Humidity control plays significant role in various industries to get desired product output. An experiment was conducted in the laboratory scale to model a humidity process and to control the same. Open loop test was conducted for the humidity process by introducing the step change and the relative humidity of the exit air was measured. The data was best matched with first order process with time delay. Proportional Integral (PI) controllers based on Ziegler Nichols (ZN), Direct Synthesis (DS) and Internal Model Control (IMC) methods were simulated in MATLAB environment and the time domain specifications showed better performance for IMC and DS based PI controllers.
PDF Abstract XML References Citation
How to cite this article
Humidity is the term related with water vapor present in the gaseous mixture. The total amount of water vapor contained in the air is expressed as absolute humidity. The Relative Humidity (RH) is the term related to ratio of partial water vapor pressure to saturation water vapor pressure at the given temperature. Thus, RH is temperature dependent. The measurement is expressed as a percentage. The temperature and relative humidity both are dependent to each other. The relative humidity decreases with increase in temperature. Humidity is the most important parameter to be maintained in the industries such as pharmaceuticals, textiles, food processing and tobacco to achieve the desired product quality. Venkatesh and Sundaram (2012) have reviewed the effect of humidity on product quality and human comfort in various industrial processes as well as air conditioning systems and also different control schemes for maintaining the same. Young et al. (2007) concluded that formulation of dry powder inhaler and aerosolisation efficiency are affected if proper humidity level is mot maintained. Better performance in terms of time and quality in the preparation of photonic crystal films are achieved if desired RH level is maintained (Liau and Huang, 2008). Rao (1989) suggested that the lesser heat is recovered in power generation if humid air turbine is utilized.
Hence, it is essential for the industries as well as human comfort to control and maintain the relative humidity to the desired level. A good controller is needed to maintain the relative humidity to the desired value. The designed controller should be optimum in all aspects of performance. An independent control of humidity in Heating Ventilating and Air Conditioning system (HVAC) is proposed by Liu et al. (2006), which utilizes the combination of refrigerator and liquid desiccant system. A decoupled independent control of relative humidity and temperature is proposed by Gomez and Reyes (2001), which uses multivariable cascade control. The inner loop uses decoupling and the outer one uses PD controller. Han and Zhang (2011) reported the independent control of humidity and temperature for air conditioner. In this method, the sensible and latent heat loads are removed separately to maintain relative humidity and temperature so that indoor comfort and consumption of energy are improved. Soundaravalli et al. (2007) have proposed model based on control for the humidifying process with transportation lag. They designed different control schemes such as ZN, IMC, Smith predictor and IMC based PID for humidity process and tested for servo and regulator problems.
Model based controller for different processes have been addressed in the literature. Madhavasarma and Sundaram (2007, 2008) have identified the model for linear and non linear processes and different controllers were also designed. Vijaya Selvi et al. (2006) designed and implemented internal model controller based on identified model for a conductivity process with transportation lag. Ziegler and Nichols (1993) proposed a conventional tuning technique for PI controller for stable first order process with dead time. Hagglund and Astrom (2008) suggested tuning rules for PI controller from step response. Chen and Seborg (2002) proposed direct substitution of formula for direct synthesis based PI/PID controller. Fruehauf et al. (1994) discussed IMC based PID tuning rules. Erdal et al. (2001) have developed a module for PID controller using current feedback amplifier with the help of active components.
Genetic algorithm based PID controller has been developed and implemented for a nonlinear plant (Ajlouni and Al-Hamouz, 2004) which resulted better integral squared error for set point changes for a particular operating point. Bhaba et al. (2007) studied Wiener model based PI controller for a conical tank process and compared with Ziegler and Nichols tuning strategy in terms of integral squared error. Ajlouni and Al-Hamouz (2004) proposed a neural based PID controller as well as feed forward controllers using evolutionary approach in genetic algorithm. Outperforming performance in time domain was achieved in this approach even in the presence of uncertainties in the plant parameter. Internal model controller was realized for induction motor to control the speed which uses artificial neural network to eliminate the sensor (Mouna and Lassaad, 2007).
After reviewing the existing methods, this study addresses the problem of identifying the black box model, for the humidifying pilot plant was performed and for the identified model, different controllers were designed and implemented.
HUMIDITY PROCESS AND SYSTEM IDENTIFICATION
A Humidity Chamber (HC) is present in the experimental setup which is shown in Fig. 1. The valve V1 is manipulated to bubble the primary air into the HC. The primary air flow rate is measured using the rotameter R. The humidified air at the outlet is made to pass through the coil of 1.25 cm long and 3 m diameter. HIH-3610 Honeywell made sensor is used to measure the relative humidity of the exit air. The air flow rate was adjusted in steps of 0.05 LPM from 0.35-0.6 LPM to achieve different operating regions. The step response was recorded for all the six operating regions. The next step input was given by manipulating the valve V1 after the steady state was reached. Process reaction curve as shown in Fig. 2 was obtained using the recorded data for six different operating regions. The process model was identified as suggested by Bequette (2003) from the process reaction curve. The identified model was validated against the experimental data with calculated data. The model was best fitted to a first order process with time delay with an accuracy of 5%. Table 1 shows the model parameters for six different operating regions.
DESIGN OF CONTROLLERS
Maintaining the relative humidity to a desired value for human comfort and product, quality is achieved by different control strategies. Lot of control techniques is available in literature for control of humidity and/or temperature. Proportional only controller leads to offset because the controller output and process output attain new equilibrium value. Hence the integral of error is taken into account and the past history of the process is considered (Nithya et al., 2008). So the controller output keeps on changing till the error becomes zero. In this study, several Proportional Integral control (PI) schemes are analyzed and simulated using MATLAB environment. The PI controller output is expressed as:
|Table 1:||Model parameters identified for different operating regions|
|LPM: Liters per minute|
|Fig. 1:||Flow diagram of the experimental setup. HC: Humidity chamber, R: Rotameter, V1, V2: Manual valves and NRV: Non return valve|
|Fig. 2(a-b):||Process reaction curve for system identification for different operating regions (a) 1, 3 and 5 and (b) 2, 4 and 6|
The transfer function for the same is given as:
PI controller tuning using ZN technique (ZNPI): Tuning of PI controller was suggested by Ziegler and Nichols which was a closed loop technique. The step response of the system was analyzed. After making the system response to sustained oscillation, the ultimate period and ultimate gain were determined. Direct substitution of the formula was also proposed by Ziegler and Nichols. From the frequency response plot also the ultimate gain and ultimate period may be found. The ZN tuning rules are suggested by Coughnwar and Leblanc (1991). The tuning values are determined by substituting the ultimate gain and ultimate period. The proportional term is obtained as Kc = 0.45 Ku and the integral term as τI = Pu /1.2. Coefficient Diagram Method (CDM-PI) for a nonlinear pH neutralization system was proposed by Meenakshipriya et al. (2012), in which the set point tracking is compared with Ziegler Nichlos which described that CDM-PI gives better performance when compared with ZN-PI.
Direct synthesis technique of PI controller design (DSPI): The controller was designed based on the closed loop transfer function and the process model. In this approach, a significant relationship is existing between the process model and the controller (Foley et al., 2005). Gain of the controller is closely associated with system gain K. Both are having inverse proportion so that the stability of the system is achieved. In this technique, controller gain multiplied by model gain remains constant. The PI controller settings are as follows (Seborg et al., 2004):
The key decision in this approach is the choice of tuning parameter. Rivera et al. (1986) has proposed choice for this design parameter.
IMCPI controller design: IMC based controller is best suited for first order process with time delay. But the IMCPI controller may produce better performance if the dead time is approximated in the controller design. The controller settings are as follows (Bequette, 2003):
|Fig. 3(a-b):||Response of the controllers for set point and load changes for different operating regions (a) 1, 3 and 5 and (b) 2, 4 and 6|
|Table 2:||Tuning parameters Kc and τI obtained for different controllers|
|Kc: Proportional term, τI: Integral term, ZNPI: Ziegler nichols proportional integral controller, IMCPI: Internal model control proportional integral controller, DSPI: Direct synthesis proportional integral controller|
The filter coefficient is playing significant role in designing the IMCPI controller.
RESULTS AND DISCUSSION
The model was identified for laboratory scale humidity. For the identified model, tuning technique such as ZNPI, DSPI and IMCPI were simulated using MATLAB. The tuning parameters are shown in Table 2. Unit step input was given to the process. The process was tested for both servo and regulatory problems. The initial condition of the process was taken as zero and unit step input was given to analyze the set point change. After the steady state was reached, a disturbance input was given to the process. Figure 3 show both set point and load changes of the process for all the six regions.
|Table 3:||Time domain specifications for different controllers|
Table 3 shows the time domain specifications for all the three different controllers.
Since, humidity plays an important role in maintaining the quality of the product in industry as well as human comfort, an attempt has been made to maintain and control the same in the lab scale environment. The process model was obtained for the pilot humidity plant which was approximated with FOPDT model. The model also validated with calculated data. For the model obtained different PI control schemes were designed and simulated using MATLAB environment. The time domain specifications such as rise time, peak time, peak overshoot and settling time were found for both servo and regulator problems. From Table 3 it is evident that DSPI and IMCPI controllers are equally outperforming ZNPI in terms of faster settling time and lesser overshoot.
- Gomez, C.R. and M.V. Reyes, 2001. Decoupled control of temperature and relative humidity using a variable-air-volume HVAC system and non-interacting control. Proceedings of the IEEE Conference on Control Applications, September 5-7, 2001, Mexico, pp: 1147-1151.
- Ziegler, J.G. and N.B. Nichols, 1993. Optimum settings for automatic controllers. J. Dyn. Syst. Meas. Control, 115: 220-222.