When applied finite element method to numerical simulation, the discretization of the domain is essential part of the procedure. There are few reports about the quadrilateral mesh generation method in comparison with triangular mesh generation. The characteristic of the proposed algorithm is that the quadrilateral of the inner nodes is carried out by bucket method in computational geometry. And then the triangulation of the residual region is carried out utilizing advansing front method. Therefore the mixed elements of triangular and quadrilateral elements are generated in this step. In the next step, a quadrilateral element is constructed connecting two triangular elements with a common side. Finally, the quadrilateral elements are generated by the subdivision of the mixed elements. The processing scheme is developed and applied to several examples.

View full abstract

For large-scale forging simulation by rigid plastic finite element method, it is important to reduce or accelerate the calculation time which is mostly occupied with the solver part of stiffness equation. In this paper, we introduce the parallelization tool of PVM (Parallel Virtual Machine) to improve the solver part into our cluster system of workstations. We employ TPDDI network as fast communication path between workstations, so PVM provides two excellent functions of Multicasting and Direct-linking which enable the faster inter communication about 50% less consumption of calculation time in comparison to the case of standard procedures on PVM as if the case of traditional socket sequence. From the calculation test for forging simulations, our parallelized solver accelerates the calculation of single workstation about 5.4 times on the system developed by using two PVM's functions above. Then, we conclude that PVM provides a new convenient means to high speed execution of large-scale numerical simulations.

View full abstract

A new analysis method for bending problems of beam by the boundary element method has been developed. Main points of this study are (1) to improve the composition of the simultaneous equations by introducing a new formulation process, (2) to establish a scheme without any variables at intermediate points and (3) to establish a generalized solution scheme for an inhomogeneous beam. This report will describe (1) and (2). Size of matrix as well as the computing time are greatly reduced owing to these new algorithm, and therefore, high efficiency on the repetitive calculations is brought about. Using these algorism, it will become much more efficient to build an optimal design system by such as the genetic algorithm and so on.

View full abstract

A new analysis method for bending problems of beam by the boundary element method has been developed. This report will describe how to solve the problems of an inhomogeneous beam. A generalized solution scheme for an inhomogeneous beam is established owing to the new strategies developed in the 1st report, which are to improve the composition of the simultaneous equations and to establish a scheme with the non-dividing solution scheme. Using the present solution scheme, it will become easier and more efficient to construct the search problems for optimal shape using such as the genetic algorithm and so on.

View full abstract

The Japanese Experiment Module (JEM) is the Japanese contribution to the International Space Station (ISS). The ISS program has established a strategy for managing the risk associated with the micrometeoroid and orbital debris (MM-OD) threat. A shield configuration was designed for the JEM Pressurized Module. Hypervelocity impact (HVI) tests and hydrocode simulations have been conducted to verify the JEM Pressurized Module shield and structural performance for MM-OD impacts. In this paper, we first report the latest hydrocode simulation results for JEM debris shield design evaluation against the strikes of debris. Next, we report the results of hydrocode simulation for large size (5 and 10 cm in diameter) debris impacts on the JEM shield and pressure structure which were conducted to evaluate the penetration hole size in the module wall.

View full abstract

Optical tomography is a new modality of non invasive measurement of physiological information, and some algorithms using the technique of the ill-posed inverse problem have been developed to obtain images. These algorithms, however, are still at the early stage and significant improvement is expected to obtain better images toward practical and clinical applications. This paper describes the rationale of a novel image reconstruction method that estimates the nullspace contribution using prior knowledge in a maximum-a posteriori-probability (MAP) framework. We illustrate and verify our concept for two cases of optical tomographic reconstruction ; one using simulated and the other using experimental data. The improvement by using our nullspace-MAP algorithm over the conventional Newton-Raphson scheme is found to be significant for the two examples.

View full abstract

A simplified method of estimating welding residual stress using modified inherent strain analysis was develoed. And residual stress in a box structure fabricated by plural welded joints was evaluated using this method. Inherent strain distribution in the box structure was inferred from that of simple welded joints by using a relation between the inherent strain distribution and the welding conditions of welded joints. Residual stress in the box structure was elastically computed with actual welding sequences. The analytical results of residual stress agree well with strain-gauge measurements from a mock-up specimen. It is shown that the developed method of elastic analysis can simply compute welding residual stress in complicated welded structures with plural welded joints.

View full abstract

In order o evaluate the interface properties of Glass Fiber Reinforced Polypropylene in the different adhesive condition, a surface of polypropylene stampable sheet is modified by O₂ plasma treatment method. Deformation and fracture properties are investigated by the static tensile test, and the stress-strain relation is regressed by 3-elements visco-elastic model to estimate the interface properties. The interface estimation is expressed following two parameter ; γ : composites'stress transferability to reinforcements through the interface (=fiber's strain/composites'strain), and a : strain intensity in matrix for composites'strain (=matrix's strain/composites'strain-1). The tensile strength of the treated composites is increasing about 40% with the 40% increment of the interlaminar strength by the plasma treatment. The regression result says the fiber reinforced effect is promoted with the interface adhesion (γ=60% to 85%), and this makes the ultimate strength be increasing. The value of a=45% in non-treated composites (in treated one, a=15%) shows the partial strain presents with a stress concentration around adhesive failures. Fractgraphies show the alternation of γ and a, so that the critical fracture is due to this partial plastic strain in non-treated one, hence, due to the interlaminar fracture because of the higher stress transferability in treated one.

View full abstract

A constitutive model that describes the macro-stiffness characteristics for fiber reinforced composite materials is introduced. The model utilizes the Eshelby's equivalent inclusion method for describing a particle but with authors'self consistent compliance scheme instead of Mori-Tanaka's meanfield theory, which most other composite models are based on. A separated meso-mechanics analysis verified the accuracy and the reliability of the present model for a fiber dispersed composite over the existing model. A numerical analysis with the developed model revealed that fiber orientation, aspect ratio of fiber, rigidity ratio of fiber/matrix, and volume fraction of fiber are the major controlling factors of the fiber reinforced composite, thus can be regarded to take the role of design parameters of the composite structures.

View full abstract

It is well known that the stress state in 2-dimensional dissimilar materials depends on mutually independent two elastic constants (parameters) and the problem can be put in order effectively by two parameters called Dundurs'ones. In this paper, the dependence of stress state on elastic constants in axisymmetric dissimilar materials is studied by using Michell's stress function and it is made clear first that, different from 2-dimensional problem, the stress state for axisymmetric case depends on three mutually independent parameters. Subsequently, which combination of three parameters is useful to deal with the problem is discussed and a combination of three parameters is presented as parameters for axisymmetric problems that play the role corresponding to Dundurs'parameters for 2-dimensional problem. Moreover, finite element analyses are carried out and the usefulness of the parameters to study the effect of material combination on stress state is demonstrated.

View full abstract

In this study, an analytical method of development of thermoelastic problems for a medium with nonhomogenous material properties is developed. For the isothermal problems of such a nonhomogeneous body, an analytical method of development has already given by Kassir under the assumption that the shear modulus of elasticity G changes arbitrarily with the variable z of the axial coordinate according to the relation G (z)=G₀z^m. However, no analytical procedure for the thermoelastic field has been established up to present. In this study, introducing the thermoelastic displacement potential function and two kinds of displacement functions, an analytical method of development of the three-dimensional thermoelastic field is proposed. Assuming that G, thermal conductivity λ and coefficient of linear thermal expansion a vary with the variable ζ connected with the dimensionless axial coordinate according to the relations G (ζ)=G₀ζ^m, λ(ζ)=λ₀ζ^l, α(ζ)=α₀ζ^k, the three-dimensional temperature solution in a steady state for a slab is obtained, and the associated thermal stress components are evaluated theoretically. Numerical calculations are carried out for several cases taking into account the variation of nonhomogeneous material properties, and the numerical results are shown graphically.

View full abstract

A residual stress state was produced in a circular plate of steel by welding a small hole in the middle of the plate, and nondestructive determination of this stress distribution was studied by the acoustoelastic spectrum method. The amplitude spectra of reflected transverse waves were first analyzed to obtain the polarization directions and the velocity difference of two transverse waves, and it was noticed that, in the resulting distribution of the velocity difference, apparent peaks existed near the elastic-plastic boundary. The results of the polarization directions and velocity difference were then applied to the acoustoelastic law, and the stress distribution in the elastic region was determined, either employing the shear difference method or assuming axial symmetry of the distribution. These results were shown to agree well with those of destructive measurements by the strain gauge method, and the peak locations in the velocity difference distribution were also confirmed to be very close to the elastic-plastic boundary. From this, it was concluded that the present method was very useful for nondestructive stress measurements.

View full abstract

This paper deals with an interaction problem of a row of ellipsoidal inclusions under asymmetric uniaxial tension using singular integral equations of the body force method. The problem is solved on the superposition of two auxiliary loads ; (i) biaxial tension and (ii) plane state of pure shear. These problems are formulated as a system of singular integral equations with Cauchy-type or logarithmic-type singularities, where unknown functions are densities of body forces distributed in the γ, θ, z directions in infinite bodies having the same elastic constants as those of the matrix and inclusions. In order to satisfy the boundary conditions along the ellipsoidal boundaries, the unknown functions are approximated by a linear combination of fundamental density functions and polynominals. The present method is found to yield rapidly converging numerical results for stress distributions along the boundaries. For any fixed shape and spacing of inclusions, the maximum stress is shown to be linear with the reciprocal of the squared number of inclusions.

View full abstract

Electromagnetic acoustic transducers (EMATs) are made for exciting and receiving longitudinal wave in ferromagnetic steel wires. The EMAT consists of a coil wound around the wire and an electromagnet to apply a static magnetic field in the axial direction. Ferromagnetic materials show a dimensional change, called magnetostriction, when an external magnetic field is applied. If the dynamic magnetic field is induced by the coil, the resultant dynamic magnetostriction launches the longitudinal wave, which propagates along the wire. In this study, we investigated the generation and detection mechanisms of the EMATs by studying the amplitude response to the bias field and pulse intensity. The results are compared with the magnetostriction curve, ensuring that the magnetostriction is the dominant mechanism for this type of EMAT. We find that the optimum static magnetic field is different for transmitter and receiver. Thus, the use of the dual-EMAT configuration is recommended to improve the signal to noise ratio for detecting flaws in wires.

View full abstract

Optimization of excitation signal is proposed for inspection of a thin steel wire by longitudinal wave. In flaw detection of the wire throughout a longe range, the longitudinal wave, which propagates along the length, will be a useful tool. The short pulse is suitable for achieving high spatial resolution of location of the flaw. However, the velocity of longitudinal wave in the wire shows the frequency dependence. This dispersion distorts the waveform, and it results in the decrease of the signal amplitude. If the effect of dispersion is removed, the signal to noise ratio is expected to be improved and the high flaw detectability can be obtained. In this study, we suggest to use the optimum excitation signal, which is so modified to compensate for the dispersion during the propagation of assumed path length. It is ensured that the flaw, which occupies 10% of the cross-sectional area of the wire, can easily be detected from 10 m away.

View full abstract

Simple numerical model is developed to simulate magnetoelastic interactions in polycrystalline steel. Elastic properties and magnetic properties of ferromagnetic steel are related to each other, both of which depend on the magnetic domain structure. In the demagnetized state, six types of domains, each of which is magnetized parallel to one of the crystallographic axes, are randomly distributed in individual grain. The magnetic field increases the volume of the domains magnetized parallel to the field at the sacrifice of the other domains. When the domain realignment is almost completed, the domain magnetization starts to rotate. The stress also expands the domains parallel to the tension or perpendicular to the compression. In this study, we propose the simulation model which considers the volume change of the domains and the rotation magnetization, both induced by the magnetic field and the stress. The stress dependences of permeability and magnetostriction are predicted. Comparison of the results with the experiments ensures that the model is useful in explaining the experimental results.

View full abstract

In this paper, an inverse analysis method to identify the stress-strain relationships of elastic-plastic materials, using a nonlinear sensitivity analysis, is presented. The identification problem can be defined as a minimization problem, by choosing an appropriate objective function, which consists of displacement and force quantities. The nonlinear sensitivity analysis, which does not require repetitive finite element calculations, forms the basis of the proposed procedure. An algorithm of Nonlinear Programming, coupled with nonlinear sensitivity analysis, results in an optimization process of classical sense. In this paper, the framework of the identification process is pressented and its validity is demonstrated through example problems. Furthermore, the framework of the proposed method, is quite general so that the methodology can be applied to various types of engineering materials.

View full abstract

In the present study, the plastic singularity for problems of a crack meeting and normal to an interface and of a corner of bonded two 90°-wedges. A practical method is introduced to solve homogeneous differential equations obtained in the analysis. Except of the hardening exponent n, the relative resistance of two materials to plastic deformation can be represented by a parameter Δ defind as Δ=(α₁/α₂) (E₂/E₁) (σ₀₂/σ₀₁)^n-1. For cases of two bonded materials having the same n, it is found that the singularity values s for the problem of a crack meeting and normal to an interface are restricted in a range of 2>s>s^*, where s^* are values for a single 90°-wedge with free-clamped edges, and the values s for the problem of a corner of bonded two 90°-wedges are restricted in a range of s^*>s>2-1/n. For cases of two bonded materials having the different n. it is found that the order of singularity depends only on the hardening exponent n of the material with smaller n.

View full abstract

The case study by nonlinear FEM analysis for impact compression fatigue damage in polypropylene type elastomer damper was carried out to construct the shape optimum mech.-design of impact absorbing elastomer. Impact compression fatigue test was carried out to investigate the characteristics and damage of erastomer damper. And nonlinear FEM analysis by popular CAE was carried out to examine the distribution of stress and strain of dampers. Main results were as follows ; (1) Examination of the sectional view of tested damper revealed the presence of spherical discoloration damage parts whose diameter was about 1.5 mm. (2) From fractography with using scanning electron microscopy (SEM), it was understood that the form inside the discoloration damage part was granular. It was suggested that one of this cause was the melting of erastomer by temperature rise which depends on strain energy changing into heat energy. (3) At Load-Displacement curve, close agreement between calculated value by nonlinear FEM analysis and compression load test results value was obtained. (4) The position of damage parts within the damper by impact compression fatigue test was almost the same as that of a highest strain part which was calculated by nonlinear FEM analysis.

View full abstract

Rotating bending fatigue tests were carried out on annealed and quenched-and-tempered 0.45% carbon steel specimens with single or double extremely shallow partial notch. The fatigue strength of these specimens was discussed based on the √(area) parameter model and the concept of linear notch mechanics (LNM). The main results are summarized as follows : (1) The fatigue limits of specimens with the same √(area) are dependent on the value of notch root radius ρ when p is large. (2) When ρ is large, the estimation of LNM is reasonable. On the other hand, when ρ is small, the estimation of LNM in the original form is not reasonable. However, the error is less than about 15% and the permissible level in general use. The reason why this error occurs is that the relative elastic stress distribution is dependent on the notch depth t when ρ is small. (3) The fatigue strength of the specimens with double extremely shallow partial notches can be predicted by the concept of LNM through the results of specimen with a single notch. The reason why LNM is reasonable is that the relative elastic stress distributions of the single notch and double notch are almost the same near the notch root even in the cases of extremely shallow notch under the condition of the same ρ and the same t.

View full abstract

To study fatigue damage on low cycle fatigue, tensile tests after fatigue were carried out using cylindrical specimens, and the relationship between the change of residual fracture ductility and the cycle ratio was investigated. The main results obtained are as follows : (1) Until the last stage of fatigue life (N/N_f=0.9 ; crack length 1.5 mm), the residual fracture ductility ε_FR did not decreased remarkably and the tensile fracture mode did not vary from cup & cone type. These results differ from previous studies using hourglass specimens. (2) The exhaustion of ε_FR is caused by two factors : One is the internal damage due to cyclic work hardening and the other is the surface damage due to small cracks. The degree of these damages is bath small. (3) Most of the fatigue damage on low cycle fatigue depends on the length of surface cracks.

View full abstract

Pulsating low cycle fatigue tests of Ti-6Al-4 V alloy which is thought to be a candidate of structural material at cryogenic temperature have been conducted in a stress controlled condition at 77 K and room temperature. Attention was paid to the S-N relation using normalized maximam stress and changes in cumulative plastic strain and plastic strain range during fatigue. The normalized maximum stress was found to be a key parameter by which the difference of test condition, such as temperature, or the deviation of material property can be evaluated in the same S-N relation. The fracture type of this alloy was found to be divided into quasi-static fracture and fatigue fracture respectivily depending on the magnitude of applied load. This was verified by fracture surface observations. It was found that the resistance to failure in the criteria of both stress and strain at 77 K is higher than that at room temperature.

View full abstract

Difference in fatigue strength depending on rotational speed between in air and in water at the rotating-bending fatigue characteristics of a TiNi Shape-memory alloy wire was found in our previous work. The heat transfer characteristics around a shape-memory alloy wire was measured using Reynold's analogy. The temperature rise of metal during the rotating-bending cycles were calculated considering heat transfer result. As a result, the rotating-bending fatigue characteristics of a shape-memory alloy wire were well explained from the viewpoint of heat transfer characteristics.

View full abstract

Short carbon fiber reinforced thermo plastics (SCFRTP) fabricated with injection molding generally shows a large amount of scatter in fatigue strength. To find appropriate method of treating this scatter, fatigue tests were carried out. As the results, the scatter in fatigue strengh was estimated in terms of the initial volumetric Young's modulus, which involves the effect of the fiber morphologies such as length distribution and orientation. Parallel to this, the heat generation during fatigue testing was measured on each specimen with different fiber morphologies. The results showed that monitoring terperature rise in specimen is a suitable way to evaluate the fatigue strength of SCFRTP. The validity of this method was demonstrated by fatigue tests with variable stress amplitude ad frequency.

View full abstract

Dynamic thermal stresses and cyclic crack growth behavior produced by a single and repeated thermal shocks were studied on silicon nitride. Thermal shock was applied to a square shaped bar specimen using the improved water-quenching method. It was clarified that the cyclic crack growth rates, da/dN, under the repeated thermal shock tests are represented by the following expression, da/dN=9.35×10^-6(K_max-1.36)^3.14. In addition, the experimentally observed maximum thermal stresses, σ_max, for silicon nitride in the present study as well as those for cemented carbides and cermets in the previous report can be approximated by the expression, σ_max=0.704αEΔT-57.74, where α, E, ΔT denote a linear coefficient of expansion, Young's modulus of the materials and applied temperature differences at thermal shocks, respectively. By using this expression we can directly evaluate the maximum thermal stresses induced by the thermal shocks without temperature measurements.

View full abstract

Reflection type high-speed holographic microscopy is applied to photograph the bifurcation of fast propagating cracks in PMMA. The cracks are of the opening mode. Spatial resolution is more than 180 lines/mm, frame interval is 5μs and the number of frames is three. A crack at bifurcation is recorded as three successive holograms on one photographic plate by means of angle-multiplexing holography. Lens assisted reconstruction method is used to reconstruct separately the three crack images from the holograms. The reconstructed crack images are magnified and photographed through a conventional microscope. The photographs show that, at bifurcation, several crack tips are running simultaneously in a small region about 1 mm in diameter. And also shown is dynamic three dimensional behavior of crack front edges or microcracks at bifurcation.

View full abstract

Branched cracks are often observed in brittle materials and structures. However the mechanism of crack branching is not fully clarified. In this paper, first, dynamic crack branching phenomena in Homalite-911 and Homalite-100 were experimentally investigated. Caustic patterns during dynamic crack branching were recorded by an ultrahigh-speed camera. Next, on the basis of the dynamic J integral, energy concepts for dynamic crack branching are presented. Special attentions are focused on the energy flow into dynamically propagating crack tips. Then four possible fracture parameters associated with the dynamic J integral are proposed. The dynamic J integral values in a dynamic crack branching test are directly evaluated by the caustic patterns taken by the high-speed camera. The variations of four fracture parameters during dynamic crack propagation including at the instant of crack branching are presented. The experimental results reveal that the energy flux per a unit time into a propagating crack tip governs the crack branching.

View full abstract

Crack healing behaviours of silicon nitride have been investigated as a function of sintering additives and crack healing condition, systematically. The semicircular crack (average diameter is 110 μm) was made in the center of specimen using Vickers hardness indenter. Some of the cracked specimens were heat-treated at 800°C∼1 400°C, 1∼150 hour in air atmosphere. Bending strength was measured at room and elevated temperature up to 1 400°C. Bending strength of cracked specimen reduced to about 48% of that of smooth specimen. However, the bending strength of pre-cracked specimens recovered completely up to that of smooth specimen by the heat treatment. And in several specimens, failure occured from the outside of pre-cracked area even at 1 300°C. This strength recovery was caused by crack healing. By analysing the test results, the following conclusion was obtained. The silicon nitride which has the highest crack healing ability within this experiment is silicon nitride with 20 wt.% SiC and 8 wt.% Y₂O₃. The best healing condition for the specimen is 1 200°C∼1 300°C, 1 h in air. The crack healed member at the condition has enough bending strength and low sensitivity to static fatigue up to 1 400°C.

View full abstract

Deformation-induced martensite and reversed austenite of metastable austenitic stainless steel sheet were evaluated by a scanning acoustic microscope with frequencies 600 MHz and 800 MHz. The sheet was elongated up to 40% below the room temperature, followed by annealing. First martensite content was measured by a feritscope. Using a complex V(z) curve, leaky Rayleigh wave velocity was measured. The deformed and annealed textures were observed with the frequency 800 MHz and compared with those by the optical microscope. Rayleigh wave velocity is dependent on the elongation and ambient temperature in elongation, and the annealing temperature, which agrees well with the one by the feritscope. Deformed grains are more clearly observed by the scanning acoustic microscope.

View full abstract

In order to develop a creep life assessment technique for a directionally solidified Ni-base superalloy CM247LC used in advanced gas turbine blades, changes in electrochemical properties of CM247LC due to creep have been investigated. The electrochemical polarization technique has been applied to creep-ruptured and creep-interrupted specimens under various creep conditions in order to examine a detectability of creep damage. Experimental results on electrochemical polarization measurements revealed that the peak current density"I_p"and"I_pr"which appeared at a specific potential during potentiodynamic polarization reactivation measurements in dilute glyceregia solution linearly increased with a life fraction in early stage of the creep life (t/t_r=0.3). The increases in I_p and I_pr mainly corresponded to increases in preferential dissolution rate of γ' phase at γ/γ' interface. This preferential dissolution was considered as a result from change in chemical composition of γ'phase at γ/γ'interface which may be attributable to creep damage accumulation. As a consequence, the creep life fraction can be estimated with high accuracy by the electrochemical technique.

View full abstract

A multiaxial constitutive model to describe the history-dependent creep deformation and damage of polycrystalline metallic materials under variable loading conditions has been developed. A particular emphasis is placed on a novel modeling of the history-dependent material degradation coupled with creep hardening mechanisms. First, the history dependence of internal variables involved in the kinematic-hardening model is identified. Then, continuum damage variables coupled with different history-dependent hardening variables are newly defined, and the total damage is expressed as the sum of these damage variables. This inelastic-history dependent damage model is incorporated into the modified kinematic hardening model which has been developed to predict the complicated anisotropic hardening/softening behavior under repeated multiaxial stress changes. Finally, the characteristic of the proposed model with respect to the history-dependent growth of damage is elucidated through numerical analyses for different material parameters. The present damage-coupled modified kinematic hardening model can describe the accelerated or decelerated accumulation of creep damage depending on stress variations, together with the transient material-softening behavior.

View full abstract

Structural optimization using the homogenization method which was introduced by Bendsφe and Kikuchi in 1988 is established as a theory of optimal design of layout (topology and shape). This theory has been applied to the various kind of problems, especially in meso-micro-scale desing, for example in the microstructure of composite and piezo-ceramic material. Frecker and Nishiwaki studied the possibility of the extension of the homogenization design method for topology/shape optimization for elastic structures into the design of flexible structures such as compliant mechanisms. Compliant mechanisms is a relatively new breed of jointless mechanism in which elastic deformation is intended to be a source of motion. In the past, most flexible structural design optimization was considered only for the direction of the flexible point, by using the material density or equivalent homogenization method to identify the optimum layout. Here we shall extend it to the case of a two dimensional design for compliant mechanism which control the deflection of the flexible point. Moreover we also applied the image based design method to extract a FE model for prototype from the optimal layout and propose the integrated structural design methodology.

View full abstract

When a thin cylindrical shell immersed in fluid is subjected to sinusoidal wave excitations in the horizontal direction, subharmonic vibration accompanying shell mode vibration may occur depending on the excitation condition. This phenomenon was investigated by dynamic stability analysis and was proved to be dependent both on frequency and acceleration amplitude of excitation. Calculation results under sinusoidal wave excitation and comparison of dynamic pressure when subharmonic vibration occurs with elastic buckling pressure provide a criterion for assessing stability of subharmonic vibration under seismic wave excitation.

View full abstract

In order to detect corrosion damage like fatigue short cracks and corrosion pits and to estimate remaining life for pump shaft, an automatic inspection system by image processing was developed. The system mainly consists of a mechanical scanner for the pump shaft, an optical microscope, a CCD camera, an image processing unit, 8 mm video recorder and data processing computer. The pump shaft surface was initially ground and polished with ceramic and rubber grinders. After polishing the shaft surface was observed with optical microscope. The magnitude of observation ranges from 50 to 200 depending on the corrosion damage size. The observed surface structure was analyzed as the brightness distribution by the binary code processing. The performance of the developed system was verified with artificial fatigue cracks and actual corrosion pits. The fatigue short cracks could be detected with accuracies of 0.1 mm for length and 3 degrees for angle, and the corrosion pits could be detected with accuracy of 10μm. The system has been applied to more than 250 actual pump shaft for 9 years, and no pump shaft failure has been experienced.

View full abstract