Liquid fuel penetration was measured using an endoscopebased imaging system in an operating single-cylinder heavy-duty direct injection diesel engine with simulated turbocharging. Sprays were imaged via the elastic backscatter technique without significantly altering the engine geometry. Light loads (or pilot injections) were also studied because the spray breakup, mixing and vaporization processes can be isolated since they are less influenced by heat feedback from the flame than in a full injection case. The pilot injections included cases with three different fuel amounts (10%, 15% and 20% of the fuel injected in the baseline case, i.e., 75% load and 1600 rev/min) with different start-of-injection timings. Maximum liquid penetration lengths beyond which the fuel is completely vaporized were observed for all the cases studied. The maximum lengths varied from 23 mm to 28 mm for the different start-of-injection timings. The experimental cases were modeled using an improved version of KIVA-II. In the modified computer code, the competing breakup mechanisms of Rayleigh-Taylor (RT) and Kelvin-Helmholtz (KH) instabilities were implemented. Furthermore, a new breakup length model was introduced into the code to account for the intact core region of high pressure diesel sprays. The modified breakup model was applied to simulate the present experimental cases. In addition, the new breakup model was also validated by comparing with extensive measured in-cylinder pressure over a wide range of engine operating conditions. It is found that the numerical predictions of liquid spray penetrations and pressure histories with the RT-KH spray breakup model give satisfactory agreement with the measurements.