Smart bolts: an of self- healing s example tructures zyxwvutsrqponm By G. Park and D.J. Inman, Center for Intelligent Material Systems and Structures, Virginia Tech, Blacksburg, USA. Structural health monitoring has taken off as a research topic in the past ten years. Currently, there is at least one conference per year devoted to structural health monitoring and several biennial conferences. More importantly, many companies are looking to health monitoring as part of their new business plans to sell service rather than just products. Accordingly, companies have a keen interest in reducing maintenance schedules and downtime for failures causing an increased reliance on health monitoring. The field of structural health monitoring is a natural for combining with smart materials because many of the best algorithms require known or controlled excitation and sensing rather than just sensing. Thus, the actuation ability of smart materials zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA becomes key in health monitoring and diagnostics. The actuation ability of smart materials also has the potential to provide force to a smart structure to counteract damage once it is detected, introducing the possibility of self-healing structures. This article focuses on the use of smart materials to create self-healing structures and provides an example of a self-healing system in the form of a smart bolt. The objective of this study is to reduce significantly resources that are dedicated to inspection routine-s of joint connections and allow systems to function longer between maintenance. We investigate the feasibility of creating smart structural bolted connections, which consist of structural members joined together by bolt and nut combinations equipped with piaoceramic and shape memory alloy elements. These combinations can be used to monitor bolt tension and connection damage. When damage occurs, temporary adjustments of the bolt tension can be achieved actively and remotely in order to restore lost torque for continued operation. zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDC Damage detection problems Structural health monitoring (SHM), also called damage dcrection or diagnosrics. is rhc conccpr of using 5trucrural measurcmenrs IO derermine rhc c ondition. inte g rity o r sta te o f a struc rurc . The m o nito ring concept is b a sic a lly one of inrc rro g a ting a struc tura l syste m with a signal. a na lyzing its response and using the response in some way ro dererminc if rhc srructure has changed in any way. In particular, it is of inrercsr to know if rhc measured change would deter the system’s normal operation. A basic goal of SHM is to dercrminc if a srructure is in danger of failing or nor. Examples of common objectives of SHM are ro dererminc if c ra c ks e xist in the struc ture , if b o lrc d o r we ld e d jo ints a re inta c r and up to sp e c ific a tio n. if the re a re a ny holes or other structural damage present in rhe strucrurc. In some sense, one might fit this topic in with non-dcstructivc evaluation (NDE) mrthods. Ilowevcr. rhc difference berwccn classic NDE methodology and SHM is that of immediacy. In NDE. parts of systems or entire systems are taken out of wrvicc to be inspccrcd and a goal of SHM is to leave the srructure in service while the analysis is performed. An example of current NDE is the NASA space shuttle rhar undergoes a vibration test (modal tcsr) before and after each flight. Differences in m o d a l info rm a tio n a re rhe n used to determine damage incurred during the flight. If vibration information during flight were used to detcrminr damage (‘online’). the sysrern would then capture the funcrion of SHAM. Damage detection methods are reviewed in Docbling. et al’. The problem of srructural damage analysis can hc subdivided into four sub problems or Icvels of hcalrh monitoring: I. Dcrrct the cxi5rence of damage. 2. Dcrcct and locate the damage. 3. Detect, locarc and quantify the damage. 4. Ljctect, locate and quanrify rhc damage rhcn estimate rhc struc rure ’s re m a ining se rvic e life . Eac h level requires more modelling and hence more mathematics to provide a solution. The Level I problem has numerous 5olutions listed in rhe lirc ra rurc a nd in fa c r ha s been implemented in practice. I.evcl 1 problems can be solved without any type of mathematical model, other than an experimental model (Park et al’). Level 2 problems, on the orhcr hand, are difficult to solve without some sort of mathematical model of rhc srruc ture . However, ‘no model’ solutions ro Level 2 problems can be found in the lireraturc formulated bv training a neural network to ‘learn’ every possible damage locarion, and comparing rhe current response to rhc network model5 (Lopes. et al?). Fcw~r solutions are available for the remaining topics. Level 4 Icad to rhe emerging area called ‘Damage Prognosis’. l’his is motivaring several large programmes in g o vc rnm c nt la bo ra to rie s and agcncics and has been the focus of several workshops. Recenrly, rhis lisr has been expanded rn include a number of exrra possibilities: 5. (3ombine zyxwvutsrqponmlkjihgfedcbaZYXWV Lcvcl 4 with smart srrucrurcs IO form self-diagnostic structures. 6. Combine Level 4 with smart structures and control ro form self-healing structures. This last I~vel, self-healing, is rhe focus of inrercst hcrc. Taking the simplified point of view that health monitoring involves looking for changes in a sysrem’s physical parameters (such as mass, damping or sriffness) leads to the observation that much of the marhcmatics associated with SHA4 will COIIIK from the field of parameter identification or estimation. ‘l’his is clear in the work of Hanks et al4 who solve a Level 3 health moniroring problem by adapting techniques from parameter estimation theory. Obviously, each level problem bccomcs more difficult to solvr. Level I problems may be solved without any reference to a structural model by using simple signal analysis. One example of such a solurion is given by Carrarius and Inman5. In rhi5 rcchnique, the idea ofren used in damage dcrection that a defect will produce a small change in stithws and/or mass. and hence frequency is applied. Early vibrarion based damage detection methods often looked at frequency response funcrion (FRF) data for a small change in frcquenc): Howcvrr, small changes in frequency are generally difficulr to mcasurc using FRF data. As an alrernarivc. Cattarius and Inman’ continuously compared rhe healthy time signal of the srrucrurc under study to the current rime signal July 2001 Smart Materials Bulletin