Search Results

An accurate estimation of cycle-by-cycle in-cylinder mass and the composition of the cylinder charge is required for spark-ignition engine transient control strategies to obtain required torque, Air-Fuel-Ratio (AFR) and meet engine pollution regulations. Mass Air Flow (MAF) and Manifold Absolute Pressure (MAP) sensors have been utilized in different control strategies to achieve these targets; however, these sensors have response delay in transients. As an alternative to air flow metering, in-cylinder pressure sensors can be utilized to directly measure cylinder pressure, based on which, the amount of air charge can be estimated without the requirement to model the dynamics of the manifold.

A number of variables in a conventional automotive powertrain are scheduled on-line based on the current operating conditions with the goal to achieve the best fuel economy (FE), emissions, and performance. The functions are obtained off-line, i.e. after a mapping, data regression, and optimization process followed by in-vehicle calibration for fine-tuning the powertrain behavior. More complex engines, referred to as high degree of freedom (HDOF) engines, require a careful tradeoff between the mapping, optimization, and calibration time on one hand and the achieved accuracy on the other. Additionally, the powertrain control module (PCM) has limited computational resources. Thus, fully representing the more complex functions can be prohibitive. As a result, an HDOF powertrain in actual operation may not completely achieve the potential benefits the new technologies offer.

Current methods of storing critical engine operating parameters often rely on lookup interpolation functions. A drawback to lookup interpolations is that as the number of independent variables increases, the function's dimensions increase from tables to cubes to hypercubes of data. A sparser collection of data may serve if the strict orthogonal structure of lookup functions is not required. Instead of forming a matrix of data for every combination of input variables, a few critical data points are stored with their independent variables. The available data can be used to estimate parameters at new input values (typically from current measurements) by relating the available data by its distance from the new inputs. An inverse distance weighting scheme is used to calculate the output of the function.