This presentation summarizes the results of experimental work performed over the past two years under funding by the Washington Tree Fruit Research Commission (WTFRC). Objectives of this effort have been: 1) identification of reliable ammonia sensors suitable for detection of refrigerant leakage in controlled atmosphere (CA) rooms, 2) quantification of ammonia loss due to water uptake in room-sampling lines, 3) development of a new demand-defrost sensor for CA refrigeration systems to improve energy efficiency, and 4) an evaluation of the use of infrared analysis methods (specifically Fourier-transform infrared spectroscopy or FTIR) for continuous monitoring of ethylene, water vapor, and other vapors (especially flavor-related esters and aldehydes) in the CA environment.

Evaluation of Ammonia Sensors

The experimental setup for ammonia sensor evaluations appears in Figure 1. This system is installed at the Stemilt Growers CA storage facilities in Wenatchee, WA and has been in operation for nearly two full storage seasons (November 1998 through July 1999 and November 1999 through May 2000). Airflow for all sensors is drawn from fruit-filled CA rooms to a common sampling manifold. The 40 rooms are sequentially sampled for a 5-minute period at a nominal flow rate of 6 L/min. Following the oxygen and carbon dioxide monitors, CA room vapors enter the ammonia sensor testing section where computer-controlled injections of 10 to 15 ppmV of ammonia are made hourly. The flow is then divided amongst the ammonia sensors under test. Flow rate to each sensor is set per manufacturer's specifications using individual rotameters. Sensors evaluated during the past two storage seasons are summarized in Table 1. Table 2 summarizes the results of the ammonia sensor evaluation study to date.

Figure 1. Ammonia sensor testing system.

Performance of Individual Ammonia Sensors

All sensors were exposed to hourly, 5-minute injections of ammonia at the 10- to 15-ppmV level. Dilution air was supplied from the Stemilt CA rooms. An "optimum" sensor response consisted of 24 signal output spikes (one per hour for 24 hours) with no output or baseline drift between injections. For these evaluations, in addition to sensor outputs, a number of ancillary variables were monitored, including: 1) voltage to the ammonia injection solenoid, 2) carbon dioxide concentration, 3) oxygen concentration, 4) CA room number, and 5) relative humidities at the input and exhaust of the sensor testing system.

David Bishop Model 730
This sensor has performed flawlessly for nearly two years now and serves as a benchmark against which the other ammonia detectors are evaluated. The device requires periodic replacement of a liquid electrolyte and a permeable membrane. Its relatively high cost ($4,400) dictates that it be used with a sampling manifold (rather than dedicating one to each room). Data for this sensor appear in Figure 2. Only minor response to varying oxygen levels has been observed from this sensor.

Figure 2. Response of David Bishop ammonia sensor.

Sensidyne Model 1000
Because of its tendency to fail without notice, this sensor was eliminated from evaluation midway through the 1998-99 storage season. This sensor is not recommended for CA applications.

EIT "Sensor Stik"
Similar to the Sensidyne device, the EIT Sensor Stik is prone to fail without notice and was eliminated from study midway through the 1998-99 storage season. This sensor is not recommended for CA applications.

MST Model FMS 8710 with Model 9602-6700 Sensor Module
Contact: Scott Jones, MST Measurement Systems Inc.
Phone: (360) 254-6759
During the 1998-99 storage season the MST sensor performed flawlessly, with minimal response to varying CA room oxygen concentrations. However, during the current 1999-00 storage year we have observed erratic behavior of two MST sensors. This erratic behavior (see data below) takes the form of extraneous outputs even with no ammonia present. The spurious output spikes correlate with oxygen levels in the rooms as seen in Figure 3. Discussions with the MST technical staff in Chicago [Mark Sherrett, (847) 808-3524] suggest that: 1) at least one of the failed sensor modules we purchased through our local supplier were out-of-date and/or; 2) we have been evaluating the "wrong model" MST sensor (we have been evaluating the Model 9602-6700 sensor whereas MST manufactures a Model 9602-6702-S specifically for use in ammonia refrigeration equipment rooms). Note that the Model 9602-6700 is the one sold by our local ammonia sensor supplier specifically for CA room monitoring applications. By the end of June 2000 we will have two new MST sensors (Model 9602-6700 and Model 9602-6702-S) installed at Stemilt for a side-by-side evaluation through the balance of the storage season (July or August 2000). In the meantime, CA operators considering the use of this sensor should purchase their replacement modules directly from the MST plant in Chicago to ensure that they are "fresh" and have not surpassed their shelf life (Contact: Nelson Rivera, MST, 1-800-547-2900 ext 2502).

Figure 3. Response of MST ammonia sensor.

BW Technologies "Plant Rat"
This electrochemical sensor has performed very well throughout the latter part of 1998-98 and throughout the current 1999-2000 storage season as seen in Figure 4. However, the "Plant Rat" does exhibit a rather high background noise level when ammonia is not present. Consequently, the ammonia detection limit (i.e., the lowest detectable ammonia concentration) for this sensor will be somewhat higher than for the other sensors evaluated. No response to varying oxygen levels has been observed.

Figure 4. Response of BW Technologies "Plant Rat" ammonia sensor.

Draeger Polytron 2
Contact: Greg LeBrun, MPC International
Phone: (425) 788-7564
The limited amount of data acquired with this electrochemical sensor looks promising; however, the sensor does produce extraneous outputs even with no ammonia present (Figure 5). The level of response to varying oxygen concentrations is undetermined.

Figure 5. Response of Draeger ammonia sensor.

NexGen Ammonia Detector
Contact: Matt Landoe, NexGen Manufacturing and Controls
Phone: (509) 624-4213
This sensor is an optical device and, until recently, has performed very well (Figure 6). Recent data indicate that the sensor output is prone to erratic baseline shifts. This behavior is followed by an instantaneous "healing" and the sensor behaves well until the next dropout event. The behavior is apparently attributed to the sensor electronics and not the sensor element itself (since the sensor resumes a normal response after "healing itself"). Nonetheless, this device requires some re-engineering before it can be considered reliable.

Figure 6. Response of NexGen ammonia sensor.

Pacific Technologies (alarm type)
Contact: Dave Putnam, Pacific Technologies Inc.
Phone: (425) 823-5649
This optical ammonia sensor has performed flawlessly for nearly two full storage seasons. Note that this device is configured as an "alarm type" sensor and only shows an output response (5 VDC) when the ammonia concentration exceeds a preset level (in our evaluations, 10 ppmV). Three of these devices have been evaluated during the 1999-2000 storage season to assess manufacturing reproducibility and reliability. Responses of all three devices have been identical (Figure 7). Pacific Technologies is pursuing business partnerships to get this device into production; currently it is available only in limited quantities for beta testing.

Figure 7. Response of Pacific Technologies alarm-type ammonia sensor.

Pacific Technologies (analog type)
In response to our request for a continuous-output device, Pacific Technologies provided this prototype instrument for evaluation midway through the current storage season. Thus far, the sensor has performed well with no extraneous responses. Note that the sensor output signal (Figure 8) is inverted in the current design.

Figure 8. Response of Pacific Technologies analog-type ammonia sensor.

Industrial Scientific
Contact: Richard Miller, District Manager, Industrial Scientific
Phone: (425) 821-4809
This optically based ammonia sensor has performed flawlessly over nearly two full storage seasons. The devices evaluated in our program have been beta prototypes with quantitative ammonia concentration output and are not available for general sale. Industrial Scientific has recently made the decision to delay introduction and sale of this sensor for CA applications due to calibration issues associated with temperature sensitivity. However, the company has recently introduced an alarm-type sensor (based on the same detector) that can be purchased for CA applications. A typical response curve appears in Figure 9.

Figure 9. Response of Industrial Scientific ammonia sensor.

US Industrial
This sensor was installed for testing in late May 2000 and insufficient data are available for an assessment of its performance.

Summary of Ammonia Sensor Evaluations

By the end of this storage season (August 2000), we will have evaluated eight different electrochemical ammonia sensors (David Bishop, Sensidyne, EIT Sensor Stik, MST, BW "Plant Rat", Draeger, US Industrial, Inmet). To date, the performance data indicate:

• The response of all of the electrochemical sensors is similar in that:
  • They typically require maintenance and/or cell replacement sometime during the storage season.
  • They all respond to variations in oxygen concentration (in addition to responding to ammonia).
• Of all electrochemical ammonia sensors tested, the one manufactured by David Bishop has been the most reliable and exhibits the lowest response to varying oxygen concentrations. The relatively high price of this instrument ($4,400) probably precludes installation of one unit for each CA room and suggests that it could be better utilized in sampling manifold configuration.

By the end of this storage season, we will have also evaluated three different optical ammonia sensors (NexGen, Pacific Technologies, Industrial Scientific). To date, the data indicate:

• Both the Industrial Scientific and Pacific Technologies sensors have demonstrated very reliable performance throughout two full storage seasons.
• The NexGen sensor has demonstrated periodic drop-out (no output signal) from which it eventually recovers (without intervention). This appears to be an electronic component failure and not a failure of the ammonia-sensing module.
• None of the optical ammonia sensors (with the exception of the NexGen unit) have required any maintenance or repair.
• None of the optical ammonia sensors have shown any response to varying oxygen concentrations.

Demonstration of Ammonia Uptake in Sampling Lines

In fall 1999 we performed laboratory experiments at Battelle facilities in Richland, WA to demonstrate and quantify uptake of ammonia in CA room sampling lines. The laboratory evaluation system is depicted schematically in Figure 10.

Figure 10. Laboratory setup for ammonia uptake studies.

To conduct these investigations, 200 feet of 1/4-inch ID Tygon^TM tubing (provided by Nate Reed, Stemilt Growers) was enclosed in an environmental chamber where temperature could be lowered below the dewpoint to induce condensate formation in the tubing. Moist air (70% RH) and dilute ammonia vapor (55 ppmV) were introduced at the tubing input and the chamber temperature was lowered until condensate formation began. A plot of the experimental data appears in Figure 11.

Figure 11. Experimental data indicating ammonia uptake by water in sampling lines.

Ammonia concentration at the tubing exhaust was observed to drop to 15 ppmV (from the original 55 ppmV) due to the presence of liquid water in the tubing. Note that most commercial ammonia sensors have a low detection limit of a few ppmV of ammonia. Consequently, if our injected ammonia concentration had been lower (say 10 to 15 ppm) the ammonia level at the tubing exhaust could have gone undetected by any existing commercial ammonia sensor. Therefore, the data suggest:

• Even a relatively large ammonia concentration in a CA room could be "scrubbed" below the sensor detection threshold due to condensed water in the sampling lines.
• If sampling lines are required to be long (greater than 100 feet) it is advisable to consider installation of a dedicated ammonia sensor in each room or cluster of rooms.

Infrared Vapor Analysis in CA Rooms

In the original sensor suited installed at Stemilt Growers for ammonia sensor evaluation, a MIDAC FTIR system was utilized in the exhaust line to verify ammonia concentration. During the current storage year (November 1999 through August 2000) this infrared analyzer has been utilized to monitor ethylene and other vapors coming from the CA rooms. Representative ethylene concentration data for the Stemilt CA rooms appear in Figure 12 and 13.

Figure 12. Ethylene concentration for single CA room for two-week period in 2000.

The Midac FTIR vapor analysis system can also be used to monitor "flavor vapors" in apples and other fruit. The feasibility of monitoring butyl acetate was demonstrated in a laboratory experiment performed at PNNL facilities in Richland. Resulting data are displayed in Figure 14.

Figure 14. Detection of butyl acetate in Red Delicious apples.

Finally, the feasibility of monitoring 1-methylcyclopropene (MCP), a ripening inhibitor, was demonstrated using the Midac instrument at Stemilt facilities in Wenatchee, WA. Representative data appear in Figure 15.

Figure 15. Infrared absorption spectrum for methylcyclopropene (MCP).

The use of FTIR instrumentation for continuous, real-time analysis of CA room vapors appears entirely feasible. While the instrumentation is fairly expensive ($35,000 to $40,000), the method allows a level of continuous room monitoring which heretofore was not possible (e.g., the measurement of ethylene in CA rooms is typically performed by bagging gas samples and analyzing them in an offsite gas chromatography laboratory). In addition, a single FTIR system can simultaneously monitor a wide variety of CA room vapors including ethylene, ammonia, and butyl acetate (from Red Delicious), as well as other "flavor" gases that affect fruit quality.