Research on Contour Error Prediction of Five-Axis CNC Machine Tools Based on CLA-Net

In this paper are proposed two new approaches to maintenance strategies for Computer Numerical Control (CNC) machine tools. The analysis is done for different families of CNC machine tools from S.C. Elmet Bacau, a company specialized in aviation. In maintenance actions applied to CNC machine tools is very important to know the evolution of defects and critical state of electrical and mechanical components. The results of this analysis concludes that maintenance actions can be judged by the developing time period diagram, between failure appearance and interruptions in operation. It is also analyzed the financial impact, revealed from known maintenance strategies adopted on CNC machine tools, resulting in a positive approach of condition based maintenance.

High-speed and high-accuracy control technology has been developed in Japan and applied to CNC (Computer Numerical Control) machine tools. Therefore, a CNC machine tool, which is made in Japan, has shown a successful history of its performance. Further development of these technologies is important for Japan to maintain its leadership in advanced manufacturing. The purpose of this research is to develop a high-speed and high-accuracy machining system. Two methods are implemented in the proposed prototype system. The first method is “geometric-model interpolation,” which is used to generate high-speed and highaccuracy command positions, and the second method is “predictive compensation,” which is used to correct machining errors. In the proposed prototype system, these errors are simulated preliminarily and compensated for in the servo data by controlling the position and the feed rate. This report describes the configuration of the proposed prototype system, and the performance of the prototype is evaluated by comparing its speed and accuracy with an existing commercial CNC machine tool. The results show that the prototype system is able to control a CNC machine tool with higher speed and greater accuracy than current CNC machine tool systems.

Thermal error caused by the thermal deformation of computer numerical control (CNC) machine tools is one of the main factors to affect the machining accuracy. With monitoring data of the temperature field, establishing data-driven thermal error model is considered as a more convenient, effective and cost-efficient way to reduce the thermal error. As a matter of fact, it is very difficult to develop a thermal error model with perfect generalization adapting to different working conditions of machining tools. In this paper, a method of an ensemble modeling (EM) based on Convolution Neural Network (CNN) and Back Propagation (BP) Neural Network for modeling thermal error is presented. This ensemble model takes full advantages of two different neural networks, namely CNN having self-extracting feature to solve collinear problem in temperature field and BP can process heat source to thermal error by mapping nonlinear function, then combined into a EM. To demonstrate the effectiveness of the proposed model, an experiment platform was set up based on a heavy-duty CNC machine tool. The results show that the proposed model achieves better accuracy and strong robustness in comparison with only with BP network and CNN network respectively.

Multiaxis machining (e.g., milling) of large aerospace structures is typically carried out using large monument scale expensive computer numerical control (CNC) machine tools. In contrast, multiple degree-of-freedom-articulated robotic arm-based machining has several potential advantages over CNC machine tools, owing to its flexibility (reconfigurability) and lower cost. However, its application is limited by its much lower stiffness and accuracy compared with a CNC machine tool. This paper presents a new method that combines wireless force sensing with a mechanistic model of the milling process to increase the accuracy of robotic milling while preserving its flexibility. Articulated-arm robotic milling experiments are conducted to validate the concept. Experimental results show significant improvement (> 70%) in the dimensional accuracy of simple geometric features machined using the new method.

Most of modern computer numerical control (CNC) machine tools (CNCMTs) use standard G-code (ISO 6983) – based controllers. However, during the generation of G-codes by Computer-Aided Manufacturing (CAM) software, several machining information could be lost, which hinder the development of smart interoperable CNCMTs required for modern manufacturing environment. The STEP-NC (STEP compliant Numerical Control), which is a machine-tool control language governed by the ISO 14649 standard, was developed to overcome the G-code programming limitations. Over the last two decades, several research studies attempted to develop a new generation of CNCMTs controllers taking advantage of STEP-NC capabilities. Nevertheless, these controllers are customized for specific applications, complex to replicate, costly, and not yet accepted in the manufacturing industry. In this paper, an easy to replicate, noninvasive, and low-cost interpreted open architecture STEP-NC controller is developed. This controller uses a Single Board Computer which can be directly connected to a commercial lathe. The controller software is developed based on C++ and can interpret a ISO14649 STEP-NC file, generate a toolpath, simulate the toolpath, and communicate with the CNCMT. The developed STEP-NC controller was successfully tested by simulating the toolpath during the machining of a test part available in the ISO standard. KeywordsSTEP-NCOpen architecture controllerCNC machine toolsMachining

The paper presents a new approach to the modelling and simulation of a computer numerical control (CNC) machine tool feed drive. The hybrid model of the drive incorporates a distributed load, explicit damping factors and measured non-linear effects in order to achieve a realistic dynamic performance. In this way, the shortcomings of traditional modelling methods applied to CNC machine tool axis drives are overcome. A MATLAB/SIMULINK package was used to simulate this new model for a CNC machine tool feed drive. A novel hybrid model with a distributed load, explicit damping factors, backlash and friction was developed and shown to have a similar dynamic response to the machine tool feed drive system under the same conditions. The simulated results display resonance frequencies, while the lumped-parameter models generate only the response of a second-order element.

Nowadays, computer numerical control (CNC) machine tool undertakes more processing tasks than other common machine tools because of its highly automated machining ability and high performance. However, due to the lack of intelligence in machining process planning, machining procedure of products mostly depends on process planners rather than CNC machine tools. To make product quality less dependable on process planner’s ability and improve the efficiency of process planning in order to fulfill changeable market, this paper presents an approach to design and develop CNC machining process knowledge base using cloud technology. The general standard STEP-NC is mapped to web ontology language (OWL) to describe machining process-related knowledge in a readable and comprehensible way. This mapping relation also makes knowledge suitable for storage in HBase. Through this ontology model, descriptive and logical knowledge can be collected. Hadoop platform is used in this approach to provide the NoSQL database HBase for large-scale knowledge storage and MapReduce programming model for large-scale knowledge processing. Taking advantage of MapReduce, knowledge query engine and reasoning engine can be developed. Users can submit task and resource descriptive files to the cloud through CNC controller and get machining process solutions from knowledge base. Evaluation mechanism is also adopted to filter low-quality knoweldge.

Many techniques for the acceleration and deceleration of industrial robots and computer numerical control (CNC) machine tools have been proposed in order to make industrial robots and CNC machine tools perform given tasks efficiently. Although the techniques selecting polynomial functions can generate various acceleration and deceleration characteristics, the major problem is the computational load. The digital convolution techniques are more efficient than the techniques selecting polynomial functions. However, neither velocity profiles of which the deceleration characteristics is independent from the acceleration characteristics nor those of which the acceleration interval is different from the deceleration interval can be generated by the digital convolution techniques. This paper proposes a generalized approach for generating velocity profiles that cannot be generated by the digital convolution techniques. According to the desired characteristics of acceleration and deceleration, each set of coefficients is calculated and is stored. Given a moving distance, and acceleration and deceleration intervals, a velocity profile having the desired characteristics of acceleration and deceleration can be efficiently generated by using these coefficients. Several velocity profiles generated by the proposed technique are applied to one single-axis control system.

Previous techniques for accelerating and decelerating industrial robots and computer numerical control (CNC) machine tools either cannot be computed in real-time, fail to generate various motion profiles, or cannot fully utilise the maximum values of velocity and acceleration. An efficient technique is proposed that has been designed to generate various motion profiles that fully utilise the maximum values of velocity and acceleration.

Method for generating CNC programs based on block-processing time to improve speed and accuracy of machining curved shapes

From the middle of the 20th century, CNC(Computer numerical control) machine tool has brought a revolutionary change to the mechanical manufacturing industry. CNC machine tool is a kind of highly mechatronics product. It has the Characteristic of high precision and high productivity. Therefore it develops quite fast and improves various kind of the quality of the products. It speeds up the development of machinery industry. In order to meet the requirements of the processing accuracy and speed of modern society, the structure and components of the machine have also become complicated. In order to protect the safety of operators and reduce time of machine repaired as well as improve efficiency, the requirements of the reliability of CNC machine tool should be greatly improved. This study will briefly introduce the concept of reliability and put forward some measures to improve the reliability.

Predictive maintenance (PdM) represents a transformative approach in modern manufacturing, aiming to forecast equipment failures through data analysis rather than relying on scheduled or reactive maintenance. This research presents a comprehensive deep learning-based framework for the predictive maintenance of Computer Numerical Control (CNC) machine tools, which are critical and costly assets in precision manufacturing. The proposed system utilizes a multi-sensor data fusion strategy, acquiring real-time operational data including vibration, acoustic emission, spindle current, and temperature. A hybrid deep learning model is developed, integrating Convolutional Neural Networks (CNNs) for automatic feature extraction from high-dimensional sensor signals and Long Short-Term Memory (LSTM) networks to capture temporal dependencies and degradation trends. The model is trained on historical run-to-failure data to learn the complex mapping between multi-modal sensor inputs and the Remaining Useful Life (RUL) of critical components such as spindle bearings and ball screws. Experimental validation is conducted on a three-axis CNC milling machine under controlled operational loads. The results demonstrate that the proposed CNN-LSTM model achieves superior predictive accuracy compared to traditional machine learning benchmarks like Support Vector Regression and standalone neural networks. The system successfully identifies incipient fault conditions with a high degree of precision, providing early warnings significantly ahead of functional failure. This capability enables optimal maintenance scheduling, minimizes unplanned downtime, reduces maintenance costs, and extends the operational lifespan of CNC machinery. The study confirms the significant potential of deep learning in enhancing the intelligence and reliability of industrial predictive maintenance systems.

To address the problems of the low accuracy of geometric error identification and incomplete identification results of the linear axis detection of computer numerical control (CNC) machine tools, a new 21-item geometric error identification method based on double ball-bar measurement was proposed. The model between the double ball-bar reading and the geometric error term in each plane was obtained according to the three-plane arc trajectory measurement. The mathematical model of geometric error components of CNC machine tools is established, and the error fitting coefficients are solved through the beetle antennae search particle swarm optimization (BAS–PSO) algorithm, in which 21 geometric errors, including roll angle errors, were identified. Experiments were performed to compare the optimization effect of the BAS–PSO and PSO and BAS and genetic particle swarm optimization (GA–PSO) algorithms. Experimental results show that the PSO algorithm is trapped in the local optimum, and the BAS–PSO is superior to the other three algorithms in terms of convergence speed and stability, has higher identification accuracy, has better optimization performance, and is suitable for identifying the geometric error coefficient of CNC machine tools. The accuracy and validity of the identification results are verified by the comparison with the results of the individual geometric errors detected through laser interferometer experiments. The identification accuracy of the double ball-bar is below 2.7 µm. The proposed identification method is inexpensive, has a short processing time, is easy to operate, and possesses a reference value for the identification and compensation of the linear axes of machine tools.

With the increasingly common use of industrial automation for mass production, there are many computer numerical control (CNC) machine tools that require the collection of data from intelligent sensors in order to analyze their processing quality. In general, for high speed rotating machines, an accelerometer can be attached on the spindle to collect the data from the detected vibration of the CNC. However, due to their cost, accelerometers have not been widely adopted for use with typical CNC machine tools. This study sought to develop an embedded miniature MEMS microphone array system (Radius 5.25 cm, 8 channels) to discover the vibration source of the CNC from spatial phase array processing. The proposed method utilizes voice activity detection (VAD) to distinguish between the presence and absence of abnormal noise in the pre-stage, and utilizes the traditional direction of arrival method (DOA) via multiple signal classification (MUSIC) to isolate the spatial orientation of the noise source in post-processing. In the numerical simulation, the non-interfering noise source location is calibrated in the anechoic chamber, and is tested with real milling processing in the milling machine. As this results in a high background noise level, the vibration sound source is more accurate in the presented energy gradation graphs as compared to the traditional MUSIC method.

Lightweight heavy-duty CNC horizontal lathe based on particle damping materials