Chemical & Processing
Oil & Gas
Pharma Biotech
Infrastructure & Design

How Statistical Modelling Benefited an Ethane Cracking Furnace
Sanjay Katrekar, Director - Marketing Ingenero. Cutting-edge software tools and analytical techniques are now being widely employed for Lifecycle Integrity Management. This article gives a gist of how statistical modelling was effectively used to streamline the operation of an Ethane Cracking Furnace to glean substantial benefits in lifecycle cost.

The lifecycle of a typical project consists of a number of stages such as Conceptualisation, Front-End Engineering Design, Basic and Detailed Engineering, Construction, Commissioning, Operation and Decommissioning. Ingenero works with clients worldwide, helping them at each stage of the asset lifecycle to protect the integrity of their key manufacturing assets. Benefits include higher on-stream time of equipment, which directly translates into increased production volumes, lower losses, enhanced safety and lower environmental emissions, thereby improving overall profitability, while continuing to operate the plant in a safe, environmentally friendly and socially responsible manner.

At the Operation stage of a project lifecycle, integrity management involves the correct execution of monitoring, inspection and mitigation activities to ensure that the facility remains fit for purpose at all times. Typical tasks include surveillance, data mining, cleaning, reconciliation, interpretation, analysis, proposing theories to explain observed phenomena, hypothesis testing, drawing of inferences, reporting and corrective action. When such a process is carried out on a continuous basis and in a structured manner using innovative tools and techniques, the benefits get amplified several fold. This case study demonstrates the value addition brought about by optimising operating parameters for enhancing the life of radiant section tubes in ethane cracking furnaces at an ethylene manufacturing facility.

In ethane cracking furnaces, a mixture of ethane and steam is thermally cracked to produce ethylene as the main product and propylene as a valuable co-product. Other by-products such as hydrogen, methane, acetylene, etc which are also produced in smaller proportions, are either further converted in downstream reactors or separated out and used as fuel in fired heaters.

Since unwanted heavy materials such as coke and tar are inevitably generated during the thermal cracking process, a cracking furnace has to be periodically decoked. Thermal stresses produced in the tubes due to sudden changes in tube metal temperature upon activation of trips and during cooling and reheating cycle of a furnace while taking it off line for carrying out maintenance activities, exposure to a high temperature mixture of steam and air during the decoking

process, erosion due to dislodging of coke particles and their high velocity flow inside the tubes, etc. contribute to the daily wear and tear and eventually lead to thinning and aging of tubes with the passage of time. Elongation and bowing of radiant section tubes due to non-uniform heating and cooling, carburisation, creep, corrosion and erosion together restrict the useful life of tubes, which in turn necessitates their periodic replacement. Since these tubes are typically constructed of high temperature nickel alloys, tube replacement is a capitalintensive affair and forms a significant part of the lifecycle cost of a furnace.

The Issue
The conversion of ethane to ethylene in a furnace improves with increasing cracking severity. Hence when plant throughput is not limited by availability of feedstock, there is a tendency to operate at a high Coil Outlet Temperature (COT) in order to maximise the proportion of ethylene in the furnace outlet stream. This operating strategy exposes the radiant section tubes to higher heat flux and occasionally to flame impingement and localised hot spots, consequently also increasing the rate of coking and inevitably hastening the aging of tubes, thereby compromising on lifecycle integrity.

Aging of tubes can be slowed down if temperature swings are minimised. One of the ways to do so is to reduce the number of heating and cooling cycles over the life of the tubes.

Amongst the end-of-run (EOR) criteria specified by furnace vendors, mostly it is either the Tube Metal Temperature (TMT) or the Coil Pressure Ratio (CPR) limit that is reached first. Conscious efforts were, therefore, directed towards extending the run-length of the furnaces by closely monitoring operating conditions, analysing the collected data using high-end analytical techniques and taking appropriate and timely pre-emptive actions on the panel and in the field.

Capturing the Low Hanging Fruits A cursory look at the operating data showed that typically a productive run was being terminated when just one or two tubes reached EOR conditions, while all other tubes were still away from the maximum allowable TMT and CPR limits. In line with the popular adage, ĄThe strength of a chain is the strength of its weakest linkČ, so also the run-length of a furnace between successive decokes is governed by the tube that cokes fastest. Ingenero s actions, therefore, focused on mapping the coking rate in individual tubes, identifying the outliers and controlling the rate of fouling in these rapidly coking tubes, to the extent possible, so that all tubes coke more or less at a uniform rate, as seen from the reduced spread in Figure 1. Some actions that helped achieve more uniform coking of tubes included adjusting the COT / feed trim, changing the frequency of cleaning burners and throttling or even completely cutting off the fuel gas flow to individual burners located near tubes that were coking faster, for ensuring a more uniform heat distribution across all tubes. The effect of applying COT trim and timely correction of fuel gas flows to relevant burners on the rate of rise of CPR in a fast-coking tube is captured in Figures 2 and 3. Each step of the decoking procedure was also critically reviewed, particularly with respect to the time for which excess air and steam at high temperature flowed inside the tubes for effectively burning off the coke deposits. Extending the air-polishing step beyond the minimum requirement has an adverse impact on tube life.

In-depth analysis for greater insight Furnace run-length also depends upon the quality of the last decoking operation. The extent of residual coke is difficult to measure directly; so indirect measures such as percentage of CO2 at furnace outlet are widely used to decide when to end a particular decoking cycle. Traditionally, when two or more successive CO2 readings are observed to be within stipulated limits, it is presumed that all coke has burned off and the furnace is put back on line. Some plant operators prefer to carry out decoking operation for a fixed length of time, without bothering about the CO2 content.

Ingenero adopted a data driven computational technique, Artificial Neural Networks, to capture the quality of the last decoke and predict the next run-length. For this purpose, historical data for the rate of rise of TMT and CPR was analysed to identify similar runs / patterns in each furnace as shown in Figure 4. Common patterns across furnaces were picked and the data across all furnaces was finally reclustered into six different groups as shown in Figure 5.

For each of the six identified clusters, a set of dependent variables was identified and for each such set, a regression model was built. These models are of auto-learning type in order to adequately capture the uniqueness of every run. For each run, the initial five days  operating data was taken for retraining the model. Finally, a comprehensive integrated model, which was a combination of all six models, was built so that the runs lying in between the six identified clusters could also be covered.

As soon as any furnace was brought on line after decoking, the operating parameters of the first few days were fed into the model to predict the expected run-length.

Whenever it was found that the furnace will not last long enough due to poor quality of the previous decoking operation, the partial pressure of hydrocarbons in the tubes was reduced by increasing the steam to hydrocarbon ratio so that the runlength could be stretched. Simultaneously, excess air availability for combustion of fuel gas in the firebox was suitably adjusted by changing the opening of the stack damper.

Using the innovative analytical and practical approaches described above, the average run-length of the cracking furnaces in this plant was successfully doubled, thereby halving the frequency of decokes per annum and contributing towards slowing down the aging process of the tubes.

Ultrasonic thickness measurements on the tubes taken during maintenance shutdowns, along with corrosion curves and mechanical data were used for estimation of damage due to creep, as per API- 530 procedure. The residual tube life was estimated for the observed TMT profile, based on minimum strength curves and corresponding Larson Miller values.

These calculations have confirmed that tube replacement, which was earlier being necessitated once in about 4 to 5 years, can now be done once every 6 to 7 years, thereby significantly improving lifecycle integrity and reducing lifecycle costs.