Tool life predictions in milling using spindle power with the neural network technique

doi:10.1016/j.jmapro.2016.03.010

Journal of Manufacturing Processes

Volume 22, April 2016, Pages 161-168

https://doi.org/10.1016/j.jmapro.2016.03.010 Get rights and content

Abstract

Tool wear is an important limitation to machining productivity and part quality. In this paper, remaining useful life (RUL) prediction of tools is demonstrated based on the machine spindle power values using the neural network (NN) technique. End milling tests were performed on a stainless steel workpiece at different spindle speeds and spindle power was recorded. The NN curve fitting approach with different MATLAB™ training functions was applied to the root mean square power (P_rms) values. Sample P_rms growth curves were generated to take into account uncertainty. The P_rms value in the time domain was found to be sensitive to tool wear. Results show a good agreement between the predicted and true RUL of tools. The proposed method takes into account the uncertainty in tool life and the percentage increase in nominal P_rms value during the RUL prediction. Using MATLAB™ on an Intel i7 processor, the computation takes 0.5 s Thus, the method is computationally inexpensive and can be incorporated for real time RUL predictions during machining.

Introduction

In the manufacturing industry, controlling the process to machine a workpiece is vital for enhancing both productivity and quality. There are limitations that can influence the manufacturing process, such tool wear and life, surface roughness, surface location error, and machining stability. Tool wear that occurs due to its interaction with the workpiece is a major limitation to machining processes. Tool failure in the machine can lead to unscheduled downtime and damage to the workpiece. It is estimated that the average machine tool downtime due to tool wear is 7–20% [1], [2], which can result in a significant loss of productivity.

Tool wear is considered stochastic and in general, difficult to predict. There exist many empirical/mechanistic models to predict tool life based on the Taylor tool life equation [3]. However, the models are deterministic and do not consider the underlying uncertainty in tool wear. Furthermore, many experiments are necessary to determine the model constants. Manufacturing enterprises always demand for higher productivity, and the automation of the monitoring process is an effective way to meet this demand. A generic methodology described in [4] for using an intelligent automatic process monitoring system is simplified as shown in Fig. 1, and that can be applied in predicting the tool condition. Tool condition monitoring (TCM) can significantly reduce downtime, and improve productivity and quality by using sensors such as dynamometer, accelerometer, acoustic emission (AE) sensors, and current/power sensors. The use of artificial intelligence (AI) algorithms to process the sensor data is widely implemented in the TCM [5], and the most commonly used techniques include neural networks, neuro-fuzzy models, fuzzy logic, and Bayesian networks. The idea is to train a system to learn from its past behavior and use it to make predictions on tool life.

Modern machine tools are loaded with power sensor in order to track the machine and tool conditions, among others. Taking the advantage of such sensor technology, current demand of (near) real-time data processing can be enhanced for advancing digital manufacturing. However, a literature review presented in the next section clearly reveals that knowledge of using power sensor for TCM and tool life prediction is still very limited. Thus, power sensor data processing using neural networks for predicting remaining useful tool life is the main focus of this research.

Section snippets

Literature review: data sensors

The selection of the sensors, sensory features, and modeling approach is a crucial step for accurate monitoring of the process. Sensors can be separated in two types:

1)
direct measurements such as lasers, optical microscopes, and optical and ultra-sonic sensors which measure the actual dimensions of the worn area on the tool [6];
2)
indirect measurements as dynamometers, accelerometers, and current sensors which measure the signal that can be correlated with tool condition [7].

Principle and overview

The neural network (NN) technique is based on the same function as a neuron in the brain. A simple neuron can be described as shown in Fig. 2. Each input X_n can be set as a vector with an information x_n and a weight $w_{n}$ . The information passing through a transfer function is considered as an output [28]. As illustrated in Fig. 3, the Multi-Layer Perceptron (MLP) consists of an input layer of neurons, one or more hidden layers of neurons, and an output layer of neurons.

The supervised

Machining setup and procedure

In this section, the experimental steps followed to collect tool wear data for a 25.4 mm diameter inserted end mill, Stellram C7792VXP06CA1.0Z4R6.1 (high feed milling cutter) are described. Fig. 6 shows the machining setup on the Okuma Millac 44V vertical machining canter with the Kennametal Stellram tool holder. Four milling carbide inserts (product XPLT060308ER-D41) are inserted in the holder. The workpiece material was a block of stainless steel SS403. The down milling tests were completed at

Uncertainty generation

The growth of P_rms is not perfect and there is uncertainty as a function of machining time and tool life, resulting from different cutting conditions, measurement errors and error in true machining parameters values given by the machine during cutting. According to the trend of the P_rms, the function was approximated using a second order polynomial. The uncertainty was assumed to happen in the following three stages:

•
the beginning of cutting (i.e., no wear yet)
•
the end of the initial wear stage
•

Predicted RUL results and discussions

This section presents results of the RUL prediction for tests at 1380 and 2700 rpm. After the uncertainty generation, a sample lot of 1000 curves were used to predict the RUL. Each curve was compared with the power signal to select the best fitted curve. Once the best fitted curve was extracted, it was used to predict the RUL. Fig. 13 illustrates the algorithm describing this procedure.

To verify the accuracy of the results, the mean absolute error and the maximum error between the true RUL and

Conclusions

Using the spindle power sensor data and a curve fitting method of Artificial Neural Network (ANN), remaining useful life (RUL) of the tools for milling operation was predicted. The root mean square power (P_rms) value in the time domain was found to be sensitive to tool wear. Random sample P_rms growth curves were generated representing the true curve. Each sample curve was attached to a true RUL curve. By combining these sample curves with the curve fitting method, the prediction of the RUL has

References (34)

R.L. Kegg
One-line machine and process diagnostics
CIRP Ann
(1984)
S. Kurada et al.
A review of machine vision sensors for tool condition monitoring
Comput Ind
(1997)
P.W. Marksberry et al.
A comprehensive tool-wear/tool-life performance model in the evaluation of NDM (near dry machining) for sustainable manufacturing
Int J Mach Tools Manuf
(2008)
D.-W. Cho et al.
The state of machining process monitoring research in Korea
Int J Mach Tools Manuf
(1999)
S. Kurada et al.
A review of machine vision sensors for tool condition monitoring
Comput Ind
(1997)
T. Obikawa et al.
Monitoring of flank wear of coated tools in high speed machining with a neural network ART2
Int J Mach Tools Manuf
(2004)
S.C. Lin et al.
Tool wear monitoring in face milling using force signals
Wear
(1996)
X.L. Li et al.
Tool wear monitoring with wavelet packet transform fuzzy clustering method
Wear
(1998)
X.L. Li et al.
Discrete wavelet transform for tool breakage monitoring
Int J Mach Tools Manuf
(1999)
H. Shao et al.
A cutting power model for tool wear monitoring in milling
Int J Mach Tools Manuf
(2004)

N. Ghosh et al.

Estimation of tool wear during CNC milling using neural network-based sensor fusion

Mech Syst Signal Process

(2007)

T. Hill et al.

Artificial neural network models for forecasting and decision making

Int J Forecast

(1994)

J.V. Tu

Advantages and disadvantages of using artificial neural networks versus logistic regression for predicting medical outcomes

J Clin Epidemiol

(1996)

J.V. Abellan-Nebot et al.

A review of machining monitoring systems based on artificial intelligence process models

Int J Adv Manuf Technol

(2009)

A.G. Rehorn et al.

State-of-the-art methods and results in tool condition monitoring: a review

Int J Adv Manuf Technol

(2004)

C. Giardini et al.

A neural network architecture for tool wear detection through digital camera observations

H.H. Shahabi et al.

In-cycle monitoring of tool nose wear and surface roughness of turned parts using machine vision

Int J Adv Manuf Technol

(2008)

Cited by (177)

Research on online identification of surface burnishing tool machining conditions by spindle current signal analysis
2024, Tribology International
The surface burnishing tool (SBT) is optimized for aluminum alloys' surface hardness and surface roughness. The SBT's rolling body wears during machining and reaches a certain threshold where the machining performance decreases. Sandpapers are used to abrade the rolling bodies to simulate surface damage. The preconditioned tools are utilized for machining, and the spindle motor current signals are recorded. The SBT's hardening capacity is weakened when the rolling body's Sa exceeds 0.581, and the finishing capacity is weakened when the Sa exceeds 0.684. The PSO-SVM model accurately identifies the failure point of the SBT's hardening capacity with an accuracy of 96.67%. Another PSO-SVM model accurately identified the failure point of SBT's finishing capacity with an accuracy of 85.83%.
A new nonlinear ensemble framework based on dynamic-matched weights for tool remaining useful life prediction
2024, Engineering Applications of Artificial Intelligence
In the field of remaining useful life (RUL) prediction, the most prominent task is constructing an accurate prediction model. However, it is difficult for single prediction models to satisfy multiple application situations. Therefore, a new nonlinear ensemble RUL prediction framework based on dynamic-matched weights is proposed in this paper. In the proposed framework, the neural network-based method and the stochastic process-based method are first aggregated through a nonlinear weighting formulation to mitigate data limitations and lack of a priori knowledge. Then, a novel ensemble weight dynamic matching algorithm is designed to achieve time-varying weight matching and improve the prediction accuracy. Finally, the ensemble RUL prediction result is characterized by the probability density function (PDF) of the remaining life. Through two milling cutter experiments, the proposed nonlinear ensemble RUL prediction framework is verified with better comprehensive performance. The cumulative relative accuracy (CRA) of the prediction results is greater than 0.6, which outperforms the commonly used tool RUL prediction method.
Tool wear prediction through AI-assisted digital shadow using industrial edge device
2024, Journal of Manufacturing Processes
Flank wear of drilling tools in manufacturing is among the main factors affecting product quality and productivity. In this study, an AI-assisted digital shadow was created for the instant prediction of the drilling flank tool wear. The drilling data were collected simultaneously using an industrial edge device and a rotary dynamometer. Feature engineering was conducted on the collected data from devices in time and frequency domains. A recurrent neural network (RNN) based on bidirectional long short-term memory (Bi-LSTM) and bidirectional gated recurrent unit (Bi-GRU) architectures was implemented on specified tool wear regions dataset. The digital shadow was created using the industrial edge device and the predictive AI model to minimize costs by reducing the need for expensive multi-sensors, manufacturing downtime, and tool underuse or overuse in a smart manufacturing system. The proposed model predicts with high accuracy and computational time efficiency and can be integrated into digital twin systems.
Hybrid prognostics to estimate cutting inserts remaining useful life based on direct wear observation
2024, Mechanical Systems and Signal Processing
The aim of this study is to monitor tool wear through recurrent direct observation, to automatically and optimally assess when it is really necessary to change tools. To achieve this goal, a hybrid prognosis algorithm is formulated to estimate cutting tools’ remaining useful life. Since cutting speed, and more generally process parameters, influences the rate of tool degradation, an adaptive prognosis strategy is presented on the basis of flank wear measurements through the application of a Particle Filter (PF) framework. The adaptability feature allows tracking changes in flank wear evolution. The idea is to fit available degradation curves of cutting tool flank land measurements, through the use of data-driven models, i.e. Multi-Layer Perceptrons (MLP) and cubic polynomials (P3). The Remaining Useful Life of the cutting tool is estimated together with its probability density function, by using a PF framework to adapt MLP weights (or P3 coefficients) along with online flank wear measurements. The devised algorithm was proven to adapt to wear trends from the field, obtained with cutting parameters not previously tested, making it suitable for a robust implementation. The approach was tested when trained upon a single run-to-failure, and validated upon four run-to-failures in different cutting conditions according to a cross-validation inspired technique. P3 was found to be more reliable (from the metrics perspective), whereas MLP allowed to be accurate with greater advance, offering a practical advantage. The proposed algorithm may also be adapted to integrate physical features, like specific force coefficients, with direct wear measurements.
Improving the useful life of tools using active vibration control through data-driven approaches: A systematic literature review
2024, Engineering Applications of Artificial Intelligence
In the present era of sustainable smart manufacturing within the industry 4.0 framework, industries thrive to achieve sustainable development. Machining processes play a substantial role in smart manufacturing. At the same time, the cutting tool is the most significant element of any machining process. The excessive tool wear or sudden failure of the cutting tool causes unplanned downtime, and it also affects the quality of finished products, economics, and effectiveness of the process. Among all the aspects, the vibrations that occur during machining and cutting forces are the most critical parameters, which accelerates the rate of tool wear. Active Vibration Control (AVC) techniques have emerged as promising approaches for mitigating the detrimental effects of vibration on tool performance. To realise the maximum potential of AVC, however, requires a methodical and exhaustive understanding of the existing literature. This study demonstrates the significance of conducting a systematic literature review on improving the Useful Life of cutting tools employing AVC and estimating through data-driven methods. A systematic literature review on AVC and remaining useful life (RUL) estimation of a cutting tool is performed using the "Preferred Reporting Items for Systematic Reviews and Meta-Analysis" (PRISMA) methodology. However, the study primarily highlights the active vibration control through MR fluid and its characteristics, modelling, and control techniques. Moreover, the data-driven approach for the RUL prediction is discussed briefly through data acquisition, data processing, feature extraction and ranking techniques together with decision-making algorithms. This review presents a structured method for identifying, evaluating, and synthesising relevant studies, thus providing a comprehensive overview of the current state of research in the field. This review seeks to identify gaps, trends, and research directions in the application of AVC for tool longevity by analysing a wide variety of literature, including peer-reviewed journal articles, conference proceedings, and technical reports. Researchers, engineers, and practitioners engaged in tool design, maintenance, and optimization will benefit from the findings of this systematic literature review. The findings will provide a consolidated knowledge base for informed decision-making, allowing for the identification of knowledge deficits, research opportunities, and avenues for further study. The ultimate objective of this review is to contribute to the advancement of AVC techniques for extending the RUL of tools, nurturing innovation, and promoting sustainable and efficient practises across a variety of industrial sectors.
Physics-informed hidden markov model for tool wear monitoring
2024, Journal of Manufacturing Systems
Tool wear monitoring (TWM) is an important part of modern intelligent machining. An efficient and accurate monitoring system will effectively improve machining accuracy and reduce processing costs. At present, the methods for TWM can generally be divided into three categories: physics-based models, data-driven models, and hybrid models. These current approaches are suffered from either low prediction accuracy, insufficient model training, or poor result interpretability due to the nature of the model construction approaches. To overcome the shortcomings of these methods, a physics-informed hidden Markov model (PI-HMM) proposed in this study fully integrates the advantages of the above two methods. Firstly, the model constrains the division of hidden states through the tool wear physical model to improve the physical consistency of the model. Secondly, according to the process parameters and the labels generated by the tool wear physical model, the training set is expanded to solve the problem of insufficient training data of the data model. Finally, the influence of real-time features is added to the established wear state output model and the physical data fusion training strategy is used to train the model to realize the classification of tool wear state and the regression of wear value. The experimental results show that the proposed model in this study achieves an average recognition rate of 0.995 for wear classification and an average coefficient of determination (R²) of 0.968 for wear regression.

View all citing articles on Scopus

¹: Currently works at Hitachi America Ltd., Research and Development Division, Farmington Hills, 34500 Grand River Ave, MI 48335, USA.

View full text

Technical PaperTool life predictions in milling using spindle power with the neural network technique

Abstract

Introduction

Section snippets

Literature review: data sensors

Principle and overview

Machining setup and procedure

Uncertainty generation

Predicted RUL results and discussions

Conclusions

CIRP Ann

Comput Ind

Int J Mach Tools Manuf

Int J Mach Tools Manuf

Comput Ind

Int J Mach Tools Manuf

Wear

Wear

Int J Mach Tools Manuf

Int J Mach Tools Manuf

Mech Syst Signal Process

Int J Forecast

J Clin Epidemiol

A review of machining monitoring systems based on artificial intelligence process models

Int J Adv Manuf Technol

State-of-the-art methods and results in tool condition monitoring: a review

Int J Adv Manuf Technol

A neural network architecture for tool wear detection through digital camera observations

In-cycle monitoring of tool nose wear and surface roughness of turned parts using machine vision

Int J Adv Manuf Technol

Technical Paper
Tool life predictions in milling using spindle power with the neural network technique