Powered air purifying respirators (PAPRs) are devices used for protection against contaminated air. They typically contain a filter, centrifugal blower and a motor powered by a battery. Standard gas masks work on the principal of negative pressure, where the user must inhale through the filter in order to clean the air. For some cases the thickness of the required filter can be very large, creating a very large resistance to breathing. This causes discomfort for the user and can impair physical performance. The centrifugal blower in PAPRs is used to provide a positive pressure to the user and minimise the resistance caused by the filter by drawing air through it.

Until recently, PAPRs have been developed to provide a constant flow of air to the user. The fixed airflow is based on average breathing rates and has limitations to the protection it can provide when the breathing demand from the user increases due to increased physical exertion. The constant airflow supply also has consequential detriments to the efficiency of some of the components in the device. Filters have a defined lifespan that is dependent on the concentration of the gas they are filtering and the volume of airflow through it. To maximise the lifespan of the filter it should first be recognised that it is only required for use on the user's inhalation. When the user exhales, all the additional air from their lungs and the, still running, centrifugal blower is dispersed through an exhaust valve in the mask. If the centrifugal blower were to only provide air to the user upon their inhalation, the lifespan of the filter could be increased. The control of these PAPRs have been categorised as breath-responsive and the literature published for the development of these is minimal.

It is recognised in this project that small changes to individual components in a PAPR can create large changes in performance, making the development of a controller difficult. It is also recognised that creating a controller for an individual case would be limited, as developments to optimise components would make it subpar. This project aims to create a form of controller that could account for individual developments in the PAPR components, making the design of the system modular. By doing this, the controller should also be able to alter depending on changes to the system. For example, by changing the: thickness of the filter; shape of the mask; centrifugal blower; motor.

To achieve an adaptive controller, this project uses system identification. This process includes a thorough investigation into the performance of the system in order to define a dynamic model to describe the relationship between an input voltage to the motor and the output pressure of the blower. The system has been analysed in the frequency domain to assess the severity of any nonlinear effects in the bandwidth of interest and models in the time domain have been developed that provide a good fit to describe this dynamic behaviour. The challenge with this is to test the model with a variety of filter thicknesses and centrifugal blowers. If a single model can show consistently stable convergence and its performance is good, then it may be generic enough for a controller aimed towards modular design. It can then be developed for use in an adaptive controller, with the intention of making the controller capable of being identified online. Before this is achieved, though, the derived model can be used offline to develop simple PI controllers and test the effectiveness of the identification procedure.