The stability of beam pointing in a laser system depends on the consistency of the optical mirror mount. Typically, a locking mechanism is used to secure the adjustment mechanism after beam alignment, ensuring the mount’s stability. However, this process can introduce errors, causing a drift in the optical path. To mitigate this issue, in this study, an interference fit adjustment screw was designed. This development enables the mechanism to self-lock after beam alignment, thereby preventing optical path drift and enhancing overall stability. Specifically, 14 long-term thermal shock stability tests, each lasting 2500 min, were conducted to validate the proposed design. The experimental results showed that the thermal drift of the interference fit adjustment screw was reduced by 47.16%, thermal shift was reduced by 79.59%, and the long-term stability improved by at least 48.67%.

In nanometric displacement measurement, laser interferometry exhibits high sensitivity to sub-period displacements. However, the interferometric signals are often incomplete in period and limited in amplitude, making them susceptible to high-frequency noise and drift. Therefore, an improved nanoscale displacement measurement method is proposed, combining band-limited dynamic-noise-assisted complete ensemble empirical mode decomposition with adaptive noise (CEEMDAN) with sinusoidal-regression demodulation. By constraining the noise band and adaptively tuning the noise amplitude during injection, the low-frequency trend component is effectively purified. Based on the prior parameters, such as laser wavelength and displacement rate, to constrain the fitting solution space, the linear least squares fitting method is used to process the trend signal to achieve high-precision phase recovery and quantitative displacement measurement. Experiments show that the method remains stable under electromagnetic interference and impulse noise conditions, improving the signal-to-noise ratio by about 2.6 dB. The measurement error is 1–3 nm within the 20–300 nm range, and the stability is improved by approximately 62%, while also satisfying real-time feasibility. Overall, the method achieves nm-level accuracy with enhanced robustness to industrial noise and real-time computational efficiency.