Corrosion monitoring in commercial CFBs Jouni Mahanen, Kyösti Vänskä, Edgardo Coda Zabetta Amec Foster Wheeler Varkaus, Finland Presented at 22nd international conference on fluidized bedconversion Turku Finland 16.6.2015 Amec Foster Wheeler 2015.
Corrosion monitoring in commercial CFBs Jouni Mahanen 1, Kyösti Vänskä 1, Edgardo Coda Zabetta 1 * 1 Amec Foster Wheeler, Power Systems and Technology, R&D Department, Relanderinkatu 2, FI- 78201, Varkaus, Finland *) Corresponding author: phone: +358 (0)400 250 712, e-mail address: edgardo.coda@amecfw.com Abstract Hot corrosion is a common challenge in boilers when firing aggressive fuels such as recycled wood and waste. Mitigating the losses caused by corrosion requires an optimal balance between i) pricy preventive actions, ii) time-consuming scheduled maintenance, and iii) the risk of unscheduled failures causing loss of production and repairs. Such balance is often based on the understanding of corrosion phenomena and on the operational experience, which set somewhat conservative limits on process temperatures and the use of corrosion-resistant materials or protections. A less conservative approach consists in monitoring the progress of corrosion in real time with online corrosion measurements and in parallel tuning the process conditions in order to minimize corrosion. However, online corrosion measurements are not trivial, and available methods are not fully validated and accepted. This paper describes the validation efforts conducted on an online corrosion monitoring technique in several CFB boilers. The method originally developed for mildly aggressive recycled wood was deployed in boilers co-firing a variety of fuels, including coal, woody biomass, agricultural residues, and waste-derived fuels. Validation of the online method was conducted against metal wastage measured in real boiler components, in test pieces temporarily installed in the boilers, and in test probes. These three types of measurements provided full understanding of the online monitor capabilities and its validation for long-, mid-, and short-term corrosion. The online method proved to be suitable for real time corrosion monitoring, although with somewhat varying accuracy depending on the co-fired fuels. With minor procedural changes, the same monitoring equipment also proved to be suitable for online fuel quality monitoring when co-firing erratic shares of several fuels. Keywords: CFB boilers, online corrosion monitoring
1. Introduction Electrochemical online corrosion measurement methods are widely used in monitoring the corrosion of critical components in the chemical process industry. The measurement is usually based on linear polarization resistance (LPR) method, which is shortly introduced in chapter 2. The method is suitable also for monitoring the corrosion of convective heat exchangers in boilers, since the ash deposit that forms on top of the corrosion probe acts as an electrolyte. It must be kept in mind that the composition and melt behavior of the deposit greatly affects the measurement. This paper illustrates cases from several plant measurements, ranging from a boiler firing coal with moderate share of agro biomass to a boiler firing waste-derived fuels. Online measurement results were compared to results obtained from calibration rings (weight loss) which were installed in the same probe, and wall thickness measurements of boiler tubes. Online measurement results were also compared to changes in fuel quality, when applicable. 2. Test methods Electrochemical online corrosion measurements are based on determining the corrosion current within a confined system. Once the corrosion current is known, corrosion rate can be calculated according to Faraday s law, stating that the amount of material that is dissolved to an electrolyte in electrochemical reactions is directly proportional to the current between anode and cathode. In this case the system comprises a corrosion probe apparatus and the deposit forming on it when exposed in a boiler. When the corrosion current through the system is known, a corrosion rate can be calculated. In LPR measurements, the corrosion current is determined indirectly by measuring the corrosion potential and polarization resistance between a reference electrode and a work electrode. The polarization resistance is derived from inducing a small deflecting voltage (typically tens of millivolts) to the system, while measuring the current that is required to induce this deflecting voltage. Based on LPR measurement, Amec Foster Wheeler developed the MECO online corrosion monitoring system. This system has been deployed in nearly 30 boilers to monitor the corrosion of convective heat exchangers during guarantee period, especially in units where challenging fuels are fired. The MECO system consists of a measuring probe which is installed in the flue gas duct, and a control unit which controls the probe cooling and executes the necessary measurements, and based on the measurements calculates the projected corrosion rate in µm/a. MECO can be connected to power plants data collecting system, which allows monitoring corrosion from the control room. The probe can be cooled with water and pressurized air, or only with pressurized air. A conceptual figure of the probe is shown in Figure 1. Two different materials can be installed in the probe at the same time, thus allowing monitoring corrosion of two materials simultaneously. The measuring circuit consists of three test pieces work electrode 1, reference electrode and work electrode 2. In addition to the measuring circuit, there are two test pieces from which the actual corrosion rate can be determined based on their weight loss after exposure. Corrosion rate based on the weight loss test pieces is performed by weighing the test pieces before the exposure, and again after exposure after removing all deposit, oxide and corrosion products from the weight loss test pieces. These results can be further validated to metal corrosion rates obtained from ultrasonic tube wall thickness measurements, which are common practice in the industry and are performed during maintenance outages. In order to obtain meaningful statistics and detailed local corrosion data, ultrasonic tube wall thickness measurements are performed on a very extensive number of points onto the heat exchangers, in order to follow up the corrosion rate, these measurements are performed in the very same points, providing extensive local information. Statistical methods are then apply to determine a representative yearly corrosion rate.[1]
Figure 1. Conceptual figure of the MECO corrosion monitoring probe. 3. Plant measurements The following sections present selected cases from the nearly 30 references where the MECO online corrosion monitoring system was deployed. This selection is intended to demonstrate the validity of the method in boilers firing a broad variety of fuels and fuel mixtures. The selected cases also highlight side features of the system, which allow other than corrosion monitoring. 3.1. CFB boiler, coal and up to 20w-% agro biomass Online corrosion monitoring was conducted for approximately one year in a CFB boiler firing coal and up to 20 weight% biomass. The monitor s probe temperature was set to match the reference temperature in the convective superheaters. An austenitic steel with 25% chromium (1) and a ferritic-martensitic steel with 11% chromium (2) were selected as probe materials. The online measurement showed no corrosion for most of the monitoring period (~1 year). Figure 2 shows an extract of such measurements, stretching for 8 continuous months. The corrosion rate in the figure is normalized to a standardized reference rate, which is not disclosed in this paper. For comparison, corrosion rates obtained from weight loss test pieces (WL) and from ultrasonic tube wall thickness measurements (WTM) that were performed at the beginning and end of the monitoring period are also shown in the same plot. Corrosion rates determined by these three techniques are comparable to each other, confirming essentially no corrosion.
Figure 2. Corrosion probe temperature and normalized corrosion rates determined from online measurement, weight loss test pieces (WL) and tube wall thickness measurements (WTM). Appearance of the corrosion probe removed after exposure is shown in Figure 3. The deposit formed on the probe was very hard, and according to SEM EDS analyzes (Figure 4), it consisted mainly of Ca and S, which is typical for coal combustion when limestone is used to control SOx emissions. No chlorine was found from the deposit nor from the oxide scale, which confirms that the deposit was not corrosive in nature. Figure 3. Appearance of the corrosion probe after exposure.
Figure 4. Analyzed compositions (SEM EDS) of deposit and oxide layers formed on test materials. 3.2. CFB boiler, coal and up to 40w-% biomass, including agro Online corrosion monitoring was conducted for approximately 3 years in a CFB boiler firing coal and up to 40 weight% biomass. The probe temperature was set to match the reference metal temperature in the convective superheaters. The probe temperature was periodically increased in order to i) verify the sensitivity of the corrosion monitor to temperature, and ii) to quantify the corrosion sensitivity of tested materials to temperarure. A ferritic steel with 0,5% chromium (1) and an austenitic steel with 18% chromium (2) were selected as probe materials. During the monitoring period three separate but identical probes were exposed. Normalized corrosion measurements alongside daily average shares of fired biomass are shown in Figure 5. For comparison, normalized corrosion rates obtained from probe weight loss test pieces (WL) and from ultrasonic tube wall thickness measurements (WTM) which were performed on yearly basis during the 3 years testing period are also shown in the same plot. Measurements show an increase in corrosion as the share of biomass increased, although a direct correlation between the corrosion rate and biomass share is not evident because of the large variety of fired biomass, ranging between clean wood chips to challenging agricultural residues.
Figure 5. Share of combusted biomass and normalized corrosion rates determined from online measurement, weight loss test pieces (WL) and tube wall thickness measurements (WTM). The appearance of all three corrosion probes after exposure is shown in Figure 6. The deposit formed on the probe was very hard, and according to SEM EDS analyzes (Figure 7), consisted mainly of Ca and S, which is typical for coal combustion when limestone is used to control SOx emissions. However, in this case, chlorine was found beneath the deposit layer, indicating that chlorine induced corrosion took place during the monitoring period.
Probe I Probe II Probe III Figure 6. Appearance of the corrosion probes after exposure. Figure 7. Analyzed compositions (SEM EDS) of deposit and oxide layers formed on test materials Results from one of the periodical changes of probe temperature are shown in Figure 8. Increasing the probe temperature clearly increased the measured corrosion, and decreasing the probe temperature back to its original value brought the corrosion back to its initial value, with only minimal (if any) memory effect of increased chlorine corrosion that occurred at higher temperature. The best correlation between probe temperature and measured corrosion was obtained from the material (2) i.e. the austenitic 18%Cr steel. Such result demonstrated that - with the fuels and corrosion mechanisms involved in this case - the MECO corrosion monitoring probe can be used effectively as a fuel quality monitoring system.
Figure 8. Effect of probe temperature on measured corrosion. 3.3. CFB boiler, recycled wood Online corrosion monitoring was conducted for approximately two years in a CFB boiler firing recycled wood. The probe temperature was set to match the reference metal temperature in the convective superheaters. An austenitic steel with 25% chromium (1) and a nickel based alloy with 21% chromium (2) were selected as probe materials. Normalized corrosion rate measurement results, probe temperature and normalized corrosion rates determined from probe weight loss rings (WL) are shown in Figure 9. In this case, ultrasonic wall thickness were not performed during the monitoring period, so the comparison of corrosion rates can be done only between online measurement and probe weight loss rings Contrary to the previous cases, here the probe temperature suffered higher fluctuations due to accidental disruptions in coolant supply. Although unplanned, these fluctuations helped to confirm once again the correlation between temperature and measured corrosion, and demonstrated the dynamic behavior of the probe in case of temperature upsets. Boiler tubes wall thickness measurements were not performed during the monitoring period, so the online measurement result can be compared to the corrosion rates determined from probe weight loss rings only. The correlation between the measured and actual corrosion rate for the austenitic 25% chromium steel (material 1) is good.
Figure 9. Corrosion probe temperature and normalized corrosion rates determined from online measurement and weight loss test pieces (WL) Figure 10. Appearance of the corrosion probe after exposure. Appearance of the corrosion probe after exposure is shown in Figure 10. The deposit layer on it was gray and brittle. According to SEM EDS analyzes, Figure 11, the deposit was rich in Ca and S, but also Na, K, Ti and Cl were detected in the deposit. Considerably high content of chlorine was also found beneath the deposit layer, indicating that chlorine induced corrosion took place during the monitoring period. Figure 11. Analyzed compositions (SEM EDS) of deposit and oxide layers formed on test materials
3.4. CFB boiler, forest residue, recycled wood, RDF pellets Online corrosion monitoring was conducted for approximately two years in a CFB boiler firing forest residue, recycled wood and RDF (refuse derived fuel) pellets. Probe temperature was set to match the reference metal temperature in the convective superheaters. An austenitic steel with 18% chromium (1) and a ferritic-martensitic steel with 11% chromium (2) were selected as probe materials. During the monitoring period two probes were exposed. Normalized corrosion rate measurement results, probe temperature and normalized corrosion rates determined from probe weight loss rings (WL) and tube wall thickness measurements (WTM) are shown in Figure 12. Similarly to the previous case, also here the probe temperature fluctuated during the monitoring period. Once again, the temperature variations highlight the effect of temperature on corrosion and the dynamic behavior of the probe. Additionally, here large variations in corrosion rate also resulted from changes in the fuel quality (mixture) during the measuring period. Once again, the corrosion rates determined with the online measurement are well in line with the corrosion rates determined from probe weight loss rings and tube wall thickness measurements. Figure 12. Corrosion probe temperature and normalized corrosion rates determined from online measurement, weight loss test pieces (WL) and tube wall thickness measurements (WTM). Probe I Probe II Figure 13. Appearance of the probes after exposure. The appearance of corrosion probes after exposure is shown in Figure 13. Deposit layer was gray and brittle. According to SEM EDS analyzes, Figure 14, deposit was rich in K, S, and Na. Chlorine
was also found beneath the deposit layer, indicating that chlorine induced corrosion has taken place during monitoring period. Figure 14. Analyzed compositions (SEM EDS) of deposit and oxide layers formed on test materials 4. Discussion and conclusions During the past two decades several publications have reported effective online corrosion measurements based on the LPR method in boilers. Few monitoring systems have been developed based on such technique by different organizations, each demonstrating the potential of this technique but also arising drawbacks, the most critical being: i) the difficulty to adapt the same system to different corrosion mechanisms (i.e. in boilers firing different fuels or widely changing fuel diet), and ii) the unavailability of boiler data such as tube wall material losses to validate the online measurement. This paper illustrates the successful use of one and the same system (MECO) to monitor online the corrosion rate of superheater tubes in boilers that fired fuels ranging from coal to biomass and waste, as well as their mixtures. Only four cases were selected here to represent the extensive experience collected from nearly 30 boilers in over 15 years. Such extensive experience not only provided the bases for a qualitative knowhow of corrosion monitoring, but also provided concrete validation data for its calibration. Not surprisingly, when co-combusting moderate shares of biomass with coal, LPR measurements showed no corrosion, which was confirmed by probe calibration test pieces and tube wall thickness measurements. The deposit (electrolyte) that forms on the probe in these conditions is not corrosive in nature. With higher shares of biomass, the deposit composition and properties change, and the LPR measurements detect this change and warn of increased corrosion. Additionally, testing activities confirmed other two applications for the MECO system, which can be adopted when lagging phenomena such as the chlorine memory effect are avoided. Firstly, the system can be used to monitor fuel quality in real time. Secondly, the probe can scan the sensitivity of corrosion to temperature and find corrosion limits for given combinations of fuels and materials. Based on the above, the MECO is best applied: - in plants that fire aggressive fuels to help estimating the lifetime of convective superheaters and to test the effectiveness of changing tube materials or process temperatures, or. - by in cofiring plants to monitor the corrosivity of fuel mixtures and help avoiding too aggressive combinations.
Acknowledgements This work is partly conducted in projects supported by the Finnish funding agency for innovation (TEKES/CLIFF). References [1] Kunnossapitoyhdistys ry, Korroosiokäsikirja, 2 nd ed. Hamina: Oy Kotkan kirjapaino Ab; 2004. ISBN 951-97101-7-5.