Several years ago, the facility's operators had reduced the concentration of a particular neutralizing chemical in the process. Now, they thought that this reduction might have been responsible for unusually high corrosion rates.
Using the system to monitor corrosion rates, a test was run to see weather increasing the concentration of the neutralizer would reduce corrosion rates. Quite the opposite happened. Increasing amounts of neutralizer dramatically increased corrosion rates rather than reducing them. This new information not only helped to reduce corrosion rates, but also provided chemical engineers at the plant with new insight into the chemistry of the process.
During the process, parts-per-billion concentrations of catalyst are added to enhance the chemistry of the process. After noticing a dramatic increase in corrosion rates, a plant technician pointed out that the increase occurred right after a new batch of catalyst was mixed. The procedure used to mix the catalyst resulted in small variations in catalyst concentration, and the online monitoring system allowed those variations to be tracked along with the corrosion rate.
The corrosion rate was shown to vary quite significantly with process and operational events. For example, the Pitting Factor and LPR corrosion rate show an inverse relationship when localized corrosion is occurring. Monitoring of general corrosion alone could not have brought these relationships to light. In addition, operators found that the LPR corrosion rate of carbon steel was directly connected with the amount of a key gaseous chemical used in the process, whose addition level inadvertently varied.
Another process stream monitored at the plant is primarily aqueous with normally low corrosion rate. Early in the testing program, short-term spikes to relatively high corrosion rates were sometimes observed. The corrosion rate spikes coincided with pumping of laboratory samples upstream of the corrosion monitoring probes. When plant operators changed their procedure and disposed of lab samples differently, the corrosion spikes ceased.
Operators have taken baseline ultrasonic thickness measurements on various parts of the piping in the vicinity of the corrosion probes. After enough time has elapsed to show reasonably accurate wall-thickness losses, the data will be compared with the cumulative metal loss data to assess the accuracy of the real-time corrosion measurements.The company is considering the system for other plant applications, for instance, corrosion troubleshooting work, chemical treatment of open-circuit cooling tower water.
So far, BASF has found the system effective for monitoring corrosion behavior, Kintz says, and expects to be able to detect and resolve corrosion problems before significant equipment damage occurs rather than reacting once the damage is done.
Figure 4: The probes were housed in explosion-proof and weatherproof (show above) enclosures.
In January 2003, the system was integrated with the site's process information management system (PIMS), enabling plant operators and engineers to view real-time corrosion data alongside standard process variables (temperatures, pressures, flow rates, vessel liquid levels). During the early phases of corrosion testing, a list was made of process variables that could affect corrosion rates. Experiments were then run by adjusting one process variable at a time while monitoring the effect on corrosion rates and pitting factors.
This method soon revealed an intricate relationship between process variables and corrosion behavior. Many process changes influence corrosion, and combinations of process changes produce unexpectedly complex effects on corrosion. In order to address this complexity and maximize the efficiency of the testing program, the site began using the multi-variable testing (MVT) protocol. With this system, a set of process variables is randomly modified all at the same time. BASF has identified 11 key process variables, and will run 12 experiments with a specific process recipe to identify the variables that influence corrosion and synergies between them, Kintz says. This type of corrosion evaluation would not be possible without real-time data.