Parikshit Mehta

Condition based maintenance-systems integration and intelligence using Bayesian classification and sensor fusion

System integration in condition based maintenance (CBM) is one of the biggest challenges that need to be overcome for widespread deployment of the CBM methodology. CBM system architectures investigated in this work include an independent monitoring and control unit with no communication with machine control (Architecture 1) and a data acquisition and control unit integrated with the machine control (Architecture 2). Based on these architectures, three different CBM system applications are discussed and deployed. A verification of the third system was done by performing a destructive bearing test, causing a spindle to seize due to lubrication starvation. This test validated the CBM system developed, as well as provided insights into using sensor fusion for a better detection of bearing failure. The second part of the work discusses intelligence in a CBM system using a Bayesian probabilistic decision framework and data generated while running validation tests, it is demonstrated how the Naïve Bayes classifier can aid in the decision making of stopping the machine before catastrophic failure occurs. Discussing value in combining information supplied by more than one sensor (sensor fusion), it is demonstrated how a catastrophic failure can be prevented. The work is concluded with open issues on the topic with ongoing work and future opportunities.

Model Based Prediction and Control of Machining Deflection Error in Turning Slender Bars

Model based control of machining processes is aimed at improving the performance of CNC systems by using the knowledge of machining process to reduce cost, improving machining accuracy and improving overall productivity. In this paper, real time control of the machining process to maintain dimensional quality when turning a slender bar is addressed. The goal is to actively control the machining feed rate to maintain constant and predicable deflection through a combined force-stiffness model integrated to the process controller. A brief review is presented on manufacturing process models, process monitoring, and model based control strategies such as Model Predictive Control (MPC). The main objective of this paper is to outline a method for deploying such models to process control. To demonstrate this, model of the deflection of the workpiece under tool cutting forces is developed. Unknown process parameters have been calculated using series of FEA simulations and verified with basic experimental data. A simple but effective control strategy has been formulated and simulated. In the initial results, the diameter of bar is maintained within 1.04% error with controller as opposed to up to 4% error without controller. Ultimately, the goal is to deploy such control strategies in the industrial control system. With the continual development in physical understanding of machining processes and affordable computing technology (both software and hardware) coupled with Open Architecture Control (OAC) applied to CNC machine tools, such approaches are now computationally feasible. This will be an enabling factor to deploy model based control in an industrial environment. The last section discusses the proposed hardware architecture to achieve this. The paper concludes with a brief plan for the future work and a summary.Copyright

Correlation of the Volumetric Tool Wear Rate of Carbide Milling Inserts With the Material Removal Rate of Ti–6Al–4V

The objective of this paper is to assess the correlation of volumetric tool wear (VTW) and wear rate of carbide tools on the material removal rate (MRR) of titanium alloys. A previously developed methodology for assessing the worn tool material volume is utilized for quantifying the VTW of carbide tools when machining Ti–6Al–4V. To capture the tool response, controlled milling experiments are conducted at suitable corner points of the recommended feed-speed design space, for constant stock material removal volumes. For each case, the tool material volume worn away, as well as the corresponding volumetric wear profile evolution in terms of a set of geometric coefficients, is quantified—these are then related to the MRR. Further, the volumetric wear rate and the M-ratio (volume of stock removed to VTW) which is a measure of the cutting tool efficiency, are related to the MRR—these provide a tool-life based optimal MRR for profitability. This work not only elevates tool wear from a 1D to 3D concept, but helps in assessing machining economics from a stock material-removal-efficiency perspective as well.

Vibration Analysis of Thin Plates Subject to Piezoelectric Actuation: A New Perspective in Modeling and Numerical Analysis

This paper undertakes model development and numerical simulations of vibration problem of piezoelectrically actuated thin plates with a holistic perspective. Constitutive laws governing piezoelectric actuator are integrated with the potential and kinetic energies of combined plate-actuator system. The equations of motions are derived using variational approach and verified with results obtained by Newton’s equilibrium approach. It is verified that the field coupled components associated with piezoelectric actuator appear as distributed moments over the area of the actuator. The equations of motion are solved using modal analysis deploying Raleigh Ritz method utilizing Boundary Characteristic Orthogonal Polynomials (BCOP). The shape functions generated using this method is used in Assumed Mode Method (AMM) to numerically simulate forced vibration analysis. Since Raleigh Ritz analysis with BCOP can be deployed with the plates of all the geometries, minor modifications in selecting the shape functions enables one to use the same method to calculate natural frequencies of annular plate as well.Copyright

Machining Process Power Monitoring: Bayesian Update of Machining Power Model

Monitoring the CNC machine tool power provides valuable information that aids condition based maintenance, machine efficiency and machining process monitoring. Cutting force in machining process is an interesting variable to measure from monitoring and control point of view. Although the direct methods of measuring the cutting force exist, prohibitive costs do not allow deployment in industrial environment. In the indirect methods of measuring force, measuring the spindle motor current to estimate the cutting power and consequently the cutting force is popular.This work discusses the calibration of spindle current based torque sensor for the estimation of the cutting force in turning operation. The work undertakes handling uncertainty in measurement of the cutting torque measurement. Considering the steady state value, the cutting torque is represented as a polynomial function of the speed and measured power. Though the identification of the unknown coefficients can be done based on the offline tests, in current work, the Bayesian update of coefficients is proposed. This method allows online learning of these coefficients. The cutting torque value based on the model has some variability due to variation in the coefficients and unmodeled dynamics. The iterative learning happens in three stages, namely — Prior belief, likelihood function establishment and update in prior belief with observed data producing posterior belief. The establishment of the priors is done through some offline tests. The likelihood function accounts for noise in the measurement of torque. And finally, Markov Chain Monte Carlo (MCMC) simulations help sampling from unknown posterior distribution. This scheme has ability to sample from any distribution. A single update cycle shows high reduction in the variability of the torque. Experimental data is produced to verify the effectiveness of method; the Bayesian update scheme outperforms least-square polynomial fit method consistently for different cutting speeds and cutting load values.Copyright

A Hydraulic Actuated Automotive Thermal Management System: Theory and Experiment

The performance of engine cooling systems can be improved by replacing the traditional mechanical driven radiator fan and water pump assemblies with computer controlled components. The power requirements for electric servo-motors increase with larger cooling demands which necessitate larger motors and/or a distributed configuration. One solution may be the use of hydraulic-based components due to their high power density and compact size. This paper investigates a thermal management system that features a computer controlled hydraulic actuated automotive fan and water pump. A mathematical model was derived for the hydraulic and thermal system components. To experimentally study the concept, a hydraulic driven fan and coolant pump were integrated with electric immersion heaters and radiator to emulate a vehicle cooling system. The dynamic model exhibited a strong correlation with the experimental test data. For a series of operating profiles, the servo-solenoid proportional control valves successfully tracked prescribed temperature set points to demonstrate that a hydraulic cooling system can maintain engine operating conditions.© 2008 ASME

The Correlation of Volumetric Tool Wear and Wear Rate of Machining Tools With the Material Removal Rate of Titanium Alloys

The objective of this paper is to assess the correlation of volumetric tool wear (VTW) and wear rate of carbide tools on the material removal rate (MRR) of titanium alloys. A previously developed methodology for assessing the worn tool material volume is utilized for quantifying the VTW of carbide tools when machining Ti-6Al-4V. To capture the tool substrate response, controlled milling experiments are conducted at suitable corner points of the feed-speed design space for constant stock material removal volumes. For each case, the tool material volumes worn away, as well as the corresponding volumetric wear profile evolution in terms of a set of geometric coefficients are quantified — these are then related to the MRR. Further, the volumetric wear rate and the M-ratio (volume of stock removed to VTW), which is a measure of the cutting tool efficiency, are related to the MRR — these provide a tool-centered optimal MRR in terms of profitability. This work not only elevates tool wear from a 1-D to 3-D concept, but helps in assessing machining economics from a stock material removal efficiency perspective as well.Copyright

Development of a Condition Based Maintenance Program for a CNC Machine: Part 1—Signal Acquisition, Processing, and Network Communication

Condition based maintenance (CBM) of machine tools is an important maintenance strategy to invoke for a manufacturing company to run as lean as possible. CBM does this by indicating, in advance, the failure of the machine tool components or system, thus reducing the machine downtime. In this paper, the development of such a system is sought. A background review of the need and structure of such a system has been provided as well as the design considerations for the system are discussed. Having those considerations as the target requirements for a CBM system, discussion of a demonstrative system is presented, being implemented on an OKUMA LB 3000EX CNC lathe. Leveraging the Open Architecture Control (OAC) technology built into OKUMA CNC systems, the proposed system shall enhance machine monitoring by integrating the internal and external sensors aboard the machine tool. This work lays the foundation for the framework of a proposed CBM system. Coolant temperatures and spindle vibration signals are acquired and processed using a high speed data acquisition system. Towards the end of the paper, descriptions of how to best use this data and integrate it with the machine tool CNC system have been provided.Copyright

Model Learning in a Multistage Machining Process: Online Identification of Force Coefficients and Model Use in the Manufacturing Enterprise

This work presents a systems approach in machining process control. Traditional force-based machining process control has been focused on single machine-single operation. The force or power sensor is used to measure the instantaneous force/power, and control action is taken by changing the feedrate in real time to follow a given force setpoint. The application of such control has successfully been implemented to prevent chatter and to elongate tool life by minimizing tool wear. This research seeks to extend the application of control algorithms to learn about the machining system (comprised in this context of a workpiece being operated on in progressive machining), and how knowledge generated by the process can be passed on to the next process for optimization. To demonstrate this, turning of a partially hardened bar is explored. A nonlinear mechanistic force model-based control framework attempts to control the cutting force at a designated setpoint, with material properties changing over the cut. The force coefficients for the material are calculated offline using experimental data and Bayesian inference methods. Since the hardened part of the bar will shift the force coefficient values, an online estimation strategy (Bayesian Recursive Least Square estimator) is used to learn the new coefficients as well as satisfying the control objective. With the newly learned coefficients passed downstream, the subsequent operation experiences no compromise of control objective as well reduces the maximum values of force encountered. Numerical analyses presented show the adaptation and control scheme performance.Copyright

Cutting Force Control in Machining: Bayesian Update of Mechanistic Force Model

For closed loop control of machining forces in the turning process, it is well established that identification of the mechanistic force model is necessary to ensure stable operation of the process. This work proposes a novel approach to update the mechanistic force model by incorporating uncertainty in the deterministic framework. Force coefficient values reported in literature are based on wide spectrum of machining conditions and so cause difficulty in predicting the machining force using the mechanistic force model. This variability stems from variation in material workpiece input quality variation. This work proposes to treat force coefficient and process variables (shear stress and friction angles) as random variables and use Bayesian Statistical techniques to infer true distribution of force coefficients via observing cutting force and feed force values and updating shear stress and friction angle joint probability distribution. A numerical analysis is performed for calculating force coefficients for Titanium alloy (Ti6-Al4V) Markov Chain Monte Carlo (MCMC) simulation is performed to sample from the posterior distribution of the force coefficient. A single update cycle shows high reduction in the variability of the force coefficient. Numerical simulations presented indicate that it is possible to implement Bayesian update scheme in a closed loop control of cutting force for online identification of force coefficients and shear stress and friction angle distributions with few required update cycles and efficiently rejects the disturbance caused by changing machining parameters.Copyright