II. Papel de KlGpd1 en la regulación de la vía KlSNF1-Cat8/Sip4 y otras rutas de
9. Producción de etanol en el mutante Klgpd1Δ
Leak detection tests were repeated multiple times (minimum 5 repeats) in order to ensure repeatability and reliability of the results. Data from all of the sensors attached to the pipe were successfully measured during the tests. As mentioned previously in Chapter 4, the proposed leak detection method is intended for relative pressure measurement; therefore, measurements taken from the FSR based sensors are normalised based on their maximum stabilised pressure and baseline (zero pressure) values. The relative nature of the readings removes the need for precise calibration of the sensor, which in turn will result in easier and lower cost installation. However, a rough calibration at the time of installation by means of adjusting the clips’ initial tension is required to ensure that the outputs of the sensors are in the range of the analog to digital converter (normally 0-5V). Figure 5.2 illustrates an example of the normalised output of the sensors during a leak test (more examples are presented in
Figure 5.2 Example of the normalised relative pressure output from the five sensors during a leak test
Four main stages of the experiment (pump on, stabilisation, leak/burst and pump off) are clearly visible from the output of the sensors. It can be clearly seen in Figure 5.2 that the pressure increased in the pipes from the baseline as the pump was started. The burst/leak event is also clearly visible in the data, as the relative pressure suddenly dropped as the rubber insert was forced out of the hole due to pressure. Finally, the drop in pressure due to switching off the pump is shown in the output of the sensors. Figure 5.2 clearly shows that the simulated burst/leak can be detected from the output of the sensor. The location of the simulated leak is not clearly determinable from Figure 5.2 as all sensors seem to respond similarly to the leak. However, further analysis of the data at the time of the leak showed that sensors 1, 2 and 3, which are upstream of the leak, have a different pressure profile compared to sensors 4 and 5 during the leak incident. Figure 5.3 shows a close up of the output of the
Figure 5.3 The close up of the normalised pressure at the time of leak taken from Figure 5.2. As can be seen from Figure 5.3 sensors 4 and 5 show a more gradual drop in the relative pressure compared to sensors 1, 2 and 3; which show a rapid drop in relative pressure as the rubber insert is forced out of the hole. This difference can be used to locate the zone in which leak had happened (i.e. between sensors 3 and 4). The first derivative of the relative pressure output of the sensors can be used to compare rate of change in the output of the sensors. Figure 5.4 illustrates the average of minimum value of the first derivative of the normalised relative pressure for the sensors 1-5 based on five repetitions (Appendix B). Error bars in Figure 5.4 are based on the standard deviation of the results from these repetitions.
I0.6! I0.4! I0.2! 0! 0.2! 0.4! 0.6! 0.8! 1! 1.2! 174! 174.5! 175! 175.5! 176! N or ma lis ed !r el at iv e! pr es su re ! Time!(seconds)! Sensor1! Sensor2! Sensor3! Sensor4! Sensor5!
Figure 5.4 Maximum rate of change in normalised pressure
As can be seen from Figure 5.4 the rate of pressure drop is higher in sensors upstream of the leak (sensors 1, 2 and 3) compared with sensors downstream of the leak (sensors 4 and 5). Moreover it can be seen from Figure 5.4 that the rate of change in sensor 3 (closest sensor to the leak) is higher than the rate of change in the other sensors. This is to be expected as sensor 3 was placed very close to the burst and is affected by the localised pressure drop in addition to the systematic pressure drop caused by the burst. However from Figure 5.4 there appears to be no trend between the amplitude of the pressure drop and distance for the sensors downstream of the leak. However time analysis of the data during the leak event showed that the order in which the sensors respond to the burst depends on their distance from the leak. Figure 5.5 shows the normalised relative pressure output of the sensors at the time of the leak and the delay of each sensor responding to the leak based on their location.
Figure 5.5 Delay of the output of the sensors depending of their location
As can be seen from Figure 5.5 as expected sensor 3 (closest to the leak) responded to the leak first followed by sensor 2 (≈20ms delay) and sensor 1 (≈30ms delay). Sensors 4 and 5 had a much slower response to the leak at ≈50ms delay and ≈90ms delay respectively. This shows that it is feasible to detect the location of the leak based on response time difference between the sensors where a high sampling rate (≥100Hz) is possible. These tests were repeated multiple times (see Appendix B) to investigate the repeatability of the results. All of the repeat tests showed similar trends to ones presented in Figure 5.4 and Figure 5.5. In addition, Li (2014) carried out these tests on a similar test rig and obtained similar results to ones presented in this thesis. A hybrid sampling rate can be used where a high sampling rate is not feasible. In this method an interrupt is configured to react to the drop in the output of the sensors and activate a higher sampling rate. Based on the reaction time of the interrupt an initial section of the pressure drop profile will be lost. However this partial pressure drop profile can still be used to detect difference between the response of the sensors based on their location.