Aircraft and method for inspecting equipment in coking plants to detect sources of error

A method for inspecting equipment in coking plants in order to detect sources of defects using an aircraft, wherein the aircraft includes a heat-resistant housing and acquires data by an inspection apparatus and transmits them by a transmitting and receiving device to at least one external analysis device for detecting the sources of defects, wherein the inspection apparatus or components thereof are configured to be retracted into an interior of the aircraft through an automatically closable opening, including a hatch/flap, in the housing, wherein the inspection apparatus includes a gas-concentration measuring device as a chemically reactive device which measures gas concentrations in an atmospheric airspace outside the coke oven in the immediate vicinity of an object of the coke oven, an infrared temperature-measuring device which determines a temperature field on a surface to be inspected in a contactless manner, and a phototechnical device which visually detects damage on the surfaces to be inspected, wherein the method comprises:

• converting acquired phototechnical and optical 2 D data recorded by the phototechnical device into a georeferenced 3D thermal point cloud using a conversion unit of the analysis device and photogrammetry, in order to generate at least orthophotos or surface textures or 3D models;
• processing acquired data from the gas-concentration measuring device, the infrared temperature measuring device, and the phototechnical device in a comparison unit of the analysis device in order to determine deviation data;
• and
• combining all the converted data and all the deviation data as well as further operating data in an evaluation unit of the analysis device in order to detect sources of defects by image processing algorithms and/or machine learning algorithms;
• wherein the orthophotos are pre-processed in such a way that at least noise is suppressed and contrast and brightness are adjusted so that, on the one hand, normalization for comparability of different images and, on the other hand, highlighting of characteristic and relevant components/objects/structures is enabled;
• wherein for the identification of relevant components/objects/structures at least the Hough transformation or the template-matching method or deep-learning models based on convolutional neural networks are used;
• wherein from the data relating to the identified relevant components/objects/structures various features are extracted for the definition of the components/objects, which serve at least for regular comparison;
• wherein data relating to the identified relevant components/objects/structures are segmented from the remaining data-set area;
• wherein at least the data from the temperature-measuring device, are superimposed with the data relating to the identified relevant components/objects so that a temperature distribution becomes visible and thermal deviation data are determined;
• wherein data from chemically reactive and physically reactive devices are superimposed with data relating to the identified relevant components/objects/structures so that chemical deviation data and physical deviation data are determined;
• wherein at least the data relating to the identified relevant components/objects/structures as well as the data from chemically reactive, physically reactive, and temperature-measuring device and the deviation data as well as the operating data which were recorded and evaluated are stored as historical data in a memory unit of the analysis device;
• wherein with the aid of the historical data a correlation model is trained with which sources of defects are predicted.