this study in was considered. Data related to research variables including respiratory capacity (FEV, FVC, FEV / FVC, FEF), blood parameters (WBC, RBC, PLT, MCV), as well as age and work experience of the exposed group and blank group the results of medical examination records Personnel were extracted.The niosh 1501 method was used to evaluate and determine the respiratory exposure of employees to benzene. Ambient air sample was taken from the respiratory area of each group by individual sampling pump and using absorbent activated carbon. GC device and mass detector were used to analyze the samples. Results of respiratory exposure to benzene (OEL-TWA) in the exposed group and the blank group were different and in the personnel of loading units 7 ppm, production 0.523 ppm, laboratory 0.178 ppm, quality control 0.224 ppm and administrative control group were zero (occupational expouser limit-time weighted average : 0.5 ppm). The results showed that the respiratory exposure to benzene was not the same in the exposed and blank groups and in the personnel of loading and production unit more than the allowable limit and in the personnel of laboratory units quality control is less than the allowable limit and in the personnel of the blank group is zero. After determining the respiratory exposure to benzene, the effects of benzene on respiratory capacity and blood parameters as well as the effect of individual factors of age and work experience on these parameters by neural network perceptron (MLP) and adaptive network fussy inference (ANFIS) in matlab 2019 software with a 5-core processor and 8 gb of ram were modeled. To do this, after extracting the data from the personnel medical file, first the data (variables) enter the pre-processing stage and after normalizing and selecting the feature, and dividing the training and experimental data in the fuzzy algorithm, the extracted features are classified. Finally, after evaluating the performance based on the rate of correct and incorrect diagnosis, the relationship between respiratory capacity and blood parameters, age and work experience were modeled. One of the most important steps in data processing and analysis to lead to the best or weakest result is the preprocessing of the data of that research (noise and discarded data removal, data sorting, data labeling, data normalization in the interval 0 up to +1 and finally the data is divided into 70% of data for training, 30% of data for testing of neural network perceptron and adaptive network fussy inference). In fact, pre-processing determines the results and its importance is so great that it can lead to the best result or the worst result. In the neural network perceptron, the best training function with the highest correlation coefficient and efficiency, Levenberg-Marquardt function was selected in matlab library as trainlm with correlation coefficient of 0.9356. Then the best model for network architecture equal to 10 layers and 20 neurons with a performance of 0.0015 and a correlation coefficient of 0.9860 was selected. In neural network perceptron with 100% accuracy, precision and recall in training and 99.7702% accuracy, 99.9321% precision and 99.997% recall in experimental section and error convergence tolerance 0/5*10-3, results it showed that respiratory capacity and blood parameters were not the same in the exposed and blank groups and also the age and work experience factors did not affect the blood parameters and respiratory capacity in the exposed and blank groups. After determining the best neural network training function (Levenberg-Marquardt function in matlab library called trainlm with a correlation coefficient of 0.9356), we placed it in the adaptive network fussy inference (ANFIS) to present the results. The settings related to the adaptive network fussy inference model using the neuro-fuzzy designer toolbox in the apps section in matlab software. Mamdani method was used for the principles of fuzzy modeling inference. In adaptive network fussy inference model after 100 repetitions and error convergence tolerance of 0.54*10-3 with 100% accuracy, precision and recall in training and accuracy 99.8163%, 99.9526% precision and 99.9863% recall In the experimental section, the results showed that respiratory capacity and blood parameters were not the same in the exposed and blank groups and also the age and work experience factor did not affect the blood parameters and respiratory capacity in the exposed and control groups. The use of integrated ANN, GA and PSO algorithms is recommended to compare the results with the obtained results and the accuracy of the system performance.