Fathy Ismail

Prognosis of machine health condition using neuro-fuzzy systems

A reliable machine fault prognostic system can be used to forecast damage propagation trend in rotary machinery and to provide an alarm before a fault reaches critical levels. Currently, there are several techniques available in the literature for time-series prediction. Among the most promising methods are recurrent neural networks (RNNs) and neuro-fuzzy (NF) systems. In this paper, the performance of these two types of predictors is evaluated using two benchmark data sets. Through comparison it is found that if an NF system is properly trained, it performs better than RNNs in both forecasting accuracy and training efficiency. Accordingly, NF system is adopted to develop an on-line machine fault prognostic system. In order to facilitate the automatic monitoring process, reference function approach is proposed here to enhance feature representation. The performance of the developed prognostic system is evaluated by using three test cases including a worn gear, a chipped gear, and a cracked gear, as well as using data sets from previous studies corresponding to a gear pitting damage and a shaft misalignment. From these tests, the NF prognostic system is found to be a very reliable and robust machine health condition predictor. It can capture the system dynamic behaviour quickly and accurately.

Surface swept by a toroidal cutter during 5-axis machining

This paper presents a method of determining the shape of the surface swept by a tool that follows a 5-axis tool path for machining curved surfaces. The method consists of discretising the tool into pseudo-inserts and identifying imprint points using a modified principle of silhouettes. An imprint point exists for each pseudo-insert and the piecewise linear curve connecting them forms an imprint curve for one tool position. A collection of imprint curves is joined to approximate the swept surface. This method is simple to implement and executes rapidly. The method has been verified by comparing predicted results of a 3-axis tool path with analytical results and predicted results of a 5-axis tool path with measurements of a part made with the same tool path.

Tool path planning for five-axis machining using the principal axis method

A study of the effect of feed direction on five-axis tool paths generated using local surface properties for tool orientation and positioning is presented in this paper. The principal axis method for five-axis machining defines the placement of the cutting tool at a single point on the workpiece surface and assumes that a preferred feed direction will be maintained. This preferred direction may not represent a practical choice for tool path planning. In this work, numerical simulations are used to evaluate tool paths with different feed directions. Numerical simulations are then verified experimentally by machining two example surfaces. The results show that both gouging and the effective cutter profile will dictate the optimal choice of feed direction.

A neuro-fuzzy approach to gear system monitoring

The detection of the onset of damage in gear systems is of great importance to industry. In this paper, a new neuro-fuzzy diagnostic system is developed, whereby the strengths of three robust signal processing techniques are integrated. The adopted techniques are: the continuous wavelet transform (amplitude) and beta kurtosis based on the overall residual signal, and the phase modulation by employing the signal average. Three reference functions are proposed as post-processing techniques to enhance the feature characteristics in a way that increases the accuracy of fault detection. Monitoring indexes are derived to facilitate the automatic diagnoses. A constrained-gradient-reliability algorithm is developed to train the fuzzy membership function parameters and rule weights, while the required fuzzy completeness is retained. The system output is set to different monitoring levels by using an optimization procedure to facilitate the decision-making process. The test results demonstrate that the novel neuro-fuzzy system, because of its adaptability and robustness, significantly improves the diagnostic accuracy. It outperforms other related classifiers, such as those based on fuzzy logic and neuro-fuzzy schemes, which adopt different types of rule weights and employ different training algorithms.

Multi-point tool positioning strategy for 5-axis machining of sculptured surfaces

Multi-point machining (MPM) is a tool positioning technique used for finish machining of sculptured surfaces. In this technique the desired surface is generated at more than one point on the tool. The concept and viability of MPM was developed by the current authors in previous works. However, the method used to generate the multi-point tool positions was slow and labor intensive. The objective of this paper is to present efficient algorithms to generate multi-point tool positions. A basic multi-point algorithm is presented based on some assumptions about the curvature characteristics of the surface underneath the tool. This basic algorithm is adequate for simple surfaces but will fail for more complex surfaces typical of industrial applications. Accordingly, tool position adjustment algorithms are developed that combine the basic algorithm with non-linear optimization to achieve multi-point tool positions on these more complex surfaces.

Rolling ball method for 5-axis surface machining

Curvature matching for 5-axis surface machining has been plagued by the complexity of the task. As a result the current tool positioning strategies are likewise computationally complicated. Gouging the surface has been the main concern and has presented the greatest difficulty in the algorithms. Some of the methods perform exhaustive searches of the surface to avoid gouging while others incrementally adjust the tool orientation until gouges are no longer detected. In this paper a new positioning strategy is presented that is simple to implement and is not difficult to compute. The rolling ball method rolls a variable radius ball along the tool path and positions the cutting tool to cut the rolling ball. A small region of the balls surface is used to approximate a small region of the surface being machined. The radius of each ball is computed by checking a grid of points in the area of the surface that the tool casts a shadow for each tool position. A pseudo-radius is computed for each grid point and the most appropriate radius is selected to be the rolling balls radius. The selection process follows a hierarchy of surface profiles ranging from convex to concave. Convex, concave, and saddle (mixed) surface regions are all computed in a similar fashion and there are no special cases for which the positioning strategy must be changed to compute a tool position. Local gouge checking is automatically built-in to the positioning computations so that the typical iterative strategy of checking for gouging, then incrementally tilting the tool until no gouges are detected is eliminated. The method is robust and simple to implement and it only requires surface coordinates and surface normals. A simulation of the method and a cutting test were performed and are presented in this document.

Arc-intersect method for 5-axis tool positioning

A new method for 5-axis CNC tool positioning is presented in this paper that improves upon a previous tool positioning strategy named the rolling ball method (RBM), which was developed by the present authors [Gray P, Bedi F, Ismail S. Rolling ball method for 5-axis surface machining. Comput Aided Des 2003;35(4):347-57]. The special property of the RBM is that it computes tool positions by considering the area beneath the tool that the tool will be positioned to cut instead of using surface curvatures computed at a single point on the surface. This enables the RBM to generate gouge-free tool positions without secondary iterative gouge-check and correction algorithms. However, the RBM generates conservative tilt angles in order to guarantee gouge-free tool positions. The new arc-intersect method (AIM) presented in this paper improves upon the RBM by directly positioning the tool to contact the surface and thereby eliminates the conservative nature of the RBM to give optimal tool positions. Like the RBM, the AIM is an area-based method that generates gouge-free tool positions without the use of iterative gouge-check and correction algorithms. The implementation described in this paper uses triangulated surfaces and the computers graphics hardware to assist in the tool position calculations. However, the method can be applied to any surface representation since it only uses surface coordinates and surface normals for computation. A section of a stamping die was machined to demonstrate the AIM and to show the improvement over the RBM and for comparison with 3-axis ballnose machining. The results showed that the AIM was 1.33 times faster than the RBM and that the AIM, with single direction parallel tool passes, was 1.62 times faster than a zig-zag pattern 3-axis ballnose tool path for the same feed rate, cusp height and tool diameter. The workpieces were measured with a CMM and the data were compared to the CAD model to show no gouging occurred and to check the cusp heights.

Comparison between multi-point and other 5-axis tool positioning strategies

Abstract A method of generating sculptured surfaces at multiple points of contact between the tool and the workpiece was developed and proven viable by the current authors in previous work. They denoted this finish machining method, “Multi Point Machining”, or simply MPM. This paper compares MPM with two other 5-axis tool positioning strategies; namely: the inclined tool, and the principal axis method. It is also compared with 3-axis ball nose machining. Comparisons are conducted using computer simulations and experimental cutting tests. Results obtained show that MPM produced scallop heights that are much smaller than those produced by the other tool positioning strategies.

Graphics-assisted Rolling Ball Method for 5-axis surface machining

In this paper, a graphics hardware-assisted approach to 5-axis surface machining is presented that builds upon a tool positioning strategy named the Rolling Ball Method presented in an earlier paper by the present authors [Comput. Aided Des. 35 (2003) 347]. The depth buffer of the computer’s graphics card is used to compute the data needed for the Rolling Ball Method, which generates gouge-free 5-axis curvaturematched tool positions. With this approach, the tool path for a workpiece can be computed with triangulated data instead of parametric surface equations. It also permits the generation of tool paths for multiple surface patch workpieces that have only positional continuity. The method is easy to implement and it is robust since every tool position is computed with the same algorithm regardless of the type of surface. For illustration, tool paths were generated for a workpiece with two bi-cubic surface patches, connected with only position continuity. Simulations for gouge-checking and machining tests were performed. Workpiece cusp heights were measured using a coordinate measuring machine. The maximum undercutting measured in the machining examples was 0.07 and 0.05 mm, which is within the expected NC machine accuracy and measuring capabilities for surfaces. q 2003 Elsevier Ltd. All rights reserved.

Chatter suppression by adaptive speed modulation

In this paper a PD fuzzy logic controller is designed and implemented to suppress chatter in peripheral milling. The controller selects combinations of the amplitude and frequency of spindle speed modulation, online, to keep a chatter indicator, the R-value, close to a prescribed set point. The controller was successful in suppressing chatter at certain operating points. The effect of the control action delay time and the degree of process instability on the controller performance were addressed through simulations and experiments. At some operating points the controller resulted in an oscillatory behaviour of the process and the R-value fluctuated significantly. Stability lobes were used to explain this phenomenon. Strategies for enhancing the performance of the controller were proposed.