smart technologies for safety engineering

124 Smart Technologies for Safety Engineering
Table 4.1 Comparison and summary of impact load identification procedures (F denotes force; aA
accelerometer (falling body); aB accelerometer (piston rod); p pressure; b photo switch)
Solution map Force and acceleration Momentum conservation
Sensors F F, aB F, aA p, b F, aA
Solution maps 2 1 — 1 1
Operation time [ms] 100–110 80 5/80 5 5
Accuracy (mass) 5 % 5 % 10 % / 5 % 15 % 5 %
Accuracy (velocity) 5 % 10 % — Measured 3 %
In general, the identification of both impact parameters based on the characteristics of the
first impulse is possible relatively quickly, already 5 ms from the very beginning of the impact
process. Therefore, the approach can operate in real-time.
4.1.7 Comparison of Approaches
A real-time impact mass and velocity estimation is extremely difficult with only one sensor.
Much better results can be obtained with two sensors; in the tested case good results were
obtained with a combination of an accelerometer and a force sensor. However, the location of
the accelerometer is very important: fixing the accelerometer to the falling body allows precise
mass identification in a short time after the impact, while with the accelerometer placed on the
impacted body the identification is far more difficult and only feasible after a longer period.
Therefore, mass determination of the impacting body is possible in a short time and with high
accuracy only when one of the sensors is fixed to the impacting object or when the parameters
of its movement are directly measured.
For velocity identification on the basis of acceleration, the initial condition is needed. Thus
in the case of the analysed structure, the velocity identification was feasible only when a joint
movement of the piston rod and the falling mass was observed.
The tested algorithms are summarized and compared in Table 4.1.
4.2 Observer Technique for On-Line Load Monitoring
The concept of the observer was introduced by Luenberger in the beginning of 1970s [7].
Originally it was devoted to feedback control systems, since first multivariable control strategies
had assumed that the whole state vector was available for feedback [8]. However, from the
practical point of view, it was not always possible to measure all the components of the state. The
state observer, which allows the state vector to be estimated based on only a few measurements,
encountered great interest among the control engineering community [9,10]. Another important
issue connected with real-life measurements was noise corrupting the measured signal. With
the aid of the theory of stochastic processes, this problem was solved by Rudolf Kalman and
nowadays his stochastic estimator is frequently called the Kalman filter [11].
Currently, a lot of attention is paid to simultaneous reconstruction of both the state and the
inputs in linear and nonlinear systems [12–15]. In the case of a mechanical system, the state
can be represented by displacements and velocities, while the inputs can be external forces.