Schatting van toestanden, belastingen en parameters in een breed frequentiebereik voor toepassingen

Accompanying the advent of new technologies like electrification and active control, along with a general improvement in system performance, there is an increasing need for reliable, fail-safe, and accurate measurements of system variables. However, capturing all relevant system states through direct sensing alone presents significant challenges due to geometric, safety, and economic constraints. While specialized sensors exist to measure quantities such as force, sound pressure, acceleration, and temperature, their application is often limited. This PhD research focuses on model-based sensing approaches to overcome these limitations by integrating estimation techniques with available measurements, enabling a more comprehensive understanding of system states. State-of-the-art model-based sensing techniques are typically constrained to low-frequency ranges due to model inaccuracies or sensor limitations, restricting their application in real-time control. Consequently, the automotive industry and other fields often rely on frequency-domain characterization, which is not well-suited for dynamic, real-time applications. To address this gap, this dissertation develops novel frameworks that extend state, input, and parameter estimation across a broad frequency range. Key contributions include the use of observability Gramians in modal state-space models to assess estimation quality, enabling more effective sensor placement and enhanced estimation accuracy. Additionally, methodologies are proposed to extend experimental mechanical system characterization from the frequency domain into the time domain, facilitating virtual point-based force estimation and parameter identification. Further advancements include a detailed investigation of modeling errors and their impact on estimation accuracy in both time and frequency domains. To improve broadband estimation, sensor dynamics are explicitly modeled, addressing bandwidth mismatches that could otherwise introduce biases. All proposed methodologies are validated through numerical and experimental studies on complex mechanical systems, including an automotive body-in-white car door and a rubber bushing. These validations confirm the effectiveness of the developed approaches in enhancing model-based sensing, providing reliable estimation of system states, inputs, and parameters in demanding real-world applications.