An Experimental Study of Clogging Fault Diagnosis in Heat Exchangers Based on Vibration Signals
The water circulating heat exchangers employed in petrochemical industrials have attracted great attention in condition monitoring and fault diagnosis. In this paper, an approach based on vibration signals is proposed. By the proposed method, vibration signals are collected for different conditions through various high precision wireless sensors mounted on the surface of the heat exchanger. Furthermore, by analyzing the characteristics of the vibration signals, a database of fault patterns is established, which therefore provides a scheme for conditional monitoring of the heat exchanger. An experimental platform set up to evaluate the feasibility and effectiveness of the proposed approach, and support vector machine based on dimensionless parameters is developed for fault classification. The results have shown that the proposed method is efficient and has achieved a high accuracy for benchmarking vibration signals under both normal and faulty conditions.
Heat exchanger is a common device used to transfer heat. In modern refineries, the cost of heat exchangers is around 40% of the total investment in a facility; the corresponding maintenance workload can be as high as60% to 70% of the total maintenance workload. Therefore it is extremely important to make sure that heat exchangers operate safely and economically. Instead of purified water, the cooling medium used in a head exchanger is industrial grade circulating water, which continues to evaporate and effloresce during use; thus its salt content constantly increases. Meanwhile, carbon dioxide in the water turns into its gaseous form and escapes from the cooling tower, leading to the continuous formation of calcium carbonate scales within the heat exchanger. Further, when the circulating water is exposed to the air, a large amount of dust as well as mud pellets and microorganisms dissolve into the water. Much of this material is then deposited on the inner walls of the heat exchangers tubes, forming scales. According to a survey by Steinhagenet al. , of 3000 heat exchangers of different types used by 1100 companies in New Zealand, more than 90% showed varying degrees of scaling within their tubes.
Over time, scaling on the inner walls of the heat exchangers tubes will cause the fluid inside to change its direction of flow, thus generating a transverse force that causes the tubes to vibrate. If some of these tubes are heavily scaled or blocked given that the velocity of fluid at the inlet is fixed the flow rate and pressure in the other tubes will be increased and the heat exchanger will vibrate violently. If this condition is not treated in a timely manner, leakage will ensue, leading to a significant financial cost. According to the statistical data, some 30% of heat exchangers break down because of vibration problems, indicating that the vibration of tube bundles is a major cause of damage . This issue has become a key cause of concern in petrochemical industries; that is, it is considered vital to monitor, in a timely manner, the clogging conditions in heat exchangers during use so as to reduce the frequency of their maintenance and ensure the safe operation of these devices by analyzing their vibration signals. Major facts discussed in this paper include the following:
•Starting with the problem of monitoring, the vibration mechanism of the water circulating heat exchangers described and the corresponding research reviewed. Finally a condition monitoring approach based on the heat exchanger’s vibration signals is proposed.
•The following procedure is described: An experimental platform for a water circulating heat exchanger was established and a support vector machine (SVM) used to classify the dimensionless parameters of its vibration signals. This procedure verified the divisibility of the vibration signals generated by the heat exchanger under both normal and fault conditions.
- CONDITION MONITORING OF HEAT EXCHANGERS
The fault diagnosis of heat exchangers has posed a major challenge in the industry because faults such as leakage and clogging within the heat exchange tubes are not visible. The fault monitoring techniques used in todays refineries are relatively outdated . , the major methods being the monitoring of PH values, oil content, chemical oxygen demand, residual chlorine, the functioning of specific instruments and etc. For example, in the Maoming Ethylene Factory, the main fault monitoring method used with its water circulating heat exchanger is to determine the difference in temperature at the water inlet versus the water outlet. However, the accuracy of this method is influenced by many external factors. For instance, the temperature of the medium entering the exchanger may not be stable; it can fluctuate to some extent, and this will have a direct effect on the temperature difference. However, by developing an online device to monitor the heat exchanger, sensors can be installed in different parts of it so as to monitor changes in the circulating water continuously, thus enabling online detection and identification. This method has achieved considerable success , with commercial products available for use in countries including the United States, Germany, and Japan . As described in reference , a single chip microcomputer (SPCE061A) was utilized to construct an online monitoring and controlling system designed to detect the temperature at the inlet and outlet as well as that of the water vapor; it can also measure the flow rate and control the flow of both water and vapor. Stojan Persin et al.  applied analytical fault detection techniques to do real time fault diagnosis for an industrial scale pilot heat exchanger. An online monitoring and prediction system is proposed in reference , which explores the way in which this device monitors and predicts a heat exchangers performance, resulting in improved production efficiency of each individual unit, lower costs of production and maintenance, as well as a higher level of safety during production. The online performance monitoring of a shell and tube type heat exchanger using steam and water has been developed on the basis of a theoretical model to monitor input and output process variables . Adaptive observers used to estimate the overall heat transfer coefficient and detect a performance degradation of the heat exchanger . Fuzzy models based on clustering techniques are used to detect leaks in a complex heat exchanger . SVM and relevance vector machine (RVM) have been widely used in the condition monitoring and fault diagnosis of machinery . The particle swarm optimization algorithm is applied to estimate the parameters of a SVM which is used to predict faults in a heat exchanger . The Dynamic Principal Component Analysis (DPCA) method and a set of Diagnostic Observers (DO) were designed to online detect and isolate faults related to sensor or actuators malfunctions in a shell and tube industrial heat exchanger , . Dragan  presented a two stage modeling procedure based on prior knowledge and recorded data to detect faults for a heat exchanger in an incineration unit. Ingimundardóttir and Lalot  presented a method for the detection of fouling in a cross flow heat exchanger using wavelets. Paper  proposes a new method for fouling detection in a heat exchanger. It is based on the modeling of the system in a fuzzy TakagiCSugeno representation. With this representation, the design of a fuzzy observer with unknown inputs of polynomial types is obtained via a LMI formulation. Shuet al.  and Zhuet al.  report a fault diagnosis method in petrochemical water circulating heat exchangers based on vibration and sound. Preliminary results show that it was difficult to identify faults by sound and that further research is required concerning vibration signals. From these reports it is clear that several researchers have studied the online monitoring and fault diagnosis system of water circulating heat exchangers; however, these investigations are limited to the monitoring of parameters such as temperature, pressure, and flow rate, with little progress in the fault diagnosis of heat exchangers. Besides, since a heat exchanger is a static device and vibrations generated by the circulating water within the tubes is far less than that generated by mechanical rotation, little research has focused on condition monitoring utilizing vibration or sound parameters.
- RESEARCH ON VIBRATIONS IN HEAT EXCHANGERS
The movement of fluid in a heat exchanger is very complicated, including cross flow, axial flow, and bypass flow in tube bundles; moreover, there are stagnant areas at the inlets and outlets of the tube bundles. The rate and direction of flow of various fluids continue to change irregularly, creating heterogeneous force field for the heat transfer tube. Then the heat transfer tube vibrates under the influences of several forces initiated by the flowing fluid. When the frequency of induced vibration comes close to the machines inherent frequency, the heat exchanger will vibrate violently, causing damage to the baffles and the joint between the tube and baffle. Further, the resonant vibrations of tube bundles, pumps, and compressors; the direct pulsating impact generated by rotating machinery; and the frequent on off switching of the heat exchanger all induce the tube bundles to vibrate, finally causing them to fail. Investigations of vibration conditions in heat exchangers have been conducted by several research groups. Feng et al.  utilized Fluent software to simulate the characteristics of flow induced vibrations in a singe straight tube, two parallel tubes, two tandem tubes, and tube bundles, obtaining corresponding dynamic responses and characteristics of the flow field. Considering aspects of modal shape and dynamic responses, Zheng et al.  analyzed the influence of vibrations induced by pulsating flow on the vibrations of heat exchangers. Wang et al.  studied the inherent vibration characteristics of planar elastic tube bundles and vibration patterns of the same elastic tube bundles under the influence of pulsating flow. They used Fluent software to achieve a simulated calculation of the pulsating flow generator; as a result, a vibration frequency in agreement with the real frequency was obtained. Chen et al.  fulfilled equivalent simulation analyses of vibrations in a heat exchanger using Ansys software and studied various factors that caused vibration of tube bundles. According to them, vibration of the tube bundles can be effectively avoided if the excitation frequency is kept away from the inherent frequency. Nevertheless, all this research was carried out based on the inherent vibration characteristics of the heat exchanger without any analysis of vibration conditions with regard to fault diagnosis. In the present paper, the flow rate of the fluid is considered to be the direct factor that causes the tube bundles in a shell and tube heat exchanger to vibrate. When fluid runs through the tube and the Reynolds number reaches a certain value, asymmetrical alternately shed vortex waves namely Karmans vortex streets appear periodically on both sides and the back of tubes. In the past couple of decades, studies off low induced vibrations have made significant progress in other countries and a highly accepted mechanism of flow induced vibration has been proposed, e. g. , vortex shedding and turbulent buffeting. The alternate formation and shedding of vortex generates a periodic exciting force perpendicular to the direction of fluid flow on both sides of tubes and causes the tubes to vibrate. The erotically, different vibration signals will be generated at distinct flow rates and under varying fault conditions in a heat exchanger, thus laying the foundation for this research.
A condition monitoring and fault diagnosis method based on wireless vibration sensors in water circulating heat exchangers has been proposed. By establishing an experimental platform for a water circulating heat exchanger and performing a series of experiments that simulate clogging faults, it has become clear that vibration signals generated by heat exchange under normal and clogging conditions are significantly different. Based on dimensionless parameters in an SVM classification model, it is clear that the difference between vibration signals collected under normal conditions versus those collected under fault conditions is greater than 85%. Therefore it is feasible to set up a condition monitoring system for a water circulating heat exchanger based on vibration signals. However, although this SVM model provides relatively high accuracy in identifying whether the heat exchanger is clogged, further research is required to improve its diagnostic efficiency and accuracy (e. g. , with regard to the exact number of clogged tubes). In the future, an improved multilevel fault classification device can be designed based on the number of clogged tubes, thus providing better monitoring.