Revamps 2026 Issue

in parallel twice per shift, and to feed-rate stepped trials at 95%, 100%, and 105% of design. Pyrometer-to-waveguide agreement at the time of survey was within 6°C across all 36 instrumented tubes after a one-time emissivity correction. A soft-sensor model was developed on the I-PAMS plat - form during the same period, combining a physics-based heat balance with a data-driven correction trained against the new skin temperature field, producing per-row remaining-useful-life estimates and a coking-and-carbur - isation propensity index based on the rolling integral of high-skin-temperature minutes. Results and performance improvements The first significant finding emerged within 72 hours of com - missioning. Across the 288 new measurement points, the peak observed tube-skin temperature during normal oper - ation was 928°C, against a pyrometer survey peak from the previous week of 902°C, which is a 26°C difference traced to a single tube on the south wall that had been running in a blind zone between two pyrometer survey paths. The tube was located adjacent to a burner tile that had degraded to the point of generating localised flame impingement; the tile was replaced at the next opportunistic intervention. Tube-to-tube variation, recorded systematically for the first time, had a standard deviation of 28°C across the 36 instrumented tubes at constant feed rate and constant fuel composition. The hottest tube was approximately 50°C warmer than the coolest tube. Row-to-row variation in mean skin temperature was 22°C, with the south rows run - ning consistently hotter than the north rows owing to the unequal ageing of the burner tiles. The intervention campaign was conducted in three steps over two months. First, the wall burners were trimmed level by level: six burners on the south side were partially throt - tled, while four on the north side were brought up. Second, a controlled rebalancing of the steam-methane feed across the inlet manifold was performed to lift flow through the cooler tubes. Third, excess oxygen was reduced from 2.4% to 1.7% in the radiant box, made possible by the new vis - ibility on localised hotspots. After these interventions, the standard deviation of tube-to-tube temperature fell from 28°C to 12°C, and the row-to-row mean spread fell from 22°C to 9°C. Quantitative outcomes are summarised in Table 1 . The 24°C reduction in peak tube wall temperature,

and the waveguide was secured with stainless-steel band- ing at 200 mm intervals. Cable routing was through three observation doors per side wall, with high-temperature ceramic-fibre sleeves protecting the cables from direct radi - ant flux. Two acquisition units, weather-rated for ambient operation between -20°C and +60°C, were installed on the tunnel-platform structure outside the radiant box. The acquisition units delivered the 288 channels through a Modbus TCP gateway to the plant DCS and to the historian, where dedicated tags were configured for each measure - ment. A heat-map screen was added to the DCS operator console, with a colour-mapped tube layout showing peak skin temperature per instrumented tube and an anomaly flag for points exceeding the per-tube alarm threshold (see Figure 6 ). The commissioning campaign covered four weeks. The first ten days were dedicated to baseline acquisition at the prevailing feed rate. The following two weeks were devoted to correlation runs against the gold-cup pyrometer, operated Figure 6 Field-mounted acquisition unit and panel-mounted human-machine interface (HMI). Per-tube peak skin tem - perature trends and current values are visible on the HMI; the acquisition unit feeds the DCS, the historian, and the I-PAMS platform in parallel

applied to the Larson-Miller correlation for the installed alloy and the design wall thick - ness,3 , 5 projects an additional 22 months of useful tube life before the design rupture margin is consumed. The tighter distribution allowed the operator to lift the average tube wall temperature by 8°C closer to the design optimum without violating the per-tube alarm, which translated to a 1.4% increase in hydrogen production at the same firing rate. Specific energy consumption fell by 0.6 GJ per tonne of hydrogen, equivalent to approx - imately 9,000 tonnes of CO₂ per year on a refinery-wide carbon accounting basis.

Operating performance comparison before and after the multi-point sensor deployment and the subsequent burner trim, feed rebalancing, and excess-air optimisation campaign

Parameter

Before

After

Change

Peak tube wall temperature, °C

928

904

-24 -16 -13

Tube-to-tube spread, std deviation, °C Row-to-row mean spread, °C Average tube wall temperature, °C Excess oxygen, radiant box, % Hydrogen production, % of design Specific energy, GJ per tonne H₂ High-skin-temperature minutes, monthly

28 22

872

880

2.4

1.7

-0.7 +1.4 -0.6

98.6 13.4

100.0

12.8 240

1,860

-87%

Table 1

Revamps 2026