The rising demand for exacting performances from manufacturing systems has led to new challenges for the development of complex tool condition monitoring techniques. Although a wide range of monitoring methods have been investigated and developed, there has been very little migration of these innovations into industrial practice. The principal factor behind this phenomenon is the stochastic nature of the environment in which the system must function. A truly universal application has yet to be developed. The work presented here centres around the application of an unsupervised neural network model to the said problem. These networks learn without the aid of a human teacher or supervisor and learn to organise and re-organise themselves in accordance to the input data. This leads to the network structure reflecting the given input distribution more precisely than a predefined model, which generally follows a decay schedule. The dynamic nature of the process provides an evaluation of the underlying connectivity and topology in the original data space. This makes the network far more capable of capturing details in the target space. These networks have been successfully used in speech recognition applications and various pattern recognition tasks involving very noisy signals. Work is in progress on their application to robotics, process control and telecommunications. The procedure followed here has been to conduct experimental drilling trials using solid carbide drills on a Duplex Stainless Steel workpiece. Duplex Stainless Steel was chosen as a preferred metal for drilling experiments because of this high strength, good resistance to corrosion, low thermal expansion and good fatigue resistance. During the drilling trials, forces on the workpiece along the x, y and z axes were captured in real time and moments of the forces were calculated using these values. These three axial forces, along with their power spectral densities and moments were used as input parameters to the Artificial Neural Network model which followed the Self-Organising Map algorithm to classify this data. After the network was able to adapt itself to classify this real world data, the generated model was tested against a different set of data values captured during the drilling trials. The network was able to correctly identify a worn out drill from a new drill from this previously unseen set of data. This autonomous classification of the drill wear state by the neural network is a step towards creating a “universal” application that will eventually be able to predict tool wear in any machining operation without prior training.

The rising demand for exacting performances from manufacturing systems has led to new challenges for the development of complex tool condition monitoring techniques. Although a wide range of monitoring methods have been investigated and developed, there has been very little migration of these innovations into industrial practice. The principal factor behind this phenomenon is the stochastic nature of the environment in which the system must function. A truly universal application has yet to be developed. The work presented here centres around the application of an unsupervised neural network model to the said problem. These networks learn without the aid of a human teacher or supervisor and learn to organise and re-organise themselves in accordance to the input data. This leads to the network structure reflecting the given input distribution more precisely than a predefined model, which generally follows a decay schedule. The dynamic nature of the process provides an evaluation of the underlying connectivity and topology in the original data space. This makes the network far more capable of capturing details in the target space. These networks have been successfully used in speech recognition applications and various pattern recognition tasks involving very noisy signals. Work is in progress on their application to robotics, process control and telecommunications. The procedure followed here has been to conduct experimental drilling trials using solid carbide drills on a Duplex Stainless Steel workpiece. Duplex Stainless Steel was chosen as a preferred metal for drilling experiments because of this high strength, good resistance to corrosion, low thermal expansion and good fatigue resistance. During the drilling trials, forces on the workpiece along the x, y and z axes were captured in real time and moments of the forces were calculated using these values. These three axial forces, along with their power spectral densities and moments were used as input parameters to the Artificial Neural Network model which followed the Self-Organising Map algorithm to classify this data. After the network was able to adapt itself to classify this real world data, the generated model was tested against a different set of data values captured during the drilling trials. The network was able to correctly identify a worn out drill from a new drill from this previously unseen set of data. This autonomous classification of the drill wear state by the neural network is a step towards creating a “universal” application that will eventually be able to predict tool wear in any machining operation without prior training.

The rising demand for exacting performances from manufacturing systems has led to new challenges for the development of complex tool condition monitoring techniques. Although a wide range of monitoring methods have been investigated and developed, there has been very little migration of these innovations into industrial practice. The principal factor behind this phenomenon is the stochastic nature of the environment in which the system must function. A truly universal application has yet to be developed. The work presented here centres around the application of an unsupervised neural network model to the said problem. These networks learn without the aid of a human teacher or supervisor and learn to organise and re-organise themselves in accordance to the input data. This leads to the network structure reflecting the given input distribution more precisely than a predefined model, which generally follows a decay schedule. The dynamic nature of the process provides an evaluation of the underlying connectivity and topology in the original data space. This makes the network far more capable of capturing details in the target space. These networks have been successfully used in speech recognition applications and various pattern recognition tasks involving very noisy signals. Work is in progress on their application to robotics, process control and telecommunications. The procedure followed here has been to conduct experimental drilling trials using solid carbide drills on a Duplex Stainless Steel workpiece. Duplex Stainless Steel was chosen as a preferred metal for drilling experiments because of this high strength, good resistance to corrosion, low thermal expansion and good fatigue resistance. During the drilling trials, forces on the workpiece along the x, y and z axes were captured in real time and moments of the forces were calculated using these values. These three axial forces, along with their power spectral densities and moments were used as input parameters to the Artificial Neural Network model which followed the Self-Organising Map algorithm to classify this data. After the network was able to adapt itself to classify this real world data, the generated model was tested against a different set of data values captured during the drilling trials. The network was able to correctly identify a worn out drill from a new drill from this previously unseen set of data. This autonomous classification of the drill wear state by the neural network is a step towards creating a “universal” application that will eventually be able to predict tool wear in any machining operation without prior training.

Magister Chirurgiae Dentium / Dental implants are artificial fixtures that are surgically inserted into the jaws to replace missing teeth. The success of dental implant treatment is dependent on achieving successful osseointegration (Branemark et al. 2001). Drills used for implant site preparation are made of different materials such as stainless steel (SS), zirconia and ceramic. Most of them do not have sufficient cutting efficiency and wear resistance (Oliveira et al. 2012). Recently diamond-like carbon coating (DLC) has been added as a drill coating to increase the cutting efficiency, increase wear resistance and drill hardness (Batista Mends et al. 2014).

Most drill wear monitoring research found in the literature is based on conventional vibration technologies. However, these conventional approaches still have not attracted real interest from manufacturers for multiples of reasons: some of these techniques are not practical and use complicated Tool Condition Monitoring (TCM) systems with less value in industry. In addition, they are also prone to give spurious drill deterioration warnings in industrial environments. Therefore, drills are normally replaced at estimated preset intervals, sometimes long before they are worn or by expertise judgment. Two of the great problems in the implementation of these systems in drilling are: the poor signal-to-noise ratio and the lack of system-made sensors for drilling, as is prevalent in machining operations with straight edge cutters. In order to overcome the noise problems, many researchers recommend advanced and sophisticated signal processing while the work of Rehorn et al. (2005) advises the following possibilities to deal with the lack of commercial system-made sensors:  Some research should be directed towards developing some form of instrumented tool for drill operations.  Since the use of custom-made sensors is being ignored in drilling operations, effort should be focused on intelligent or innovative use of available sensor technology. It is expected that the latter could minimize implementation problems and allows an optimal drill utilization rate by means of modern and smart sensors. In addition to the accelerometer sensor commonly used in conventional methods, this work has considered two other sensor-based methods to monitor the drill wear indirectly. These methods entail the use of an instrumented drill with strain gauges to measure the torque and the use of an encoder to measure the Instantaneous Angular Speed (IAS). The signals from these sensors were analyzed using signal processing techniques such as, statistical parameters, Fast Fourier Transform (FFT), and a ii preliminary Time-Frequency (TF) analysis. A preliminary investigation has revealed that the use of a Regression Analysis (RA) based on a higher order polynomial function can very well follow and give prognosis of the development of the monitored parameters. The experimental investigation has revealed that all the above monitoring systems are sensitive to the deterioration of the drill condition. This work is however particularly concerned with the use of IAS on the spindle of the drill, compared to conventional monitoring systems for drill condition monitoring. This comparison reveals that the IAS approach can generate diagnostic information similar to vibration and torque measurements, without some of the instrumentation complications. This similitude seems to be logical, as it is well known that the increase of friction between the drill and workpiece due to wear increase the torque and consequently it should reduce or at least affect the spindle rotational speed. However, the use of a drill instrumented with a strain gauge is not practical, because of the inconvenience it causes on production machines. By contrast, the IAS could be measured quite easily by means of an encoder, a tachometer or some other smart rotational speed sensors. Thus, one could take advantage of advanced techniques in digital time interval analysis applied to a carrier signal from a multiple pulse per revolution encoder on the rotating shaft, to improve the analysis of chain pulses. As it will be shown in this dissertation, the encoder resolution does not sensibly affect the analysis. Therefore, one can easily replace encoders by any smart transducers that have become more popular in rotating machinery. Consequently, a non-contact transducer for example could effectively be used in on-line drill condition monitoring such as the use of lasers or time passage encoder-based systems. This work has gained from previous research performed in Tool Condition Monitoring TCM, and presents a sensor that is already available in the arsenal of sensors and could be an open door for a practical and reliable sensor in automated drilling. iii In conclusion, this dissertation strives to answer the following question: Which one of these methods could challenge the need from manufacturers by monitoring and diagnosing drill condition in a practical and reliable manner? Past research has sufficiently proved the weakness of conventional technologies in industry despite good results in the laboratory. In addition, delayed diagnosis due to time-consuming data processing is not beneficial for automated drilling, especially when the drill wears rapidly at the end of its life. No advanced signal processing is required for the proposed technique, as satisfactory results are obtained using common time domain signal processing methods. The recommended monitoring choice will definitely depend on the sensor that is practical and reliable in industry. / Dissertation (MEng)--University of Pretoria, 2012. / gm2013 / Mechanical and Aeronautical Engineering / MEng / Unrestricted