Embedded Systems and FPGAs for Implementation of Control Oriented Models, Applied to Combustion Engines

University dissertation from Lund University

Abstract: Performance demands put on combustion engines are ever increasing, e.g. demands on emissions and fuel consumption. The increased demands together with new combustion concepts increase the need for feedback engine and combustion control. Mathematical models are considered important in order to implement high performance feedback control, as well as to perform diagnostic functions in vehicles. Various implementation platforms which can be used to implement mathematical models in vehicles are described in this this thesis; embedded processors, FPGAs and ASICs. Which of these implementation platforms to choose must be decided based on the intended application and current demands on performance. Embedded systems, ASICs and FPGAs are discussed based on literature found in the field, covering a wide span of considerations. Furthermore a number of considerations which are important when implementing algorithms and logic in embedded processors and FPGAs are described.The theory is put into practice in the thesis, implementing a heat release calculation on an FPGA and developing and implementing a NOx model in an embedded processor. To be able to implement a fast NOx model several techniques were used. Parts of the model were tabulated, difficult operators such as division were avoided and the properties of fast C code was kept in mind. This thesis combines the areas of automatic control, electronic hardware design and development of embedded software, and applies it to combustion engine control. The work undertaken indicate different possibilities when implementing high speed control oriented models in FPGAs and embedded processors. This thesis aims to fill a gap between state space models, common in automatic control, and high fidelity physical models, commonly used for simulation, by providing a method to develop high fidelity control oriented models which are low in computation demand and implementable in FPGAs and embedded processors.