Development and Usage of a Virtual Mass Air Flow Sensor

Electronic technologies continue to provide ever-increasing options in computational capabilities. This in turn, enables the use of advanced signal processing techniques that can allow sensors or even actuators to be less complex or possibly eliminated. Alternatively, additional information could also possibly be extracted from existing sensors.

This paper will discuss one such example: a virtual mass airflow sensor. The engine airflow system is a multi-variable nonlinear system and as such, use of a virtual mass airflow sensor represents many automotive systems. In this paper we'll discuss the impact of using a virtual mass airflow sensor on the rest of the powertrain control system. We'll discuss the impact in terms of vehicle performance, the impact on the downstream algorithms that use the virtual signal and the impact to the electronic hardware. We'll also discuss how the resulting increases in the cost of the electronics to accommodate the increasing computations are often less than the cost of the sensor and associated wiring.

The specifics of designing a neural net system as a representative of a virtual mass airflow sensor for this project are discussed in a companion paper entitled “A Neural Network Based Methodology For Virtual Sensor Development”.