Eksploatacja reaktorów przemysłowych ze wspomaganiem tomografii ultradźwiękowej i algorytmów głębokiego uczenia

Grzegorz Kłosowski,Antoni Świć,Tomasz Rymarczyk,Konrad Kania,Tomasz Cieplak

doi:10.17531/ein.2020.1.16

Abstract

Monitoring of industrial processes is an important element ensuring the proper maintenance of equipment and high level of processes reliability. The presented research concerns the application of the deep learning method in the field of ultrasound tomography (UST). A novel algorithm that uses simultaneously multiple classification convolutional neural networks (CNNs) to generate monochrome 2D images was developed. In order to meet a compromise between the number of the networks and the number of all possible outcomes of a single network, it was proposed to divide the output image into 4-pixel clusters. Therefore, the number of required CNNs has been reduced fourfold and there are 16 distinct outcomes from single network. The new algorithm was first verified using simulation data and then tested on real data. The accuracy of image reconstruction exceeded 95%. The results obtained by using the new CNN clustered algorithm were compared with five popular machine learning algorithms: shallow Artificial Neural Network, Linear Support Vector Machine, Classification Tree, Medium k-Nearest Neighbor classification and Naive Bayes. Based on this comparison, it was found that the newly developed method of multiple convolutional neural networks (MCNN) generates the highest quality images.

Highlights

The developed method is a new proposal in the field of tomographic algorithms, and it cannot be said that its use will always outweigh the effectiveness of other known methods, in all tested cases the metoda wielu splotowych sieci neuronowych (MCNN) algorithm proved to be the most effective
Known and currently used methods of monitoring tank reactors are still burdened with problems resulting in a relatively low resolution of reconstructed images, it was necessary to take up the analyzed subject
An important achievement of the research is the noise-resistant algorithm based on multiple convolutional neural networks (MCNN), which, despite being trained on simulation data, effectively reconstructs objects hidden inside the tank, regardless of their shape, quantity, location or dimensions, based on real measurements

Summary

Introduction

Due to the time-varying, non-linear and uneven nature of this process, it is very difficult to determine the exact mathematical model of these processes, which necessitates their monitoring. For this reason, to ensure a high level of reliability and trouble-free maintenance of tank reactors it is necessary to effectively monitor the processes taking place inside them. A tank chemical reactor is, in the simplest sense, a vessel adapted to carry out a specific chemical reaction in it. Chemical reactors and the processes taking place in them are usually an essential element of a technological process aimed at producing a specific chemical product.

Methods

Results

Discussion

Conclusion