The Design And Implementation Of Thermal Analysis System For Nodular Cast Iron

The spheroidization of nodular cast iron can be a decisive factor to it’s mechanical properties, like hardness, tensile strength and toughness. However, the spheroidization is very susceptible to the raw material and the spheroidizing treatment. The traditional prediction algorithm of nodularity and component is often based on multiple regression technique, which is luck of adaptability. A more accurate and adaptable thermal analysis system used to predict the quality of nodular cast iron was designed in this paper, which can predict the content of carbon, silicon, carbon equivalent and the spheroidization rate.(1) Research of the thermal analysis system for nodular cast iron at home and abroad was studied systematically, there are some problems in the current research achievements, then the research content and objective were determined. General scheme of the system in this paper was presented combined with the working principle of the thermal analysis system, which is consisted of the software and the hardware system.(2) The design works of the prediction algorithm based on BP neural network contain the selection of each input neurons, the design of network architecture, and determination of the parameters. Then using the MATLAB Neural Network Toolbox to simulate and train the network. Finally, the algorithm can be applied on the embedded systems.(3) The hardware system mainly consists of sampling device, thermocouple circuit, controller, interactive interface, data storage and the debugging section. The main control unit consists of the microcontrollers of Freescale’s K60 series, which is responsible for the implemention of data processing, forecasting algorithm and some other accessibility; AD7193 was used in the temperature acquisition unit to collect the signal from the K-type thermocouple precisely.(4) The software system was designed based on MQX, the design work is focused on task of the application layer, including the functions of collecting, analyzing, displaying, and saving data. The most important function is predicting the quality parameters using BP neural network algorithm.(5) The entire system was tested and verified. Each module was working properly, the data shows that the system has high accuracy in predicting the nodularity and the component parameters, some point needed to be studied more were given.