(213a) The Transport Phenomena within the Intasense Indoor Air Quality Monitor Design

As the energy efficiency of buildings increases, there has been rising concern about the reduction of indoor air quality. This is because ventilation systems frequently use open loop control: the amount of fresh to recycled air entering the building is frequently held constant irrespective of the current air quality of the building. Using open loop control in a ventilation system can cause too much fresh air to enter the building, leading to energy inefficiencies, or too little fresh air to enter the building, leading to poor air quality. Poor air quality in turn has been shown to cause or trigger a variety of illnesses including asthma and sick building syndrome.  Within the European Union the productivity loss equivalent of sick leave as a result of poor air quality has been estimated at more than 80 billion euros annually [1]. Additionally, space heating accounts for more than half of the energy consumption of public and residential buildings within the European Union [2]. Therefore, to improve the energy efficiency of the building, it would be better to use closed loop control in ventilation systems, employing negative feedback from a network of air quality monitors, to determine the amount of fresh air which should be added to the environment. This would have the potential to reduce both sick leave and medical costs while simultaneously decreasing the energy consumption of a building.

There are a few devices on the market meant for permanent installation which measure indoor air quality. However, they are frequently limited to combinations of total volatile organic compounds (VOCs), carbon dioxide and humidity concentrations. As VOCs such as benzene, formaldehyde are carcinogenic and their OSHA exposure limits vary by more than an order of magnitude, it is clear that measuring only total VOC concentration is not sufficient. Therefore, the INTASENSE Air Quality Monitor, which houses three metal oxide gas sensors, was designed [3]. These sensors target formaldehyde, benzene, nitrogen dioxide and carbon monoxide by chemical reactions between the target gases and oxygen species at the semiconductor surface, which in turn causes a change in resistance across the surface [4].

Metal oxide sensors, particularly those in prototype stages, need to be tested and used in controlled environments. Specifically, it is very important to control the transport phenomena within the device because water can change adsorbed species at the metal oxide surface. Additionally, the temperature of the surface of the sensor can also change its reactivity. Therefore, change in flow rate, leading to a change in temperature, or humidity fluctuations can cause a fluctuation in the sensor response. Therefore, to incorporate three prototype metal oxide sensors into a single device three transport problems needed to be solved.

First, metal oxide sensors have a surface operating temperature between 150 and 400 degrees Celsius. In commercial sensors heat transport from the sensor packaging to the fluidic system is not a problem because the sensor surface is well insulated from the sensor packaging. However, part of sensor prototype development is perfecting this insulation. Therefore, the prototype device needed to be able to withstand high temperatures, while minimizing heat transport to components with melting points around 50 degrees Celsius.

To accomplish this, a modular, plug and play design was chosen so that components in contact with the sensor packaging could be exchanged if necessary. Computational models of heat transport from the sensor, through the sensor packaging and to the fluidic device were conducted to verify designs prior to construction. Initially, all components were machined from PMMA. However, during sensor prototype experiments, where the sensor packaging reached temperatures in excess of 60 degrees Celsius, PMMA components in direct contact with the sensor packaging were replaced with PEEK.

The second transport problem involved insuring that each of the three sensors received an equal flow, with the same delay from the inlet of the device to the sensor surface, with gas flow centered on the sensor surface, all in parallel. To do this the device was broken up into three subsections: the channels before the sensor, the channels after the sensor and the compartment which housed the sensor. So that the delay between air entering the system and arriving at each sensor would be equal, the channels before the sensors were made as identical as possible and the pressure drop was held equal by changing the width of the channels after the sensors. Because the pressure drops were equal, the flow through each channel was also equal. Lastly, the sensor compartment architecture was designed so that it would direct the flow onto the sensor surface.

To determine if the gas flow through each channel was equal, nine experiments were conducted on the device. Specifically, a syringe pump was used to repeatably generate gas flow while three pressure channels and three differential pressure sensors were connected to the inlet and outlet of each sensor inlet and outlet. Each item was used once with each other item. The differential pressure was recorded. Least squares methods and an ANOVA test was then used to determine if any variance in sensor reading could be attributed to fluidic system. The results of the experiments show that the flow rate through each channel is equal within statistical bounds.

The last transport problem involves modeling the adsorption, desorption, convection and diffusion of analyte molecules through the preconditioning unit. A preconditioning unit, which stabilizes the humidity of the air passing through the device while minimally adsorbing analyte molecules, was necessary to incorporate in to the device because metal oxide sensors are also sensitive to humidity fluctuations. Therefore, a layer of silica gel, which acts as a reversible adsorbent, sandwiched between particle filters was placed before the gas sensors, to dampen humidity fluctuations. However, as in liquid chromatography, the analyte peak concentration will widen as the length of the column increases. Additionally, there was concern that molecules of interest could also be reversibly adsorbed, leading to a decrease or loss of the sensor signal.

In order to minimize signal loss, while achieving sufficient humidity fluctuation control, experiments were conducted and a computational model was generated to determine the correct amount of adsorbent to use based on the expected humidity fluctuations, and desired sensor flow rates. These experiments took two forms: pulse studies and isotherm measurements. In the pulse studies, pulses of contaminant gas and humidity were run through the system to see if sensors were responding to fluctuations in humidity or gas concentrations. Using 2g of silica gel and a gas flow rate of 0.4 l/min and a starting relative humidity of 35 percent, pulses of nitrogen dioxide at 12ppm, 16ppm, 20ppm; formaldehyde at 2ppm, 8ppm, 14ppm, 20ppm; CO at 20ppm, 40ppm, 60ppm, 80ppm, 100ppm; and benzene at 2ppm, 4ppm, 6ppm 8ppm and 10ppm were all detectable with a delay on the order of seconds. In addition to these pulse studies the isotherm curve, which correlates the amount of a chemical on an adsorbent to the amount in the air at equilibrium, was measured for water. This was used in combination with literature values for contaminant gases to generate a computational model of the transport of gases through the preconditioning unit.

Bibliography: