The purpose of initial monitoring is first to measure the as-built structural modal characteristics and use these data to evaluate the accuracy of the designer's model, and through this, the adequacy of the design. The second purpose is to measure the baseline performance to monitor stability over time.  These data can also be used to assist in modeling older structures for the purpose of better retrofitting and upgrading.

Advantages:  Better model accuracy means better predictions of performance under extreme loads and lower cost because structural upgrades can be planned more accurately.

Baseline performance is vital to long-term monitoring.

Limitations:  None.  This step is highly recommended for significant bridges.

Features:  Two important questions are, "What characteristics should be monitored, and at how many locations?"  For each bridge, this will depend on which structural elements the design engineer feels are most critical.  Typical measurements include the following:

• Temperature and wind conditions during the study.  Bridges are non-linear structures, and response characteristics can change under different conditions.
• Acceleration, both global movement, and individual elements.  Sufficient channels are needed to measure simultaneous movements, with preferably at least 64 channels. 
• Strains in critical structural elements under specific loads. 
• Precise baseline locations (position or tilt) should be measured at critical locations.

The recording system should have the following minimum characteristics:

• Data should be recorded digitally with at least 16 bits of dynamic range.
• Sensors must be highly sensitive and capable of recording very small motions (0.005m/sec² or less).
• Bandwidth will depend on the structure.  For example, the fundamental mode of the Golden Gate Bridge has an 18-second period; therefore, sensors with DC (zero Hertz) response are required.  Typically, DC to 50 Hz is recommended.

The recording system must have sufficient capacity to record all 64 channels for many minutes.  This will provide statistical redundancy when analyzing the data.