Structural Health Monitoring Using Evolutionary Computing

In this system, the problem of damage detection is formulated as an inverse problem. Solving the inverse problem, it is possible to obtain the damage location and degree. Evolutionary computing is applied to solve the damage problem. PSO-GA, based on Particle Swarm Optimization (PSO) and Genetic Algorithms (GA), is used to solve the problem. Furthermore, Improved PSO-GA is proposed here, which can find multiple solutions in order to overcome the difficulty of the inverse problem. A numerical example of water supply network is presented to demonstrate the applicability and efficiency of the proposed method. keywords; Hearth Monitoring, Particle Swam Optimization, Genetic Algorithms 1.Introduction Recently, Health Monitoring system has gained attention in Japan because of the great loss due to the Hyogoken-Nambu Earthquake in 1995 and the damage and deterioration of existing structures. In addition, a lot of structures constructed during high economic growth period are becoming superannuated, so that there arises a big problem that bridge structures cause falling of concrete fragment in Japan. Then, the necessity of the health monitoring of structures is taken up as the major subject. The health monitoring technology has been so far advanced especially in the fields of mechanical systems, nuclear power plants, and aero and astronaut ical engineering . On the other hand, various researches have been made in the field of civil and architectural engineerings. As representative examples, there are such two applications as the damage detection of structures using the vibration test and various sensoring based on optic fiber sensors. Optic fiber sensors have gained a lot of attention because of their stable anti-noise characteristics. 2.Structural Health monitoring [1]-[7] Structural health monitoring is a technology for the continuous monitoring, inspection, and damage detection of structures. As representative examples, there are two applications as the damage detection of structures using the vibration test and various sensoring using optic fiber sensors. For the former cases Neural Networks or If-Then rules are used to identify the damage degree. For the latter cases various sensors are allocated for the expected damage portions. However, there are some problems in the applications. First, it is necessary to require the knowledge and experience concerning the arrangement of the sensors. Next, the sensor value is not always accurate because of noise. In addition, if the sensoring data are collected, the damage location and degree are not always identified uniquely. 3. Particle Swarm Optimization Using Genetic Algorithm 3. 1 Particle Swarm Optimization [8] PSO uses an adaptive algorithm based on a social-psychological metaphor; in which population of individuals search a solution with common information stochastically toward previously successful regions in the search space. Agents (particles) remember the best position of every individual that has the highest fitness value ( Pbest ) and the global best position with the highest fitness value in the entire group. In addition, they have their own present position vectors and velocity vectors, and the velocity speed toward Pbest and G is calculated based on the following equation: best vi = x ⋅ (ωvi + φ1 ⋅ (Gbest − xi) + φ2 ⋅ (Pbest − xi)) Where x that is as inertia factor with the value of from 0.9 to 1.0 and ω is the attenuation factor. φ1and φ2 are generated as a random numbers (maximum is 2). If the velocity becomes greater than the prescribed maximum velocity (V ), the value takes . This enables to keep the search point in the feasible region. Each positions ( max Vmax xi) is updated based on the following equation: