PTQ Q2 2024 Issue

Figure 4 CokeNil dashboard

CokeNil scenario 1 Figure 2 shows Scenario 1, where our model is tested on his- torical data to measure its impact on furnace performance. It was identified that the ethylene furnace, operating under the traditional pattern, reaches the threshold within 25 days from the commencement of a new start. However, upon replac- ing this conventional process with the CokeNil AI model, the forecasted threshold reaches 33 days, signifying an extended run length of roughly eight days (~27%), achieved through the optimal parameters recommended by the AI model. These bespoke enhancements ultimately contribute to increased production, aligning with the organisation’s finan - cial goals. Implementation of the parameters recommended by the model is expected to yield a considerable benefit for the 48 t/hr feed furnace, ranging from 140,000 to 150,000 $/year. This notable financial enhancement can be attributed to the projected reduction of three decoking cycles. CokeNil Scenario 2 Figure 3 illustrates Scenario 2, showing that the forecasted TMT by CokeNil could extend the run length by 18 days (~25%). These contrasting outcomes highlight the marked difference between traditional and advanced approaches to ethylene furnace operation. Upon the implementation of the recommended parame- ters by the model, it is anticipated that the 48 t/hr feed fur - nace could yield a substantial benefit of 70,000-80,000 $/ year. This financial improvement is attributed to the expected reduction in the decoking cycle by 1.6 instances. CokeNil dashboard Figure 4 shows the user interface dashboard for a cognitive solution, analysing various furnace parameters (temperature, pressure, and flow rate) in real-time and suggests adjust - ments to optimise run length by providing actual and rec - ommended parameters for each batch. This could involve

fine-tuning operating conditions, predicting potential issues before they arise, and enabling preventative maintenance. The dashboard includes an advanced analytics panel con - sisting of the following: • Predictive: Displays the actual vs predicted TMT trend with the comparative trend of critical parameters like CPR. • Descriptive: Generates alarms that promptly notify when the TMT or CPR approaches the threshold and provides the visualisation of critical parameters • Prescriptive: Provides the overall benefits statistics. CokeNil benefits for furnace run length include: • Increased run length: The AI model’s ability to optimise operations leads to longer furnace runs, resulting in higher production capacity and reduced downtime. • Improved efficiency: By optimising operating conditions, the AI model helps reduce energy consumption, leading to more efficient production. • Reduced costs: Longer runs, improved efficiency, and fewer unplanned shutdowns can translate to notable cost savings. CokeNil trend analysis The CokeNil trend analysis shown in Figure 5 indicates the key elements of the dashboard: • It displays the non-optimised run length of the furnace vs the optimised run length and the next time the prediction will be updated. It indicates the non-optimised run length is 35 days, and the optimised run length is 45 days, which means the optimised run length is 10 days compared to the non-op - timised prediction. • The user can select process parameters, namely total naphtha and coil outlet temperature (COT), for a specific date range to see its individual behaviour and trends. • The ‘Actual vs Predicted TMT’ graph indicates the accu - racy of the TMT prediction model. • The ‘Furnace RL Increase %’ chart shows that the AI model has increased the furnace run length by 25%.

PTQ Q2 2024