Elsevier

Applied Thermal Engineering

Volume 152, April 2019, Pages 24-31
Applied Thermal Engineering

Research Paper
Validation of CUSUM control chart for biofouling detection in heat exchangers

https://doi.org/10.1016/j.applthermaleng.2019.02.009Get rights and content

Highlights

Abstract

Biofouling is an undesirable phenomenon in heat exchangers, which adheres slowly and progressively to the inner surface of the tubes, results in a reduction of the heat transfer, and increases the operating and maintenance costs. Controlling this phenomenon is essential to ensure that the equipment operates under optimal conditions.

Different biofouling detection methods are currently available, depending on the properties of the fluid or by means of intrusive elements in the process. In this paper, the validation of the CUSUM control graph as an alternative technique to conventional detection methods is proposed. This type of graph is very efficient in the early detection of slow and progressive changes within a process.

In a pilot plant, formed by two tubular heat exchangers, biofouling was allowed to grow until experimental variables indicated that growth had occurred. The biofouling growth was monitored through the evolution of the heat transfer resistance Rf, which is a technique that has been widely validated for this purpose, and through the CUSUM control graphs.

The variable Rf, clearly indicates the different phases of growth. However, the evolution of this variable depends on the physical-chemical characteristics and biological activity of water. The CUSUM graphs demonstrated a greater predictive capacity with regard to changes in the biological adherence process. Additionally, their results do not depend on the characteristics of water. This new tool could be simply and economically implemented regardless of the heat exchanger’s location.

Introduction

Heat exchangers are the most commonly used equipment in industrial processes. They function by thermally exchanging one fluid with another through their walls [1]. There are different types and sizes of heat exchangers, whose principle and classification is well defined [2].

An inevitable effect on heat exchangers using seawater as the cooling fluid is the presence of biofouling [3], [4]. This undesirable phenomenon, wherein biotic deposits adhere and accumulate on an artificial surface submerged into or in contact with seawater, consists of an organic film composed of micro-organisms embedded in a self-produced polymer matrix called biofilm, wherein inorganic particles (salts and/or corrosion products) resulting from other types of encrustation processes can be captured and retained. Such biofilms can cause the accumulation of macro-organisms [5].

The biological growth in the heat exchangers consists of three clearly different phases: the initial growth or induction phase, exponential accumulation, and stabilisation phase or levelling-off [6], [7]. Throughout its three phases, the development of biofouling inside the exchangers is conventionally calculated by indirect measurements using a method based on monitoring the fluid properties. Additionally, mathematical calculations are performed to obtain two variables; namely, the frictional fluid resistance (f) and the heat transfer resistance (Rf), which indirectly define the deposition of biofouling inside the tubes [8].

Biofouling can seriously impair the ability of the surface to transfer heat under the conditions of temperature difference for which it was designed [9]. In heat exchangers and condensers, biofouling formation increases the frictional resistance of the fluid, reduces the heat transfer efficiency, increases the operational and maintenance costs [10], and leads to financial and energy losses. In industrial processes, 15% of the maintenance costs can be attributed to the maintenance of heat exchangers and boilers. Half of this percentage is caused by biofouling, whose derived cleaning costs can amount up to $50,000 per heat exchanger per cleaning [11]. The costs associated with heat exchanger fouling include losses resulting from a decline in efficiency, reduced production during unplanned stops caused by biofouling, and maintenance costs resulting from the elimination of biofouling. For these reasons, it is extremely important to detect and mitigate this phenomenon before it reaches levels that are sufficiently high to affect the capacity of such equipment by causing it to perform at levels lower than its design levels.

Other techniques have been investigated for the measurement of biofouling inside heat exchangers, such as ultrasound [12], [13], acoustic [14], optical [13], and x-rays [15]. However, these applications require a very expensive and intrusive equipment in the biofouling growth process, and can even alter it. The most recent studies in this field have focused on reducing the sensor costs and determining their optimal location [16] such that their intervention in the biofouling growth process is limited.

Computational techniques have also been used to predict the amount of biofouling prior to the limit factor being reached [17]. These techniques are based on mathematical models that have the ability to learn the different data operating conditions and develop a model capable of predicting the behaviour of the process [18]. Artificial Neural Network (ANN) [19], [20] and Kalman filtering [21], [22] are amongst the techniques that have been used to predict heat exchanger biofouling. A comparative investigation of both techniques can be found in [20]. The disadvantage of these techniques is that they require a long period of learning and have difficulties in predicting the model when small variations exist in the inputs [20], [23].

Statistical process control (SPC) has been used in other fields to predict the evolution of a process over time. Contributions in this field have been made through the Shewhart graphs for univariate processes or the Hotelling T2 control graphs for multivariate processes [24]. In this study, the Shewhart X-S graphs were used to control three measurements of the cylinder section, and were analysed separately with regard to the rectification process monitoring of a steel cylinder’s internal diameter. Additionally, the Hotelling multivariate T2 control chart was used to monitor all three measurements simultaneously. Similar results were obtained from both graphs. However, Shewhart’s control chart was easier for the users to implement and interpret.

A small group of seven variables corresponding to the cylinder lubrication process were also monitored in a 2T marine engine through the Hotelling T2 control chart, under specific working conditions [25]. The deviations in the process were effectively detected with respect to the optimal working conditions. However, small and progressive changes were not detected because this type of graph has difficulties in detecting this type of behaviour.

The Shewhart graphs for univariate control processes behave in a similar manner to the Hotelling T2 graphs. That is, they lose sensitivity with regard to the detection of small changes in the process mean vector [26]. Most of the heat exchange processes are stable and do not suffer alterations to the input variables of the process, which may affect biofouling growth.

In these cases, alternative techniques, such as cumulative sum charts (CUSUM), are the most appropriate. This type of graphs represents the cumulative sum of the deviations, which contain information from all previous samples. Therefore, they are very efficient in detecting small but progressive changes in the variable being measured [27]; In this issue, during the process of making a part for the automotive industry, the Shewhart and Cumulative Sum (CUSUM) control charts were compared for the same magnitude in the change process. Although the CUSUM charts detected the changes satisfactorily, the Shewhart control charts could not detect them, which indicates that the process was in control.

In the naval sector, the CUSUM control graphs have been used in the online monitoring of a group of variables belonging to the fuel processing of a marine diesel engine installed on a tanker. The effectiveness of the technique was demonstrated by the detection of slow and progressive deviations [28]. This technique has also been used in the detection of possible defects in the main bearing of a wind turbine. Moreover, the method was fast and reliable and provided an estimate of wear development as a function of time [29]. Additionally, this technique has been combined with the slope technique and used in the detection of biofouling [23] in the district heating industry. However, a large amount of simulated data is needed for the algorithm to function appropriately, and there also exist difficulties in establishing the reference value for which the biofouling level is considered too high for the optimal operation of the heater.

The main objective of this study was to validate the effectiveness of the CUSUM statistical technique in the detection of biofouling accumulated in the walls of a tubular exchanger using seawater as the cooling fluid. To this end, the specific objectives were set as follows: (1) allow biofouling growth inside the heat exchanger tubes until the experimental variables indicate that growth has been established; (2) define the amount of biofouling deposited during the different growth phases and assessed by the evolution of the Rf variable over time as the comparison baseline; (3) validate the CUSUM control chart as a technique for the early detection of biofouling against the previously established baseline.

Section snippets

Pilot plant

The experimental plant is shown in Fig. 1 and consists of two tubular heat exchangers in the backflow, which are manufactured according to the guidelines of the Tubular Exchanger Manufacturing Association (TEMA). The outer tube had an external diameter of 240 mm (thickness of 20 mm) and was made of AISI 304 SS. The tube bundle was formed by four independent tubes with a length of 3.1 m in (inner diameter of 10.2 mm and thickness of 2.5 mm) and made with AISI 316L stainless Steel. The outer

Evolution of heat transfer resistance

The Rf progression clearly shows the three growth phases of biofouling (Fig. 4), namely, induction, exponential growth, and levelling-off. The mean, standard deviation, and maximum and minimum values of each phase are listed in Table 3. The induction phase occurred within the first 30 days (700 h of samples). The Rf values suggest that the exponential growth phase occurred in the range between sample 700 and sample 1080, which corresponds to days 30 and 45, respectively. Finally, the

Discussion

Ideal conditions for the initiation of biofouling were created inside the tubes of an experimental plant formed by two heat exchangers using seawater as the cooling fluid. The evolution of biological growth in the heat exchanger was measured by the time progression of the heat transfer resistance, which is a widely used technique that has been validated in numerous studies.

In turbulent flow, Rf represents the sum of the thermal resistance by convection and conduction [41]. The resistance to

Conclusions

The main objective of this study was to validate the effectiveness of the CUSUM control chart in the detection of biofouling on the walls of a tubular exchanger using seawater as the cooling fluid. This was done with regard to the time measurement of the heat transfer resistance Rf, which is typically used for this purpose.

The CUSUM control charts were shown to be effective in detecting deviations in the heat exchanger with regard to previously established optimal operating conditions.

The

References (42)

  • D. Boullosa-Falces et al.

    Monitoring of fuel oil process of marine diesel engine

    Appl. Therm. Eng.

    (2017)
  • F. Agyenim et al.

    A comparison of heat transfer enhancement in a medium temperature thermal energy storage heat exchanger using fins

    Sol. Energy

    (2009)
  • Z. Xu et al.

    Experimental study on microbial fouling characteristics of the plate heat exchanger

    Appl. Therm. Eng.

    (2016)
  • Y. Cho et al.

    Validation of an electronic anti-fouling technology in a single-tube heat exchanger

    Int. J. Heat Mass Transf.

    (1999)
  • S. Singh et al.

    Performance investigation of plate type heat exchanger (a case study)

    Int. J. Eng. Res. Appl.

    (2014)
  • Tubular Exchanger Manufacturers Association, Standard of Tubular Exchanger Manufacturers Association, The association,...
  • C. Ferreira et al.

    Biofilm control with new microparticles with immobilized biocide

    Heat Transfer Eng.

    (2013)
  • S. García et al.

    Impact of the surface roughness of AISI 316L stainless steel on biofilm adhesion in a seawater-cooled tubular heat exchanger-condenser

    Biofouling

    (2016)
  • L.M. Vega et al.

    Parameter estimation of biofouling deposition models on seawater-cooled heat exchangers

  • B. Bansal et al.

    Fouling of Heat Exchangers by Dairy Fluids – A Review

    (2005)
  • A. Trueba et al.

    Influence of flow velocity on biofilm growth in a tubular heat exchanger-condenser cooled by seawater

    Biofouling

    (2015)
  • Cited by (10)

    • Real-time identification of out-of-control and instability in process parameter for gasification process: Integrated application of control chart and kalman filter

      2022, Energy
      Citation Excerpt :

      KF can be applied to state detection [28], fault diagnose [29] and process control [30] of various equipment or processes. Recently, some scholars apply control chart and KF jointly to different system or scenarios [31–34]. In order to implement real-time identification of out-of-control and instability in process parameter for gasification with high accuracy and reliability, the present paper proposed a monitoring model for gasification process using control chart combined with Kalman filter (KCC).

    View all citing articles on Scopus
    View full text