Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
Session IV: Materials State Awareness Application Issues ISSUES AND IDEAS IN STATE AWARENESS FOR REALISTIC MATERIALS AND STRUCTURES Douglas E. Adams, Purdue University Structural state awareness is a process through which loading, damage, and performance in structural material components are identified through a combination of offline nondestructive testing and online monitoring. The basic building blocks and premise of structural state awareness are first reviewed. The position taken here is that key research issues have been revealed as structural state awareness technologies are implemented. Some of the key barriers to implementing structural state awareness methods are then described using specific applications in military ground vehicles, composite weapons systems, and rotorcraft. It is believed that many of these barriers to implementation involve materials state awareness solutions. One possible method for classifying research topics in materials state awareness is then derived using modern systems theory as a catalyst. This classification method is organized according to material, component, model, measurement, environment, and data analysis issues. After providing a list of key challenges in materials state awareness, two noncompeting visions are articulated for short- term and long-term research, respectively, in materials state awareness. 31
32 Proceedings of a Workshop on Materials State Awareness COUPLING MATERIALS STATE AWARENESS WITH STRUCTURAL HEALTH MONITORING AND DAMAGE PROGNOSIS Charles R. Farrar, Los Alamos National Laboratory The process of implementing a damage detection strategy for aerospace, civil, and mechanical engineering infrastructure is referred to as structural health monitoring (SHM). This process involves the observation of a structure or mechanical system over time using periodically spaced measurements, the extraction of damage-sensitive features from these measurements, and the statistical analysis of these features to determine the current state of system health. For long- term SHM, the output of this process is periodically updated information regarding the ability of the structure to continue to perform its intended function in light of the inevitable aging and degradation resulting from the operational environments. Under an extreme event, such as an earthquake or unanticipated blast loading, SHM is used for rapid condition screening. This screening is intended to provide, in near real time, reliable information about system performance during such extreme events and the subsequent integrity of the system. Once damage is detected, damage prognosis (DP) is employed to predict the remaining useful life of a system, given some estimate of the future loading conditions that the system will experience. Currently, for most complex engineering systems accurate DP is not feasible with existing engineering capabilities. To date, SHM and DP studies have for the most part been carried out independent of the materials science community. It is the authorâs speculation that damage detection, as determined by changes in the dynamic response of systems, has been practiced in a qualitative manner, using acoustic techniques (e.g., tap tests on train wheels), since modern humans have used tools. More recently, this subject has received considerable attention in the technical literature. However, with the exception of condition monitoring of rotating machinery, there are very few instances in which SHM technology has made the transition from research to practice. A review of the literature reveals several outstanding challenges for transitioning SHM technology from research to practice. These challenges include the following: Structural monitoring versus structural health monitoring. Many sensor systems currently being deployed on real-world structures are actually structural monitoring systems, as opposed to SHM systems. They are simply sparse arrays of sensors deployed with no a priori definition of the damage to be detected and no definition of the methods for feature extraction and statistical classification that will be used to identify damage. Local versus global damage detection. The most fundamental challenge is the fact that damage is typically a local phenomenon and may not significantly influence the global response of a structure that is normally measured during operation. Defining damage a priori. The success of any damage detection technique will be directly related to the ability to define the damage that is to be detected in as much detail as possible and in as quantifiable terms as possible. Defining the requisite sensing system properties. A significant challenge for SHM is to develop the capability to define the required sensing system properties before field deployment and, if possible, to demonstrate that the sensor system itself will not be damaged when deployed in the field.
Session IV: Materials State Awareness Application Issues 33 Accounting for operation and environmental variability. When deployed on a structure outside of a controlled laboratory setting, the damage detection process will have to deal with structures that experience changing operational and environmental conditions. Need for long-term proof-of-concept studies. There are very few long-term SHM studies ongoing on real-world structures. Theses studies are difficult to perform because of costs. However, such studies are needed before structure owners and regulators will accept SHM as an acceptable means of condition-based maintenance. Lack of data from damaged systems. Few system owners will allow engineers to damage their structure in an effort to validate a damage detection approach. Even if such studies were allowed, in almost all cases damage is introduced in an artificial manner and it is questionable if such âdamageâ is truly indicative of the actual damage that will be encountered in the field. Time scales associated with damage evolution. Damage can accumulate over widely varying time scales, which poses significant challenges for the validation of SHM sensing systems in the field. Nontechnical issues. In addition to the challenges described above, there are other nontechnical issues that must be addressed before SHM technology can make the transition from a research topic to actual practice. These issues include convincing system owners that the SHM technology provides an economic benefit over their current maintenance approaches of using quantified benefit-cost analyses, and convincing regulatory agencies that this technology provides a significant life-safety benefit. Also, universities have to overcome inherent barriers to multidisciplinary research. Tenure and promotion at U.S. universities still primarily reward the individual investigator, and SHM is too multidisciplinary for most individual investigators. Masterâs and doctoral students still need focused research topics for their theses and dissertations, and so they tend to be trained as specialists. If a group of graduate students with different backgrounds work collaboratively on a project, one must be the technology integrator, and this role often does not help these individuals toward the completion of a dissertation. Finally, the U.S. university education system is not evolving to train more multidisciplinary technology integrators and leaders for the future. Such training must always be balanced with the continued need for the technology specialist. These barriers and conflicting goals have led to a state in which the SHM research community is divided into two distinct subdisciplines: (1) those developing data analysis procedures and (2) those developing sensing technology. However, to develop effective SHM solutions these technologies need to be developed in a coupled manner. These issues are coupled with many industriesâ short research time horizons that are on the order of a 12- to 18-month time to market. The materials science community can contribute significantly to the further development of effective SHM/DP technology in three general areas: (1) developing new materials for sensing, (2) developing better fundamental understanding of damage mechanisms and the associated changes in materials properties that are indicators of damage, and (3) developing more robust multiphysics damage evolution models. Significant future developments of SHM/DP technology will need to come by way of multidisciplinary research efforts in which fundamental
34 Proceedings of a Workshop on Materials State Awareness materials science is coupled with mechanical and electrical engineering to systematically address the issues listed above. STATISTICAL ISSUES RELATED TO MATERIALS STATE AWARENESS William Q. Meeker, Iowa State University Center for Nondestructive Evaluation Emerging technologies are changing the way that engineers view reliability and system health for purposes of planning and effecting maintenance, repair, and replacement. Today we have an increasing ability to measure critical parameters and gather and process large amounts of data. This, along with advances in scientific modeling of degradation processes, is providing the potential for obtaining better information more quickly for the purposes of making better decisions. Historically, reliability data consisted of time-to-failure information and predictions based on empirical models that are used to estimate lifetime distributions. Estimation is based on the use of data either from field experience or accelerated laboratory life tests, in the case of new materials or components. In the case of high-reliability components, it may be impossible (or undesirable) to observe relevant failures in the field or in the laboratory. In such cases it is sometimes possible to obtain degradation data (actual physical or chemical degradation, performance degradation, or other kinds of degradation surrogates) that are useful for predicting failure, even for individual units. Examples include light output of lasers, vibration in a motor, and chemical change in a coating. An important advantage of degradation data is that they provide a much richer basis for developing chemical and physical models of failure. If there is enough fundamental knowledge about a particular failure mode so that one can develop an adequate model for incremental damage as a function of environmental variables, then it is possible to predict the time to failure for a given environmental profile. One example relates to work done at the National Institute of Standards and Technology (NIST). NIST scientists conducted careful indoor experiments designed to obtain fundamental understanding of the degradation processes of a model epoxy coating as a function of environmental conditions (temperature, humidity, and ultraviolet intensity and spectrum). This information and the experimental results were used to develop a response surface model to predict incremental damage as a function of the environmental conditions at a point in time. The resulting model, when used with actual outdoor environmental data, can be used to predict cumulative damage. For model verification, the predictions were compared with actual observed damage in units exposed to input environment. In general, the task of predicting the failure of individual units offers both challenges and opportunities. The ability to make such predictions hinges on the ability to develop a useful model to predict degradation and/or the ability to sense relevant changes in the state of the system (component or material) of interest. Models for both the physical state of the system and for sensor data will require appropriate stochastic elements in addition to the usual deterministic structure. The detailed characteristics of these models will have to be obtained from extensive experimentation or other data-gathering methods.
Session IV: Materials State Awareness Application Issues 35 An important component of any failure-prediction methodology is the determination of a decision criterion: Which variable or variables (empirical, physical/chemical model output, or a combination) should be used to make a call that a failure is eminent or that there is too high a safety risk? Consider a traditional nondestructive evaluation (NDE) system as an analogy. For a simple scalar criterion (such as the amplitude of a reflection from a crack in an ultrasonic transducer at inspection), modeling the variable and determination of a threshold are relatively straightforward. In the more complicated multizone ultrasonic transducer system, a bivariate response (signal amplitude and signal-to-noise ratio) is used for the decision criterion. Such bivariate data require a more advanced statistical model for the decision criterion. In more complicated situations (e.g., system state and/or environmental monitoring from an array of sensors), data could be highly multivariate, and some means of dimension reduction will almost certainly be needed. Such dimension reductions could be done through a physical and chemical model for incremental damage as a function of environmental conditions or through empirical modeling. The other element of choosing a decision criterion is the determination of a decision threshold (which will have the same dimension as the decision variable[s]). Generally this is done through the quantification and assessment of the trade-off between quantities analogous to probability of detection (POD) and probability of a false alarm (PFA). In continuous monitoring, however, the concepts of POD and PFA are more complicated than they are in periodic inspection. In particular, POD is usually replaced by something like âaverage run length,â giving the amount of time that it will take to detect a subtle change in the system state when data are contaminated with noise. There has been much previous work in the area of statistical process monitoring (also known as statistical process control or change-point detection), particularly methods developed for the chemical process industries. Such methods have the potential to provide useful tools for materials state awareness. A final practical concern for a failure prediction methodology running in real time is the reliability of the sensors themselves. It is easy to imagine how faulty sensors could lead to either false alarms or failure to predict.