There has been an increasing interest in leveraging machine learning tools for chatter prediction and diagnosis in discrete manufacturing processes. Some of the most common features for studying chatter include traditional signal processing tools such as Fast Fourier Transform (FFT), Power Spectral Density (PSD), and the Auto-correlation Function (ACF). In this study, we use these tools in a supervised learning setting to identify chatter in accelerometer signals obtained from a turning experiment. The experiment is performed using four different tool overhang lengths with varying cutting speed and the depth of cut. We then examine the resulting signals and tag them as either chatter or chatter-free. The tagged signals are then used to train a classifier. The classification methods include the most common algorithms: Support Vector Machine (SVM), Logistic Regression (LR), Random Forest (RF), and Gradient Boost (GB). Our results show that features extracted from the Fourier spectrum are the most informative when training a classifier and testing on data from the same cutting configuration yielding accuracy as high as %96. However, the accuracy drops significantly when training and testing on two different configurations with different structural eigenfrequencies. Thus, we conclude that while these traditional features can be highly tuned to a certain process, their transfer learning ability is limited. We also compare our results against two other methods with rising popularity in the literature: Wavelet Packet Transform (WPT) and Ensemble Empirical Mode Decomposition (EEMD). The latter two methods, especially EEMD, show better transfer learning capabilities for our dataset.

The objective of this paper is to demonstrate the flexure properties of ABS plastic in a 3D printed object as a process to enable embedded pressure sensing capabilities. Developing the potential for non-static 3D parts broadens the scope of the fused deposition modeling (FDM) process to include printing ‘smart’ objects that utilize intrinsic material properties to act as microphones, load sensors, accelerometers, etc. In order to demonstrate a strain-based pressure transducer, strain gauges were embedded either directly on top or in the middle of a flexible ABS diaphragm. Securing a strain gage directly on top of the diaphragm traced a reference pressure more closely than diaphragms with the strain gage embedded halfway into the diaphragm. To prevent temperature-related drift, an additional strain gage was suspended above the secured gage, inside the 3D printed cavity. The additional gage allowed for a half-bridge circuit in lieu of a quarter-bridge circuit, which minimized drift due to temperature change. The ABS diaphragm showed no significant signs of elastic hysteresis or nonlinear buckling. When sealed with 100% acetone, the diaphragm leaked ∼50x slower than as-printed sensors. After pressurizing and depressurizing the devices multiple times, they output pressure readouts that were consistent and repeatable for any given pressure within the operational range of 0 to 7psi. The repeatability of each of the final generation sensors indicates that ‘smart’ objects printed using an FDM process could be individually calibrated to make repeatable recordings. This work demonstrates a concept overlooked previous to now — FDM printed objects are not limited to static models, which lack dynamic motion of the part as an element of design. Altering FDM’s bottom-up process can allow for easily embedding sensing elements that result in printed objects which are functional on the mesoscale.

In this paper, we develop and apply feature extraction and selection techniques to classify tool wear in the shaving process. Because shaving tool condition monitoring is not well-studied, we extract both traditional and novel features from accelerometer signals collected from the shaving machine. We then apply a heuristic feature selection technique to identify key features and classify the tool condition. Run-to-life data from a shop-floor application is used to validate the proposed technique.

Condition Based Maintenance (CBM) systems are crucial for today’s high accuracy machining of exotic materials. For reliable results, CBM systems need early and reliable warning based on prediction models that use multiple types of sensors. In this study, tool flank wear during end milling difficult-to-machine alloys was measured using an optical microscope. Then, vibration data collected with an accelerometer was investigated for its relationship to tool flank wear. The developed relationship between accelerometer output and tool flank wear was validated with further experiments. It was observed from frequency domain responses of these outputs that specific harmonics of the tool pass frequency were dominant, and tool flank wear can be related to the amplitude of these harmonics during machining. This way, it was shown that through accurate online prediction of tool wear, premature interruption of the process as well as machining with a worn tool can both be avoided, improving end-product quality as well as reducing machining costs.

The objective of this work is to identify failure modes and detect the onset of process anomalies in Additive Manufacturing (AM) processes, specifically focusing on Fused Filament Fabrication (FFF). We accomplish this objective using advanced Bayesian non-parametric analysis of in situ heterogeneous sensor data. The proposed method can ultimately lead to intelligent decision making and closed loop control in AM processes. Experiments are conducted on a desktop FFF machine (MakerBot Replicator 2X) instrumented with a heterogeneous sensor array including thermocouples, accelerometers, an infrared temperature sensor, and a real-time miniature video borescope. FFF process failures are detected online using the non-parametric Bayesian Dirichlet Process (DP) mixture model and evidence theory based on the experimentally acquired sensor data. This sensor data-driven defect detection approach facilitates real-time identification and correction of FFF process drifts with an accuracy and precision approaching 85% (average F-score). In comparison, the F-score from existing approaches, such as Probabilistic Neural Networks, Naïve Bayesian Clustering, Support Vector Machines, and Quadratic Discriminant Analysis was in the range of 55% to 75%.

A key concern in modern day manufacturing is developing a reliable method for monitoring tool health. This involves not only establishing a mathematical method for deriving tool failure, but also developing a reliable method for collecting and transmitting this information. This paper presents a reliable wireless network for collecting, analyzing, and predicting tool failure. The wireless network consists of three remote nodes and one coordinator node. Each remote node is mounted on a 1″ R8 mill tool holder and consists of three major components: an accelerometer, microcontroller, and wireless transceiver. These wireless nodes transmit the Z-axis acceleration (the machines vertical Z-axis) as seen by the tool holder back to a central computer. This computer utilizes LabVIEW in order to collect and analyze the acceleration data. The LabVIEW program also acts as a simple user interface for indicating states of tool failure, along with providing a means for publishing the information to the internet for process monitoring by remote users. The wireless network discussed in this paper is robust, inexpensive, and was found to reliably monitor and predict tool health conditions. The innovative aspect of this work is the ability to use commercial-off-the-shelf wireless sensors to remotely monitor multiple rotational machines from a single location on site and over the internet to off-site locations.

This paper discusses a preliminary setup for an ongoing research project with goals of off-line modeling and optimization for a CNC turning process of AISI 4137 steel alloy; followed by online monitoring, optimization, and control of the machining process. A full factorial Design Of Experiment (DOE) of three machining parameter factors was created in Minitab™ and Analysis of Variance was performed, in order to determine which parameters influenced the machining process the most. Accelerometers, acoustic emission sensors and force sensors have given researchers insights into the relationships between mechanical vibration and tool condition during the turning process. Similarly, correlations have been recognized between electrical power consumption, machining forces, tool temperature, and tool condition. While monitoring the machining process with sophisticated force and acceleration sensors is effective, implementation in a large scale factory environment may not be an economical solution to online monitoring and control. Finding an ideal combination of sensors capable of monitoring significant factors that affect the CNC steel turning process will allow process optimization and reduce the cost of machining.

A low cost, wireless vibration sensor system has been developed for noninvasive integration into commercial end milling tool holders. Electret based accelerometers are used as the sensors and a Bluetooth compatible digital transmitter is used as the sensor interface. The use of mass market consumer electronic components is low cost and plug and play with modern PC hardware. Two prototypes were built and, in both cases, were able to collect good quality data at high sampling rates. The objective of the research is to enable accurate observation of NC metal cutting system dynamics. Initial results indicate the system can be used to estimate tool runout and detect the onset of regenerative chatter, prior to workpiece damage.

Traditional piezoelectric accelerometers used for machine condition monitoring are expensive and represent a capital risk when placed in the harsh environment of a cutting process. Additionally, these components require signal conditioning hardware and are sampled on a PC via an independent data acquisition interface (DAQ card). The goal of the research discussed herein is to test an industrial-friendly electret-based accelerometer that can perform tasks similar to a traditional piezoelectric accelerometer. The sensor will be adapted to utilize Bluetooth wireless data capabilities, further enhancing the sensors industrial practicality. The output of this electret-based sensor will be compared to the output of a traditional piezoelectric accelerometer and accompanying DAQ. More specifically, the study will focus on the effects of elevated temperature on the sensor. To achieve this, a comparison of both the electret and piezoelectric accelerometer response spectra will be observed over a range of 21°C to 77°C. To further validate the sensor, turning data is collected wirelessly from the sensor and compared to the output of the traditional piezoelectric sensor. Finally, the performance of the sensor for monitoring a tool’s condition during turning is evaluated and presented. The generated trend is contrasted to the comparable trend developed from the piezoelectric-based accelerometer.

In this paper, an endmill condition monitoring technique is presented for curvilinear cutting. This algorithm operates without the need for prior knowledge of cutting conditions, tool type, cut curvature, cut direction, or directional rate of change. This technique is based on an autoregressive-type monitoring algorithm which is used to track the tool’s condition using a tri-axial accelerometer. Accelerometer signals are monitored due to the sensors relatively low cost and since use of the sensor does not limit the machining envelope. To demonstrate repeatability, eight life tests were conducted. The technique discussed herein successfully prognosis impending fracture or meltdown due to wear in all cases, providing sufficient time to remove the tools before failure is realized. Furthermore, the algorithm produces similar trends capable of forecasting failure, regardless of tool type and cut geometry. Success is seen in all cases without requiring algorithm modifications or any prior information regarding the tool or cutting conditions.

Tracking the health of cutting tools under typical wear conditions is advantageous to the speed and efficiency of manufacturing processes. Existing techniques monitor tool performance through analyzing force or acceleration signals whereby prognoses are made from a single sensor type. However, this work proposes to enhance the spectral output of autoregressive models by combining tri-axial accelerometer and tri-axial dynamometer signals. Through parallel processing of force and acceleration signals using single six degree of freedom modeling, greater spectral resolution is achieved. Two entirely independent methods of tracking the tool wear are developed and contrasted. First, using the discrete cosine transform, primary component analysis will be applied to the spectral output of each AR auto and cross spectrum (Method 1). Each discrete cosine transform of the 6 dimensional spectral data is analyzed to determine the magnitude of the critical (primary) variance energy component of the respective spectrum. The eigenvalues of these selected spectral energies are then observed for trends towards failure. The second method involves monitoring the eigenvalues of the spectral matrices centered at the toothpass frequency (Method 2). The results of the two methodologies are compared. Through the use of the eigenvalue method, it is shown that, for straight and pocketing maneuvers, both methods successfully track the condition of the tool using statistical thresholding.