Increase Plant Productivity

Process / Plant Optimization – Performance Monitoring

Increase plant productivity through online performance monitoring

The technology can increase production by 1 – 3% and reduce unplanned maintenance costs by 10 – 30%

June 2002

D. C. White, Emerson Process Management, Houston, Texas

________________________________________

The process industries have always been concerned with improving reliability of major equipment and avoiding unscheduled production slowdowns or shutdowns. As process plants get larger, the financial implications of even relatively short production outages are quite high. Reliable and efficient equipment operation is essential for profitable production. Performance of process equipment, such as compressors, heaters, heat exchangers and columns, often deteriorates with time due to wear and tear and fouling, and this deterioration has significant economic implications.

In addition, deterioration in process equipment performance in conjunction with traditional condition monitoring is often a precursor to actual equipment failure. Even when process data are monitored online, actual equipment performance, which is not directly measurable, can be masked by changes in stream compositions, operating conditions, ambient conditions and other normal process variations. To correct for these variations, it is necessary to use rigorous engineering models of the performance and the best possible estimate of the correct values of process input data to the calculation.

Fortunately, continuing advances in computers and communication have dramatically reduced costs for developing and implementing this rigorous performance monitoring. An attractive current option is to obtain these applications, as a service, over the Internet, vastly reducing the cost and complexity of implementing and maintaining them. The required investment in hardware, operating systems, databases, software licenses, IT staff and third-party support is significantly reduced.

In this article, we present the results of installing this technology on processing units. Technical details and economic benefits are given. Important associated benefits include the opportunity to benchmark similar equipment across multiple plant sites and the ability to use centralized engineering resources to support multiple sites.

Assets profitability. The process industries continue to experience negative trends in operating margins and increased regulatory oversight, particularly in the health, safety and environmental (HSE) area. To improve competitive operation and profitability and HSE performance, enhanced equipment reliability is critical. Maintaining efficient operation and performance of plant equipment is essential to long-term process profitability and safe operation.

In a typical process plant, maintenance expenditures are the largest costs after feedstock and utilities. However, the cost of poor reliability is even higher. Many plants report losses in production capacity of 3 – 7% due to unscheduled shutdowns and slowdowns of major process equipment.

Another factor significantly affecting plant profitability is timing of maintenance and cleaning schedules for major equipment items. This maintenance and/or cleaning is required since most plant equipment performance degrades with time. Not only does performance degrade between overhauls, but also with each successive overhaul the performance gains may be less.

Several approaches to maintenance can be used in process plants. One is to wait until the equipment breaks and then fix it if it is really important. The second, known as preventive maintenance, uses average times to failure for equipment and schedules maintenance before the expected failure time. However, equipment can vary widely in actual performance. Some scheduled maintenance is unnecessary and some equipment fails before its scheduled shutdown. Predictive maintenance attempts to find techniques to determine more precisely if equipment is underperforming or about to fail. Predictive maintenance techniques have been demonstrated to save as much as 30% of unscheduled maintenance costs, while simultaneously improving equipment reliability.

We are all aware of the tremendous decrease in physical size of high-performance computing equipment and the increase in communication bandwidth and capabilities. Sensors on equipment are becoming cheaper, with enhanced computing and communication capabilities included. With continuing improvement in computing and communication capabilities, predictive maintenance can be based on actual device performance data, obtained and analyzed in near real time. The objective is to catch potential equipment problems early, which leads to less expensive repairs and less downtime. Conversely, we want to avoid shutting expensive equipment down unnecessarily.

Fig. 1 illustrates the concept. We would like to detect anomalies early and then decide what they mean with respect to the equipment. Data from the process and equipment are validated and brought to performance models. These calculate the equipment performance and correct to standard conditions, and use economic information to calculate the cost of poor performance. This can be used for predicting unscheduled removal (or replacement) of part(s), disruption of service or capacity delays.

Fig. 1. The goal is to detect anomalies early and then decide what they mean with respect to the equipment.

The collective software technologies that address this area are known as asset management systems. Asset management system technologies are designed to ensure that production equipment is maintained at the maximum performance level for minimum cost. Assets in this context refer to all of the physical equipment in the plant – compressors, pumps, distillation columns, heat exchangers, boilers, etc. A key part of asset management systems is equipment performance monitoring to determine its current condition and enable predicting its operation in the future.

Model-based equipment performance monitoring and analysis. Let’s assume you are responsible for a major piece of equipment in the plant such as a large heater. Suppose you see efficiency go down and fuel usage go up – what does it mean? Variations in air temperature, equipment and fuel gas composition, as well as operating decisions, can easily cause efficiency to vary by 20%. Obviously, you need some way to bring the performance to standard conditions that can be compared with historical, design and clean conditions.

In the past, plant performance has been monitored by developing either simple input / output algorithms or by processing data that have first been fitted to a simple correlation-based equipment model. While both of these methods can produce acceptable results, they still require a significant investment in software and hardware technologies and skilled technicians to oversee the monitoring infrastructure. Some organizations still continue to retrieve raw plant data and calculate performance using homemade spreadsheet and base timing of maintenance and cleaning schedules on these calculations. Such techniques are hard to maintain long term and are prone to errors.

For a true representation of current operation of any piece of plant equipment, a model-based performance monitoring system provides greatest accuracy. This technique allows the user to extract or infer information about the machine from the operating data by use of the mathematical representation of the equipment and the calculation engine.

Technology description. Fig. 2 shows modern equipment performance monitoring technology as installed in leading process plants today.

Fig. 2. Modern equipment performance monitoring has been installed in leading process plants.

Parameters are selected for each of the unit operations being monitored which have an engineering meaning: for example, overall heat transfer coefficient (UA value) for a heat exchanger or isotropic efficiency for a compressor stage. Each of these can be expected to respond to fouling in the unit, in both the previous cases by decreasing in magnitude with time. The parameters are then corrected to standard conditions. For example, heat transfer coefficients vary significantly with the fluid processed, and its flow rate and viscosity, which are further functions of its operating temperature. Simple calculation of a UA does not permit accurate determination of fouling factors. Correcting the calculation to standard conditions is necessary to permit useful information to be extracted from the data.

The rate at which the parameters are recalculated should be chosen to reflect anticipated decay in their value. If frequency is set too high, then there will be no significant change from one estimate to the next; if too low, it is possible to miss a critical shift.

Heat exchangers’ fouling rate obviously depends on the fluid being processed, so the appropriate update may differ from one exchanger to another. A standard recommendation is to recalculate fouling indicators two or three times per day. This is a higher frequency than any expected change; however, it is necessary for the statistical verification that follows. The value used for prediction is more likely to be a daily or weekly moving average.

Other types of equipment have very different rates of change. For some compressors and reactors it may vary over many months, while an offgas CV may change from minute-to-minute.

It is important to establish that the chosen parameter is a "good" performance indicator, i.e., that it has the following properties:

• The value

...