Artificial intelligence based knuckle-boom crane safety assistant

Today, lift-loading machines form the backbone of the production and logistics network. The immediate surroundings of the lifting and loading machines are the places that are considered as dangerous areas and where the risks of work accidents are high. In the studies conducted in the operatorship courses, it has been determined that 15% of the construction equipment operators are at least high school graduates, while 75% of them are primary school graduates. In Turkey, an average of 1153 fatal occupational accidents occur annually. 26% of these accidents occur in the construction industry. The correct use of the capacity of lifting loading machines is important for reducing accident rates. In order to prevent accidents, electronic control systems are used on the machines that limit operator usage and allow only safe movements. These safety systems, called moment control, ensure that the load is safely lifted and transported away. In this study, unlike the solutions in traditional torque control systems, an artificial intelligence based load limiting system is proposed. Inclination, pressure and length sensors are placed on a 35 ton mobile hydraulic crane with a single folding boom on the vehicle for operation. Then, instantaneous values were recorded during the lifting, lowering, extension and retraction of 6 different test loads with the crane. In total, 30779 data were collected in 50ms periods, over 2 months of work. The field data transferred to the digital environment were grouped as two separate datasets on lifting and extending test weights. In the datasets, repetitive data due to recording frequency and shaking data from the test operator were filtered out. In both data sets, 4 application models with different input parameters were created for pressure and load estimation. Multiple Linear Regression, K-Nearest Neighbor, Artificial Neural Networks, Recurrent State Network structures were trained and tested with normalized data sets. Using the data set of the mobile crane, instantaneous load and pressure estimation was carried out with artificial intelligence models. By teaching previous study data with Recurrent State Network, future study data were predicted. The obtained results have been added to the torque control algorithm and the machine safety system has been made more stable. For the application of the results obtained in the digital environment, an embedded circuit design has been made in which the models trained in the computer environment can be tested. Artificial Neural Network Model was run with 4 different approaches in the designed circuit and successful results were obtained. The prediction was made with an accuracy of 98.3% from the ANN model, which was run with 30 iterations with a single middleware.