Search

The hot-dip galvanizing process is widely used in the production of steel products as an important anti-corrosion method for steel products. Due to the complexity of the galvanizing process and the immaturity of some technologies, some factory productions of hot-dip galvanizing products are often with defects such as unsuitable spangle size in the head and tail, unsuitable spangle size in the edge, diagonal edge, and penetration streaks (zinc flow lines), which seriously affect the appearance of the zinc layer and subsequent using function, and are likely to cause undesirable consequences. In order to solve these problems, the object detection method in the field of deep learning is applied to the zinc spangles defects detection. Firstly, a dataset of zinc spangles defects on zinc-coated steel surfaces was established, and the dataset is preprocessed in this paper. Data enhancement is used to improve the generalization ability of the model, and Wiener filter is used to reduce the motion blur. This paper combines transformer structure and Bi-FPN structure with the YOLO-v5 model to propose the YOLOv5-Transformer-BiFPN (YOLOv5-TB) model. The YOLOv5-TB model is used for zinc-coated steel defect detection, which can accurately and quickly detect spangles defects on the surface of zinc-coated steel. The experimental results show that the overall performance of YOLOv5-TB network applied to the detection of zinc spangles defects on zinc-coated steel surfaces is better than most of the current mainstream Object detection algorithms. [ABSTRACT FROM AUTHOR]

The efficiency of truck loading-and-unloading tasks has become one of the major problems in logistics transportation, and intelligent loading-and-unloading robots are considered as effective solutions to this problem. Recently, carton detection with deep learning is widely used for those robots; however, lack of large annotated datasets is one of major impediments for rapidly deploying carton detection models. We currently have presented the Sacked Carton Dataset (SCD) that contains carton images from four logistics scenarios, such as Comprehensive Pharmaceutical Logistics Company (CPLC), E-commerce Logistics Company (ECLC), Fruit Market (FM), and Large Furniture Market (LFM). However, due to the domain shift, the carton detection model trained with source domain (which are annotated) has poor generalization ability when applied to the target domain (which are not annotated and from new logistics scenarios). To solve this problem, a novel image synthesis method is proposed to quickly acquire a labeled dataset, which replace the foreground texture of the source datasets with the texture of the target datasets. In this paper, a Local Surfaces Segmentation Algorithm is proposed to extract the carton stacking skeleton and an Invisible Contour Reconstruction Algorithm is proposed to complete the contour of the carton. Our method can keep the context relationship of foreground objects and backgrounds unchanged, and it also can greatly augment the target datasets. Code is available here. [ABSTRACT FROM AUTHOR]

Surface roughness has played a crucial role in determining the quality and performance in service of the machined workpiece. To enhance the performance of the final product, it is necessary to quantify the final surface roughness accurately. To this end, massive physical models and data-driven methods have been devoted to modeling surface roughness. However, a high-performance physical and data-driven surface roughness prediction model is often subject to the complex modeling process and data insufficient in the milling process. To this end, a physics-informed neural network for surface roughness prediction in milling operations is proposed in this paper. By using the proposed method, the physical knowledge can be incorporated into the deep learning prediction model, which can effectively reduce the complexity and data dependencies in the modeling phase. To verify the applicability and accuracy of the model, cutting tests were conducted using various workpieces, cutting tools, and process parameters. The results demonstrated that the proposed method can effectively reduce the data dependence while depicting high performance, which is more reliable to be applied in the manufacturing industries. [ABSTRACT FROM AUTHOR]

Drilling is one of the most classical machining operations. Real-time monitoring of drill wear can effectively testify whether the product fails to meet the specifications due to drill failure. This paper proposes a tool wear assessment and life prediction model based on image processing and deep learning methods, which works effectively for small sample datasets and for low-quality images. The normal areas and worn areas of the drill bits are extracted using the U-Net network and traditional image processing methods, respectively. Moreover, the original dataset is classified using the migration learning technique. The wear level of a drill bit can be accurately evaluated through experimental tests. Testing results show that the proposed method is more convenient and efficient than previous methods using manual measurements. These results can be applied to real-time drill wear monitoring, thus reducing part damage caused by tool wear. [ABSTRACT FROM AUTHOR]

Weakly supervised learning applies image tag labels to train convolutional neural networks to locate defect. In industrial vision system, metal surface is anisotropic under light in all directions and it is inevitable to cause local overexposure due to the natural reflection of active strong light, especially on the cylindrical metal surface. In this paper, injector valve is taken as the representative of cylindrical metal workpieces. Since the variety and complexity of cylindrical metal workpiece defects which cause pixel-level annotation require expensive manual work. This problem hinders the application of convolutional neural network in industries. In order to solve these above challenges, this paper proposed an end-to-end weakly supervised learning framework named Integrated Residual Attention Convolutional Neural Network (IRA-CNN). IRA-CNN only uses image tag annotation for training and performs defect classification and defect segmentation simultaneously. Weakly supervised learning is achieved by extracting category-related spatial features from defect classification scores. IRA-CNN is composed of multiple Integrated Residual Attention Block (IRA-Block) as the feature extractor which improves the accuracy and achieves real-time performance. IRA-Block adds Integrated Attention Module (IAM) which includes channel attention submodule and spatial attention submodule. The channel attention submodule adaptively extracts the channel attention feature map to improve its bilateral nonlinearity and the robustness. IAM can be well integrated into the IRA-CNN makes the neural network suppress the interference of useless background area and highlight the defect area. Satisfied performance is achieved by the proposed method in our own defect dataset which could meet the requirements in the industrial process. Experimental results show that the method has good generalization ability. The accuracy of defect classification reaches 97.84% and the segmentation accuracy is significantly improved compared with the benchmark method. [ABSTRACT FROM AUTHOR]

Over the last few years, advanced deep learning-based computer vision algorithms are revolutionizing the manufacturing field. Thus, several industry-related hard problems can be solved by training these algorithms, including flaw detection in various materials. Therefore, identifying steel surface defects is considered one of the most important tasks in the steel industry. In this paper, we propose a deep learning-based model to classify six of the most common steel strip surface defects using the NEU-CLS dataset. We investigate the effectiveness of two state-of-the-art CNN architectures (MobileNet-V2 and Xception) combined with the transfer learning approach. The proposed approach uses an ensemble of two pre-trained state-of-the-art Convolutional Neural Networks, which are MobileNet-V2 and Xception. To perform a comparative analysis of the proposed architectures, several evaluation metrics are adopted, including loss, accuracy, precision, recall, F1-score, and execution time. The experimental results show that the proposed deep ensemble learning approach provides higher performance achieving an accuracy of 99.72% compared to MobileNet-V2 (98.61%) and Xception (99.17%) while preserving fast execution time and small models' size. [ABSTRACT FROM AUTHOR]

Penetration recognition is the critical technology to improve manufacturing quality and automation level in arc welding. In this paper, recent research advances concerning penetration recognition based on machine learning are comprehensively reviewed, focused on signals feature extraction and application of machine learning in this field. Three types of signals originating from arc welding, namely, weld pool image, arc sound, and arc voltage, are used to analyze the correlation with the penetration state and extract effective features. Next, the following contents briefly introduce some machine learning methods including conventional machine learning and deep learning, and emphatically summarize their application and performance in penetration recognition. Notably, the deep learning method possesses higher classification accuracy and better generalization ability in penetration recognition due to its deeper architecture and effective learning capacity. In the end, some challenges and tendencies, involving multi-sensor information fusion, data-driven recognition, model development incorporating attention mechanism, lightweight model deployment, and real-time control, are presented for further research. This paper is an attempt to provide a reference source and some guidance for researchers who use machine learning methods for penetration monitoring. [ABSTRACT FROM AUTHOR]

Scholarly work points to 5S + 1, a simple yet powerful method of initiating quality in manufacturing, as one of the foundations of Lean manufacturing and the Toyota Production Systems. The 6th S, safety, is often used to prevent future occupational hazards, therefore, reducing the loss of time, money, and human resources. This paper aims to show how Industry 4.0 technologies such as computer-based vision and object detection algorithms can help implement the 6th S in 5S + 1 through monitoring and detecting workers who fail to adhere to standard safety practices such as wearing personal protective equipment (PPE). The paper evaluated and analyzed three different detection approaches and compared their performance metrics. In total, seven models were proposed to perform such a task. All the proposed models utilized You-Only-Look-Once (YOLO v7) architecture to verify workers' PPE compliance. In approach I, three models were used to detect workers, safety helmets and safety vests. Then, a machine learning algorithm was used to verify if each detected worker is in PPE compliance. In approach II, the model simultaneously detects individual workers and verifies PPE compliance. In approach III, three different models were used to detect workers in the input feed. Then, a deep learning algorithm was used to verify the safety. All models were trained on Pictor-v3 dataset. It is found that the third approach, when utilizing VGG-16 algorithm, achieves the best performance, i.e., 80% F1 score, and can process 11.79 frames per second (FPS), making it suitable for real-time detection. [ABSTRACT FROM AUTHOR]

In recent years, research work on intelligent data-driven bearing fault diagnosis methods has received increasing attention. The detection of a fault, whether incipient or moderate, and the monitoring of its evolution are a major challenge in the field of fault diagnosis and are of great industrial interest. For an efficient identification of this type of fault, we propose in this paper a new method of bearing fault diagnosis ("novel BiLSTM" method). This new approach contributes to the improvement of fault diagnosis methods based on BiLSTM networks. The performance was tested under sixteen conditions and for different loads using the Case Western Reserve University (CWRU) bearing dataset under conditions higher than those proposed in the literature dealing with the same problem. The experimental results obtained show that the proposed method has excellent performance. Subsequently, the proposed method was experimentally compared with the CNN model. The results of this comparison showed that the model developed in this paper not only has a higher accuracy rate in the test set but also in the learning process. [ABSTRACT FROM AUTHOR]

Insufficient data is always a challenge for developing an accurate machine learning or deep learning model in manufacturing processes, especially in tool wear monitoring under varied cutting conditions. This paper presents a Random Forest model for predicting tool wear under varied cutting conditions as well as studies extracted signal features. The Random Forest algorithm was chosen as the machine learning model, rather than the novel deep learning model. This was due to the feature importance investigation, which was embedded in the Random Forest algorithm, thereby making it easier to study the physical meanings of signal features. The frequency domain signals were rearranged as features related to spindle speeds and machine tool structure based on domain knowledge. This is the first paper to rearrange the frequency domain signals for observing the physical meanings of selected features. When data normalization was adopted, frequency domain signals related to spindle speeds were excluded from important features. Only spectrum energy related to structure vibration and time domain signals were important features. Data normalization enhanced the weighting of structure vibration features in a machine learning model. This study showed that feature normalization made the machine learning model more adaptable to different cutting conditions. Furthermore, prediction accuracy for cutting condition of spindle speed = 42,000 rpm and feed = 1.5 μm/rev (lowest prediction accuracy among cutting tests in this study) showed an increase from 68.0 to 84.1%. In addition, spindle speed had a more significant effect than feed on classification accuracy in tool wear monitoring based on experimental results. As a result, at least two data sets of the same spindle speed as in tool wear prediction were recommended to be used for model training. When there were at least two data sets in training data with the same spindle speed as in testing data, the study showed prediction accuracies were greater than 75% without data normalization and 81% with data normalization. [ABSTRACT FROM AUTHOR]

The digital industrial revolution calls for smart manufacturing plants, i.e. plants that include sensors and vision systems accompanied with artificial intelligence and advanced data analytics in order to meet the required accuracy, reliability and productivity levels. In this paper, we introduce a surface analysis and classification approach based on a deep learning algorithm. The approach is intended to let machining centres recognise the adequacy of process parameters adopted for the milling operation performed, based on the phenomenological effects left on the machined surface. Indeed, the operator will be able to understand how to change process parameters to improve workpiece quality of subsequent parts by a reverse engineering procedure that reconstructs the process parameters that generated the analysed surface. A shallow convolutional neural network was proposed to work on surface image patches based on a limited training dataset of optimal and undesired cutting conditions. The architecture consists of a series of 3 stacked convolutional blocks. The performance of the proposed solution was validated through 5-fold cross-validation, measuring the mean and standard deviation of the f1-score metric. The algorithm arrived at outperformed the best state-of-the-art approach by 4.8% when considering average classification performance. [ABSTRACT FROM AUTHOR]

As the core machining process element in the overall manufacturing process of a part, the machining process route plays an important role in improving final manufacturing quality. In the current CAPP system, the decision-making of the process route still depends on human-computer interaction and essentially depends on human intelligence. In the past decade, deep learning technology architecture has been gradually improved, which provides a new enabling technology for intelligent process planning. Recently, some researchers have applied deep learning to process route decision-making. However, due to the challenges of data representation and deep learning network construction, this promising solution is still at infancy. To address the two challenges, this paper presents a novel process route generation approach based on deep learning. First, we propose a fourth-order tensor model to represent the geometry and technological requirements of a part. And the relation matrix is constructed to represent the relationships among machining features. The process route is represented as a sequential set of one-hot vectors. Then, we construct an encoder-decoder neural architecture to automatically generate the machining process route for the part. The 3D convolution neuron network-based encoder converts the geometry, technological requirements, and the information of the relationships among machining features into a higher layer of vector representation, and the long short-term memory network-based decoder maps this representation to the process route. The whole neural architecture including the encoder and decoder is jointly trained to maximize the conditional probability of the target process route given the training part. Finally, the paper takes slot cavity parts as examples to verify the feasibility and effectiveness of the proposed approach. [ABSTRACT FROM AUTHOR]

Intelligent monitoring of tool condition is the key to ensuring workshop manufacturing efficiency, product machining quality, and accuracy, and is also an indispensable part of intelligent processing. In the face of complex and massive, multi-source heterogeneous, and low-value density machining process data, the monitoring method based on traditional machine learning is challenging to meet the development needs of intelligent manufacturing. In contrast, with its powerful data processing and automatic feature extraction capabilities, deep learning shows broad application prospects in tool condition intelligent monitoring. Given this, this paper first systematically introduces the components of tool condition intelligent monitoring framework based on deep learning. Subsequently, the basic principles, modeling process, and application status of the four most widely used deep learning models (deep belief network, stacked auto-encoder network, convolutional neural network, and recurrent neural network) in the field of tool condition monitoring are detailed, and the advantages and disadvantages of different models are comparably discussed. Finally, the challenges and prospects of the current tool condition intelligent monitoring based on deep learning are summarized. [ABSTRACT FROM AUTHOR]

Additive manufacturing (AM) applications are rapidly expanding across multiple domains and are not limited to prototyping purposes. However, achieving flawless parts in medical, aerospace, and automotive applications is critical for the widespread adoption of AM in these industries. Since AM is a complex process consisting of multiple interdependent factors, deep learning (DL) approaches are adopted widely to correlate the AM process physics to the part quality. Typically, in AM processes, computer vision-based DL is performed by extracting the machine's sensor data and layer-wise images through camera-based systems. This paper presents an overview of computer vision-assisted patch-wise defect localization and pixel-wise segmentation methods reported for AM processes to achieve error-free parts. In particular, these deep learning methods localize and segment defects in each layer, such as porosity, melt-pool regions, and spattering, during in situ processes. Further, knowledge of these defects can provide an in-depth understanding of fine-tuning optimal process parameters and part quality through real-time feedback. In addition to DL architectures to identify defects, we report on applications of DL extended to adjust the AM process variables in closed-loop feedback systems. Although several studies have investigated deploying closed-loop systems in AM for defect mitigation, specific challenges exist due to the relationship between inter-dependent process parameters and hardware constraints. We discuss potential opportunities to mitigate these challenges, including advanced segmentation algorithms, vision transformers, data diversity for improved performance, and predictive feedback approaches. [ABSTRACT FROM AUTHOR]

Currently, the measurement of guide rail straightness error is basically a direct measurement, which cannot meet the requirements of online straightness error measurement in the machining process of the machine tool. Therefore, this paper proposes a straightness error evaluation model based on the recess pressure, which can realize the online measurement of the straightness error of hydrostatic guideways. To address the nonlinear relationship between the recess pressure and the linear deviation data of the guideway in an experiment, based on deep learning, the hydrostatic guideway straightness error was evaluated. First, the experimental data are analyzed and processed by feature, and a sliding window is processed using the data time sequence. Second, a long short-term memory network model is constructed based on an attention mechanism, the model parameters are obtained through orthogonal experiments, and the theoretical straightness error of the hydrostatic guideway is obtained via training. Finally, the theoretical values and experimental values of straightness error are compared and evaluated with the multilayer perceptron and recurrent neural network models. The results show that the model can effectively evaluate the straightness error of hydrostatic guideways. [ABSTRACT FROM AUTHOR]

In this paper, we propose a machine learning approach for detecting superficial defects in metal surfaces using point cloud data. We compare the performance of two popular deep learning architectures, multilayer perceptron networks (MLPs) and fully convolutional networks (FCNs), with varying feature sets. Our results show that FCNs (F1=0.94) outperformed MLPs (F1=0.52) in terms of precision, recall, and F1-score. We found that transfer learning with pre-trained models can improve performance when the amount of available data is limited. Our study highlights the importance of considering the amount and quality of training data in developing machine learning models for defect detection in industrial settings with 3D images. [ABSTRACT FROM AUTHOR]

This paper introduces a web-platform system that performs semi-automatic compute of several risk indexes, based on the considered evaluation method (e.g., RULA—Rapid Upper Limb Assessment, REBA—Rapid Entire Body Assessment, OCRA—OCcupational Repetitive Action) to support ergonomics risk estimation, and provides augmented analytics to proactively improve ergonomic risk monitoring based on the characteristics of workers (e.g., age, gender), working tasks, and environment. It implements a body detection system, marker-less and low cost, based on the use of RGB cameras, which exploits the open-source deep learning model CMU (Carnegie Mellon University), from the tf-pose-estimation project, assuring worker privacy and data protection, which has been already successfully assessed in standard laboratory conditions. The paper provides a full description of the proposed platform and reports the results of validation in a real industrial case study regarding a washing machine assembly line composed by 5 workstations. A total of 15 workers have been involved. Results suggest how the proposed system is able to significantly speed up the ergonomic assessment and to predict angles and perform a RULA and OCRA analysis, with an accuracy comparable to that obtainable from a manual analysis, even under the unpredictable conditions that can be found in a real working environment. [ABSTRACT FROM AUTHOR]

A safe and reliable avionics system is an important guarantee for airplanes when accomplishing all kinds of tasks. With the ever more comprehensive and intelligentized development of avionics system, the layout of cable brackets becomes more complicated, which results in the quick and reliable assembly, and inspection of cable brackets is becoming more and more demanding. Nowadays, the detection and recognition of cable brackets in aircraft assembly scenes still rely on manual labor to compare the real assembly scene with the designed two-dimensional drawings or graphs. With the increase in number and types of cable brackets, the recognition process in multi-assembly scenes becomes intricate, time-wasting, and highly mistakable. Aiming at the cable bracket recognition problems, this paper proposes a low retrain complexity hybrid model for adaptive recognition of cable brackets. Firstly, this paper builds a weight sharing feature extraction model using improved loss function and multi-scale ensemble convolutional neural network according to the analysis of cable brackets. Secondly, the support vector machine is used to classify the feature vector extracted by pre-trained feature extraction model. By this way, the hybrid model could be retrained quickly for new types of cable brackets never seen in multiple assembly scenes, under condition that bracket number and types are variable. At last, the performances of other feature extraction methods such as SIFT, HOG, and multi-layer perceptron (MLP) are evaluated for contrast with the proposed method. As shown in the results, our method obtains higher recognition accuracy. [ABSTRACT FROM AUTHOR]

As an all-in-one process that blended design, materials, and manufacturing, additive manufacturing (AM) adopted a bottom-up stacking approach, thereby avoiding excessive infrastructure and complex production steps. However, the different process stages of AM were influenced by multi-scale factors such as material characteristics, parameter settings, and process stability, resulting in uncertainty. Therefore, the effective containment and accurate evaluation of automated production processes were crucial. With the rapid development and widespread application of artificial intelligence (AI) technologies in the Industry 4.0 era, computer vision (CV) techniques, as an important component, could achieve multi-level understanding of the real world through traditional machine learning (ML) and efficient deep learning (DL) approaches based on information-dense cross-media data such as images, videos, and three-dimensional models. Hence, the organic integration of CV and AM was of great significance for the virtual simulation, global perception, real-time monitoring, and quality measurement of the machining processes. According to the chronological order and logical relationship of manufacturing engineering, this paper organized and summarized the application of CV techniques in material characteristics perception and stacking process simulation before the implementation of projects, in situ inspection, and law exploration during the manufacturing processes, as well as product evaluation and defect detection after processing, which demonstrated its powerful performance and great potential in the intelligence and integration of AM. [ABSTRACT FROM AUTHOR]

As the integration of mechanical engineering and deep learning fields becomes increasingly intertwined, the application of experimental thermal error modeling in intelligent manufacturing has gained significant importance. In this paper, the issue of spindle thermal error is treated as a multivariate time series problem due to the thermal transfer characteristics. This study aims to address the challenge of modeling intermittent multivariate time series spindle thermal errors under a wide range of environmental temperatures and various operational scenarios. To tackle this challenge, a substantial volume of experimental data, capable of effectively reflecting the patterns of spindle thermal error variations, was collected through experiments conducted at multiple speeds and under various operational scenarios. Subsequently, the acquired thermal error data underwent intermittent multivariate time series transformation (IMTS) to suit the serialized deep learning model. The study introduces the Crossformer model into the field of thermal error modeling for the first time, which is a variant of the Transformer model. The Crossformer model exhibits remarkable adaptability to temporal aspects while effectively maintaining its focus on data features. Ultimately, this study resulted in the development of the IMTS-CrossformerR experimental thermal error model. Throughout the research, a comprehensive examination of various models was undertaken, including two traditional Transformer models and other thermal error deep learning and machine learning models. The results indicate that the proposed model outperforms its counterparts across multiple model metrics and predictive capabilities. Particularly noteworthy is its substantial improvement in the range (± 5) ratio of residual fluctuations reaching 95.7%, a key engineering metric. [ABSTRACT FROM AUTHOR]

In the practical thickener cone systems, the underflow concentration is hard to measure through physical sensors while there exist the high cost and significant measurement delay. This paper presents a novel and deeply efficient long short-time memory (DE-LSTM) method for concentration prediction in the deep cone thickener system. First, the DE-LSTM for thicker systems is developed for feature learning and long temporal preprocessing. Then, the feedforward and reverse LSTM subnetworks are employed to learn the robust information without loss. At last, the experimental verification of an industrial deep cone thicker demonstrates the proposed DE-LSTM's performance outperforms other state-of-the-art methods. [ABSTRACT FROM AUTHOR]

A severe tool wear is often encountered during the process of turning/milling difficult-to-cut materials like Inconel 718. To protect the cutting process from the tool failure, the countermeasure is currently limited to frequent tool changes. Such preventive solution increases not only the machine downtime but also the manufacturing cost. Therefore, there is a strong demand for real-time tool life detection in production lines. In this study, we proposed a stacked LSTM encoder-decoder model for tool wear anomaly detection. Our model only requires "normal" datasets for model training and yields the anomaly score based on the deviation between input and output sequences. This aspect is especially valuable for the actual production lines where the operating conditions are optimized and anomalies are rare. The proposed model is also an end-to-end learning architecture and requires no pre-processing for feature extraction and selection. This characteristic enables real-time processing, which is significant for anomaly detection. In this paper, we first validated our model using artificially generated sinusoidal waveforms and demonstrated its high performance in detecting deviations in amplitude, frequency, waveform bias, and added noise components. The proposed model was then implemented into an actual turning process for Inconel 718 to detect the tool wear in real-time. After being trained by AE (Acoustic Emission) or audio signals captured in the "normal" cutting stage, the model can derive appropriate anomaly scores well matched with the cutting stages of "normal," "transitional" and "anomalous" corresponding to the tool wear. Finally, deep insight into law/rule-based, data-based, and model-based approaches are discussed. [ABSTRACT FROM AUTHOR]

Surface roughness is an important indicator for shaft grinding, and its prediction is always a core issue in practice. The surface roughness of each part for a large shaft would be different with same processing parameters, which leads to traditional processing parameter-based prediction being ineffective; thus, the surface roughness prediction of large shaft grinding is more difficult. This paper proposes a surface roughness prediction method for large shaft grinding based on deep learning, and three kinds of process signals are selected as the inputs, including spindle current, vibration, and acoustic emission. An experiment considering the contribution degree of each processing parameter is designed to generate the dataset on a large shaft grinding platform, and the appropriate surface roughness intervals are divided. The multiple process signals are fused through an attentional CNN-LSTM architecture, in which the CNN is used to extract the features of the process signals, the LSTM deals with the sequential output of CNN, and the self-attention mechanism is utilized to realize the automatic weight allocation. Moreover, the model degradation problem after long-term use of the grinding platform is solved by the parameters fine-tuning approach. The experimental results show that the proposed method achieves the best accuracy than all comparative items, and the model performance can be recovered through parameters fine-tuning; thus, the proposed method is effective. [ABSTRACT FROM AUTHOR]

Deep learning in computer vision is becoming increasingly popular and useful for tracking object movement in many application areas, due to data collection burgeoning from the rise of the Internet of Things (IoT) and Big Data. So far, computer vision has been used in industry predominantly for quality inspection purposes such as surface defect detection; however, an emergent research area is the application for process monitoring involving tracking moving machinery in real time. In steelmaking, the deployment of computer vision for process monitoring is hindered by harsh environments, poor lighting conditions and fume presence. Therefore, application of computer vision remains unplumbed. This paper proposes a novel method for tracking hot metal ladles during pouring in poor lighting. The proposed method uses contrast-limited adaptive histogram equalisation (CLAHE) for contrast enhancement, Mask R-CNN for segmentation prediction and Kalman filters for improving predictions. Pixel-level tracking enables pouring height and rotation angle estimation which are controllable parameters. Flame severity is also estimated to indicate process quality. The method has been validated with real data collected from ladle pours. Currently, no publications presenting a method for tracking ladle pours exist. The model achieved a mean average precision (mAP) of 0.61 by the Microsoft Common Objects in Context (MSCOCO) standard. It measures key process parameters and process quality in processes with high variability, which significantly contributes to process enhancement through root-cause analysis, process optimisation and predictive maintenance. With real-time tracking, predictions could automate ladle controls for closed-loop control to minimise emissions and eliminate variability from human error. [ABSTRACT FROM AUTHOR]

Contour error compensation of the computer numerical control (CNC) machine tool is a vital technology that can improve machining accuracy and quality. To achieve this goal, the tracking error of a feeding axis, which is a dominant issue incurring the contour error, should be firstly modeled and then a proper compensation strategy should be determined. However, building the precise tracking error prediction model is a challenging task because of the nonlinear issues like backlash and friction involved in the feeding axis; besides, the optimal compensation parameter is also difficult to determine because it is sensitive to the machining tool path. In this paper, a set of novel approaches for contour error prediction and compensation is presented based on the technologies of deep learning and reinforcement learning. By utilizing the internal data of the CNC system, the tracking error of the feeding axis is modeled as a Nonlinear Auto-Regressive Long-Short-Term Memory (NAR-LSTM) network, considering all the nonlinear issues of the feeding axis. Given the contour error as calculated based on the predicted tracking error of each feeding axis, a compensation strategy is presented with its parameters identified efficiently by a Time-Series Deep Q-Network (TS-DQN) as designed in our work. To validate the feasibility and advantage of the proposed approaches, extensive experiments are conducted, testifying that our approaches can predict the tracking error and contour error with very good precision (better than about 99% and 90% respectively), and the contour error compensated based on the predicted results and our compensation strategy is significantly reduced (about 60~85% reduction) with the machining quality improved drastically (machining error reduced about 50%). [ABSTRACT FROM AUTHOR]

Nowadays, new challenges around increasing production quality and productivity, and decreasing energy consumption, are growing in the manufacturing industry. In order to tackle these challenges, it is of vital importance to monitor the health of critical components. In the machine tool sector, one of the main aspects is to monitor the wear of the cutting tools, as it affects directly to the fulfillment of tolerances, production of scrap, energy consumption, etc. Besides, the prediction of the remaining useful life (RUL) of the cutting tools, which is related to their wear level, is gaining more importance in the field of predictive maintenance, being that prediction is a crucial point for an improvement of the quality of the cutting process. Unlike monitoring the current health of the cutting tools in real time, as tool wear diagnosis does, RUL prediction allows to know when the tool will end its useful life. This is a key factor since it allows optimizing the planning of maintenance strategies. Moreover, a substantial number of signals can be captured from machine tools, but not all of them perform as optimum predictors for tool RUL. Thus, this paper focuses on RUL and has two main objectives. First, to evaluate the optimum signals for RUL prediction, a substantial number of them were captured in a turning process and investigated by using recursive feature elimination (RFE). Second, the use of bidirectional recurrent neural networks (BRNN) as regressive models to predict the RUL of cutting tools in machining operations using the investigated optimum signals is investigated. The results are compared to traditional machine learning (ML) models and convolutional neural networks (CNN). The results show that among all the signals captured, the root mean squared (RMS) parameter of the forward force ( F y ) is the optimum for RUL prediction. As well, the bidirectional long-short term memory (BiLSTM) and bidirectional gated recurrent units (BiGRU), which are two types of BRNN, along with the RMS of F y signal, achieved the lowest root mean squared error (RMSE) for tool RUL, being also computationally the most demanding ones. [ABSTRACT FROM AUTHOR]

Sheet metal stamping is widely used for high-volume production. Despite the wide adoption, it can lead to defects in the manufactured components, making their quality unacceptable. Because of the variety of defects that can occur on the final product, human inspectors are frequently employed to detect them. However, they can be unreliable and costly, particularly at speeds that match the stamping rate. In this paper, we propose an automatic inspection framework for the stamping process that is based on computer vision and deep learning techniques. The low cost, remote sensing capability and simple implementation mean that it can be easily deployed in an industrial setting. A particular focus of this research is to account for the harsh lighting conditions and the highly reflective nature of products found in manufacturing environments that affect optical sensing techniques by making it difficult to capture the details of a scene. High dynamic range images can capture details of an environment in harsh lighting conditions, and in the context of this work, can capture highly reflective metals found in sheet metal stamping manufacturing. Building on this imaging technique, we propose a framework including a deep learning model to detect defects in sheet metal stamping parts. To test the framework, sheet metal 'Nakajima' samples were pressed with an industrial stamping press. Then optimally exposed, sequence of exposures, tone-mapped and high dynamic range images of the samples were used to train convolutional neural network-based detectors. Analysis of the resulting models showed that high dynamic range image-based models achieved substantially higher accuracy and minimal false-positive predictions. [ABSTRACT FROM AUTHOR]

As fault identification algorithms for rotating machinery based on deep learning are developing rapidly, convolutional neural networks (CNNs) have been attracting extensive attention due to their feature extraction capabilities. However, most of the current CNN–based fault identification models can only evaluate one aspect of the fault, and the recognition accuracy is low in the case of a complex fault. To achieve various and complex fault diagnoses, this paper proposes a multi-tasking atrous convolution neural network (MACNN). First, the network introduces sub-modules such as atrous convolutional layer, batch normalization processing, and PReLU activation function to improve the efficiency of down-sampling and better realize fault feature extraction. Secondly, a parallel multi-independent output layer and a special loss function corresponding to the multitasking structure are proposed, which make the network better in solving the problem of multi-dimensional fault assessment. Finally, based on the MACNN, experiments on rotating machinery fault diagnosis are carried out. In the experiment, three kinds of bearing fault positions and four kinds of depths are identified together, and the accuracy can reach more than 99%, which has obvious advantages over other neural network methods such as artificial neural network (ANN), traditional three-layer convolutional neural network (3L-CNN), and ACNN without multi-task learning. [ABSTRACT FROM AUTHOR]

Bearing fault is a factor that directly affects the reliability of the machine tools. Small sample bearing fault diagnosis plays an important role to improve the reliability of machine tools. However, the over-fitting and weak performance are common problems of small sample bearing fault diagnoses based on deep learning. This paper proposed a different method based on data enhancement and convolutional neural networks (CNN). The method firstly decomposes the vibration signals of the rolling bearing according to the optimal decomposition criterion of variational mode decomposition (VMD). Then, it selects the modes according to the fault frequency characteristics and filters the selected modes into multiple sub-band signals by band-pass filters. Moreover, it computes out the autocorrelation peak vector of the sub-band signals. Finally, the method uses the fault diagnosis network made from a 4-layer neural network, automatically extracts bearing fault features, and predicts the fault types of the testing signals. The experiment shows that the proposed method has a 99% accuracy rate in the rolling bearing fault data set XJTU-SY and requires fewer training samples than the latest methods of NKH-KELM and VMD-CNN. The proposed method has high accuracy under the small sample conditions, which makes it applicable in some practical CNC machine tools with difficulties obtaining bearing samples. [ABSTRACT FROM AUTHOR]

During machining processes, accurate prediction of cutting tool wear is prominent to prevent ineffective tool utilisation and significant resource waste. Tool wear conditions and progression involve complex physical mechanisms, and a promising approach is to deploy heterogeneous sensors and design a deep learning algorithm to conduct real-time tool wear monitoring and precious prediction. To tackle the challenge of deep learning algorithms in processing complex signals from heterogeneous sensors, in this paper, a systematic methodology is designed to combine signal de-noising, feature extraction, feature optimisation and deep learning-based prediction. In more details, the methodology is comprised of the following three steps: (i) signal de-noising is carried out by a designed Hampel filter-based method to eradicate random spikes and outliers in the signals for raw data quality enhancement; (ii) features extracted from heterogeneous sensors in the time and frequency domains are optimised using designed recursive feature elimination and cross-validation (RFECV)-based and Isomap-based methods; (iii) a convolutional neural networks (CNN) algorithm is devised to process the optimised features to implement tool wear prediction. In this paper, a case study showed that 80% features were reduced from the originally extracted features and 86% prediction accuracy was achieved based on the developed methodology. The presented methodology was benchmarked with several main-stream methodologies, and the superior performance of the methodology over those comparative methodologies in terms of prediction accuracy was exhibited. [ABSTRACT FROM AUTHOR]

The manufacturing sector is a fundamental pillar of worldwide economies, contributing markedly to global economic growth. However, the manufacturing industry is persistently confronted with issues impeding its development and expansion, such as manpower shortages, safety concerns, high initial investment for installation, and long return on investment. Within this context, machine tending has become a crucial component of the manufacturing process and potentially serves as a viable solution to the afore-mentioned predicaments. Over the past 5 years, implementing automated machine-tending systems has widely extended from simulation or laboratory environments to practical applications in manufacturing workshops as robotics and artificial intelligence develop rapidly. To fully benefit from the potential of machine-tending applications, it is necessary to comprehend and tackle its associated challenges. Therefore, this paper aims to contribute to the evolution of machine-tending applications by investigating the impacts of emerging trends of advanced technologies, such as autonomous mobile robots, computer vision, machine learning, and deep learning. This systematic literature review is based on the Protocol of Preferred Reporting Items for Systematic Review and Meta-Analyses to analyze the 50 scientific literature related to machine tending in the last five years. [ABSTRACT FROM AUTHOR]

Seam tracking technology is an important part of the intelligent welding field. In this paper, a laser vision-based real-time seam tracking system was built. The system consists of a self-developed laser vision sensor, a six-axis robot, a gas metal arc welding system, and an industrial computer. After building the system, the system calibration was performed. During the seam tracking, the arc light, spatter, and other welding noise have a negative impact on the image processing algorithm to extract the weld feature points, and even lead to system drift and algorithm failure. To this end, a two-stage extraction and restoration model (ERM) was proposed for processing real-time welding images to improve the robustness and accuracy of the seam tracking system. In the ERM, the region of interest was first detected and extracted by the YOLOv5s model, then the extracted images were restored by the conditional generation adversarial network. After using the ERM model, a series of image processing was performed to obtain the coordinates of the weld feature points. The total time consumed by the algorithm is 37 ms per frame on average, which meets the real-time requirement. Moreover, the experimental results show that the seam tracking system based on the ERM can achieve real-time tracking for different types of planar V-bevel welds, and the average error is 0.21 mm, which meets the requirements for actual welding. [ABSTRACT FROM AUTHOR]

With the rapid advancement of wireless technology, the problem of cybersecurity monitoring and detection of cyber-attacks has been receiving widespread attention from industry and academia. The consequences of an undetected cyber-attack in a manufacturing system are not limited to intellectual property theft and cost. It may include destroying equipment, changing product plans, or altering processes. This paper proposes a deep hybrid learning model to improve network intrusion detection systems. To this end, initially, the data set is normalized and preprocessed. Afterward, deep hybrid learning models integrating Attention-based Long Short Term Memory (ALSTM) and Fully Convolutional Neural Network (FCN) with Gradient Boosting, such as Extreme Gradient Boosting (XGBoost) and Adaptive Boost (AdaBoost) are constructed to detect anomalies in traffic data of industrial internet of things (IoT) devices, successfully. The proposed model managed to detect cybersecurity threats in seven different Industrial Internet of Things (IIoT) devices with high-performance measures. The results reveal that deep hybrid learning can ideally detect cyber-security attacks and be versatile in detecting different types of attacks. [ABSTRACT FROM AUTHOR]

Recently, Internet of things (IoT) devices have been widely implemented and technologically advanced in manufacturing settings to monitor, collect, exchange, analyze, and deliver data. However, this transition has increased the risk of cyber-attacks, exponentially. Subsequently, developing effective intrusion detection systems based on deep learning algorithms has proven to become a reliable intelligence tool to protect Industrial IoT devices against cyber threats. This paper presents the implementation of two different classifications and detection utilizing the long short-term memory (LSTM) architecture to address cybersecurity concerns on three benchmark industrial IoT datasets (BoT-IoT, UNSW-NB15, and TON-IoT) which take advantage of various deep learning algorithms. An overall analysis of the performance of the proposed models is provided. Augmenting the LSTM with convolutional neural network (CNN) and fully convolutional neural network (FCN) achieves state-of-the-art performance in detecting cybersecurity threats. [ABSTRACT FROM AUTHOR]

In the growing Industry 4.0 market, there is strong need to implement automatic inspection methods to support manufacturing processes. Tool wear in turning is one of the biggest concerns that most expert operators are able to indirectly infer through the analysis of the removed chips. Automatising this operation would enable developing more efficient cutting processes that turns in easier process planning management toward the Zero Defect Manufacturing paradigm. This paper presents a deep learning approach, based on image processing applied to turning chips for indirectly identifying tool wear levels. The procedure extracts different indicators from the RGB and HSV image channels and instructs a neural network for classifying the chips, based on tool state conditions. Images were collected with a high-resolution digital camera during an experimental cutting campaign involving tool wear analysis with direct microscope imaging. The sensitivity analysis confirmed that the most sensible image channels are the hue value H that were used to teach the network, leading to performances in the range of 95 of proper classification. The feasibility of the deep learning approach for indirectly understanding the tool wear from the chip colour characterisation is confirmed. However, due to the big effects on chip colours of variables as the workpiece material and cutting process parameters, the applicability is limited to stable production flows. An industrial implementation can be foreseen by populating proper large databases and by implementing real-time chip segmentation analysis. [ABSTRACT FROM AUTHOR]

Skin parts are widely used in the modern aerospace industry. The remaining wall thickness of pockets on a skin part must be precisely controlled to reduce overall weight and ensure strength. The in-process dimensional accuracy forecast of remaining wall thickness is urgently required for process efficiency improvement and compensation. In this paper, the spindle power signal of pocket milling is used for the in-process dimensional accuracy forecast. Based on deep learning mothed, the correlation between dimensional accuracy and features extracted from power signal is established. To improve the applicability of the forecast model, a deep transfer learning model based on pockets clustering and feature reconstruction is proposed. The pockets of a skin part are divided into different types through clustering. Then, the trained deep learning network for feature extraction of one type of pocket is transferred to a new type by feature reconstruction. The results indicate that the deep transfer learning model can achieve in-process dimensional accuracy forecast for remaining wall thickness with high precision. [ABSTRACT FROM AUTHOR]

In metal cutting processing, prediction of the tool wear or VB value can help early warning and timely tool replacement before the tool reaches the service life. Although the deep neural network is an effective method to predict the tool wear, the existing research predicts the tool wear only at its next time moment without considering the tool states at different time points. It ignores important information of the tool wear. In this paper, we propose a comprehensive model, which consists of a monitoring module and a prediction module, to monitor and predict the tool wear for the first time. In the monitoring module, a DenseNet model is constructed to monitor the tool wear via sensor signals. In addition, the prediction module based on attention mechanism is developed by simulating the human brain attention to selectively focus on the important part of processing sequence information. An encoder-decoder structure is introduced for a multi-step prediction of the tool wear. Several VB values in the nearest future are predicted by using sequential VB values monitored in the latest past. Experimental studies show that the short-term information has more influence on the tool wear prediction than the long-term information. The proposed method has been used to predict multi-step VB values in milling operations. [ABSTRACT FROM AUTHOR]

Fixtures are used to locate and secure workpieces for further machining or measurement process. Design of these fixtures remains a costly process due to the significant technical know-how required. Automated fixture design can mitigate much of these costs by reducing the dependence on skilled labour, making it an attractive endeavour. Historical attempts in achieving automated fixture design solutions predominantly relied on case-based reasoning (CBR) to generate fixtures by extrapolating from previously proven designs. These approaches are limited by their dependence on a fixturing library. Attempts in using rule-based reasoning (RBR) has also shown to be difficult to implement comprehensively. Reinforcement learning, on the other hand, does not require a fixturing library and instead builds experience and learns through interacting with an environment. This paper discusses the use of reinforcement learning to generate optimized fixturing solutions. Through a proposed reinforcement learning driven fixture design (RL-FD) framework, reinforcement learning was used to generate optimized fixturing solutions. In response to the fixturing environment, adjustments to the reinforcement learning process in the exploration phase is studied. A case study is presented, comparing a conventional exploration method with an adjusted one. Both agents show improved average results over time, with the adjusted exploration model exhibiting faster performance. [ABSTRACT FROM AUTHOR]

Gas tungsten arc welding (GTAW) is a popular technology for joining metallic parts with high stability and quality. However, many GTAW-related processes are conducted manually, which is time-consuming, and the weld quality strongly depends on the skill of the welder. Although various automatic GTAW methods and systems have been developed, controlling the weld quality remains a challenge because of the difficulties involved in predicting the weld quality, such as the penetration depth and backside bead geometry, during welding. Hence, this paper proposes an accurate and effective method for estimating the penetration depth through weld pool monitoring using a convolutional neural network (CNN) trained on weld pool images. The weld pool images contained several objects, each influencing the prediction accuracy. The CNN architecture and the structure of the fully connected layers (FCLs) also affected the prediction accuracy. To optimize the performance of the estimation model, the effects of each object in the weld pool image and the structure of the CNN architecture were analyzed and evaluated. The structure of the FCLs that outputted a quantitative penetration depth was optimized and evaluated through hyperparameter tuning. With the proposed method, the optimized model could quantitatively predict the penetration depth; the mean absolute error was 0.0516 mm, with an R2 value of 0.998. Accurately predicting the penetration depth can be employed in real-time weld quality control to ensure a sound weld back bead. [ABSTRACT FROM AUTHOR]

In the permanent mold casting process, the distribution of mold coating thickness is a significant variable with respect to the coating's thermal resistance, as it strongly influences the mechanical properties of cast parts and the thermal erosion of expensive molds. However, efficient online coating thickness measurement is challenging due to the high working temperatures of the molds. To address this, we propose an indirect monitoring concept based on the analysis of the as-cast surface corresponding to the coated area. Our previous research proves linear correlations between the as-cast surface roughness parameter known as arithmetical mean height (Sa) and the coating thickness for various coating materials. Based on these correlations, we can derive the coating thickness from the analysis of the corresponding as-cast surface. In this work, we introduce a method to quickly evaluate the as-cast surface roughness by analyzing optical images with a deep-learning model. We tested six different models due to their high accuracies on ImageNet: Vision Transformer (ViT), Multi-Axis Vision Transformer (MaxViT), EfficientNetV2-S/M, MobileNetV3, Densely Connected Convolutional Networks (DenseNet), and Wide Residual Networks (Wide ResNet). The results show that the Wide ResNet50-2 model achieves the lowest mean absolute error (MAE) value of 1.060 µm and the highest R-squared (R2) value of 0.918, and EfficientNetV2-M reaches the highest prediction accuracy of 98.39% on the test set. The absolute error of the surface roughness prediction remains well within an acceptable tolerance of ca. 2 µm for the top three models. The findings presented in this paper hold significant importance for the development of an affordable and efficient online method to evaluate mold coating thickness. In future work, we plan to enrich the sample dataset to further enhance the stability of prediction accuracy. [ABSTRACT FROM AUTHOR]

γ-TiAl alloy is a typically difficult material to machine, with common machining defects such as grain pull-out and material spalling during machining, resulting in component scrap. As a result, it is critical to investigate the monitoring method of surface roughness during milling. A new model for predicting surface roughness is proposed in this paper. The prediction problem of surface roughness is turned into a classification problem based on deep learning. The features of the force signal are extracted using a continuous wavelet transform, and the one-dimensional signal is converted into a two-dimensional time–frequency diagram to obtain additional information. The transfer learning mechanism is introduced to make the model faster and improve prediction accuracy. The overfitting problem is solved, and the model's generalization ability is improved through batch normalization and a dropout layer. The results show that the method has high recognition accuracy, with a maximum classification accuracy of 98% and an average accuracy of 96.12%, i.e., a 10.52% improvement over the traditional model, allowing it to accurately realize monitoring of surface quality in milling processing. [ABSTRACT FROM AUTHOR]

During milling operations, wear of cutting tool is inevitable; therefore, tool condition monitoring is essential. One of the difficulties in detecting the state of milling tools is that they are visually inspected, and due to this, the milling process needs to be interrupted. Intelligent monitoring systems based on accelerometers and algorithms have been developed as a part of Industry 4.0 to monitor the tool wear during milling process. In this paper, acoustic emission (AE) and vibration signals captured through sensors are analyzed and the scalograms were constructed from Morlet wavelets. The relative wavelet energy (RWE) criterion was applied to select suitable wavelet functions. Due to the availability of less experimental data to train the LSTM model for the prediction of tool wear, SinGAN was applied to generate additional scalograms and later several image quality parameters were extracted to construct feature vectors. The feature vector is used to train three long short-term memory network (LSTM) models: vanilla, stacked, and bidirectional. To analyze the performance of LSTM models for tool wear prediction, five performance parameters were computed namely R2, adjusted R2, mean absolute error (MAE), root mean square error (RMSE), and mean square error (MSE). The lowest MAE, RMSE, and MSE values were observed as 0.005, 0.016, and 0.0002 and high R2 and Adj. R2 values as 0.997 are observed from the vibration signal. Results suggest that the stacked LSTM model predicts the tool wear better as compared to other LSTM models. The proposed methodology has given very less errors in tool wear predictions and can be extremely useful for the development of an online deep learning tool condition monitoring system. [ABSTRACT FROM AUTHOR]

For wire-feed laser additive manufacturing (WLAM), the build geometrical parameters are one of the indicators of build quality; thus, it is crucial to monitor the geometrical parameters in real-time for quality assurance. However, the current research and development for in situ geometry monitoring are in the early phase due to interweaved correlation of the sensing data and their comprehensive effects on the bead geometry, as well as the high characterization cost to model these effects. This paper focuses on using machine learning techniques to enable in-process geometry monitoring by comprehensively modeling the correlation between the real-time molten pool sensing data and bead geometry properties. A deep learning-based multi-modality convolutional neural network (m-CNN) is trained to take the molten pool image and thermal profile as the input to comprehensively estimate the geometric properties of the build bead. The network is configured by the hyperparameter optimization process and experimentally validated by the real-time molten pool sensing data collected on a wire-feed laser additive manufacturing (AM) system. The effect of using the temperature data from the leading, center, and tailing positions of the molten pool on the prediction performance of the CNN model is studied and analyzed. The CNN model's performance is compared with a support vector regression model for comparison. The developed model represents an in-process monitoring framework for real-time estimation of post-processing bead geometric properties and takes a step towards developing in situ quality control strategy for the metal AM system. [ABSTRACT FROM AUTHOR]

Rapid developments in artificial intelligence and image processing have presented many new opportunities for defect detection in manufacturing processes. In this work, an intelligent image processing system has been developed to monitor inter-layer deposition quality during a wire arc additive manufacturing (WAAM) process. This system reveals the feasibility and future potential of using computer vision knowledge in WAAM. Information produced from this system is to be used in conjunction with other quality monitoring systems to verify the quality of fabricated components. It is tailored to identify the presence of defects relating to lack of fusion and voids immediately after the deposition of a given layer. The image processing system is built upon the YOLOv3 architecture and through moderate changes on anchor settings and achieves 53% precision on surface anomaly detection and 100% accuracy in identifying the fabricated components' location, providing a prerequisite for high-precision assessment of welding quality. The work presented in this paper presents an inter-layer vision-based defect monitoring system in WAAM and serves to highlight the feasibility of developing such intelligent computer vision systems for monitoring the WAAM process for defects. [ABSTRACT FROM AUTHOR]

Waste auditing is one of the tools used to quantify waste generation in construction processes, especially in industrialized building construction facilities that aim to reduce waste. These audits are organized following a regular schedule to monitor manufacturing activities with respect to the waste generated. However, the identification and quantification of waste through occasional audits of activities at any particular workstation remains a biased, manual, error-prone, and monotonous task. This paper proposes the automation of waste auditing in industrialized construction facilities, using as a case study a cutting station on a window manufacturing line. The waste generated during the cutting process is quantified using contour-based image processing algorithms, and the identification of the material is determined by optimized deep learning classification models. This approach allows the continuous acquisition of waste generation data at the workstation level and enables data-driven waste management decision-making that has the potential to support the reduction of waste in industrialized building construction facilities. [ABSTRACT FROM AUTHOR]

Battery cells are central components of electric vehicles. It is important for automotive original equipment manufacturer (OEM) to utilize high-quality battery cells to ensure high performance and safety of their vehicles. This results in the high demand for quality control measures and inspection methods in battery cell manufacturing. Particular relevant features of battery cells are welds for the internal electrical contact. Failures of these welds are often the cause for battery defects in the field and scrap during production. Consequently, there is a strong need to evaluate all welds during manufacturing. However, there is no established method that allows a quick, comprehensive, and cheap inline measurement of the weld quality. This paper presents a new eddy current-based method for nondestructive testing of seam welds as well as a machine learning approach for its validation. A deep learning model has been trained on eddy current measurements to predict results from a reference inspection method, in this case computer tomography. The results prove that eddy current measurements can be used to replicate data acquired by computer tomography, which means that eddy current measurements could be a suitable candidate for nondestructive 100 % inline inspection. More general, this study demonstrates how machine learning may help to get deeper insights into measurement results and to validate new nondestructive testing techniques whose detailed features are yet unknown. The presented evaluation method enables understanding the capabilities and the limits of a new technique and to extract hidden features from the data. In terms of defect segmentation, the trained model applied to an eddy current test data set achieves an accuracy of 93.7 %. Furthermore, the usage of machine learning allows to perform evaluations on artificial product samples with specific defects and features, which avoids the costly production physical samples. [ABSTRACT FROM AUTHOR]

Automatic welding penetration recognition is a significant and challenging research direction to improve the manufacturing quality and intelligence level of gas tungsten arc welding (GTAW). In this paper, a data-driven deep learning approach based on vision transformer (ViT) is proposed to implement welding penetration recognition. To meet the requirement of the model training on data, a large weld pool image dataset composed of four penetration categories is constructed. Twelve ViT models with different architectures are constructed and trained from scratch to explore their feasibility for penetration recognition. Then, pretrained on ImageNet, a state-of-art ViT-B/16 model, is employed, which alleviates the issue of modeling complexity and data insufficiency and improves the validation accuracy by 4.45%. The transfer learning results demonstrate a 98.11% as the test accuracy for the ViT model. Compared with the other conventional recognition methods, the proposed model has fewer misclassifications and superior generalization performance due to self-attention. Furthermore, the attention rollout method is applied to visualize the image region of interest (ROI), confirming the reliability and effectiveness of the developed approach. This study expands the application scope of the ViT to penetration recognition and achieves remarkable effects, laying a reliable foundation for automated welding quality optimization in GTAW. [ABSTRACT FROM AUTHOR]

With the development of Industry 4.0 technology including the Internet of Things (IoT) and deep learning techniques, it is important to reduce maintenance costs and ensure the safety of manufacturing process. The cutting tool degradation can cause significant economic losses and risks for machine users. Accurate prediction of cutting tool is important for making full use of cutter life. Deep learning plays an important role for tool condition monitoring. To overcome these difficulties, this paper aims to propose a new approach in the application of deep learning to estimate the tool wear during the milling process. The proposed methodology is based on the data-driven approach using Variational Mode Decomposition (VMD) and deep learning. Two deep learning machines used in this study, Convolutional Neural Networks (CNN) and Bidirectional long short-term memory (BiLSTM) to achieve collaborative data mining on (VMD) and to enhance the accuracy of modeling. VMD is a new decomposition technique used to decompose signal into sub-time series called intrinsic mode functions (IMFs). However, the VMD performances specifically depend on the constraints parameters that need to be pre-determined for VMD method especially the number of modes. The model development based on 1D-CNN and BiLSTM are selected by using the IMFs as inputs. The performance of the proposed approach is further improved by using the combined CNN and BiLSTM and has shown higher performances in prediction, compared to traditional learning techniques and adopted in previous works highlight the proposed prognostics method's superiority. Among all models, the VMD-CNN-BiLSTM achieve the best performance of modeling with respect to efficiency and effectiveness. [ABSTRACT FROM AUTHOR]

Anomaly in mechanical systems may cause equipment to break down with serious safety, environment, and economic impact. Since many mechanical equipment usually operates under tough working environments, which makes them vulnerable to types of faults, anomaly detection for mechanical equipment usually requires considerable domain knowledge. However, a common dilemma in many practical applications is that one may not be able to obtain the empirical knowledge about anomaly or the history data is completely unlabelled, which makes conventional fault identification methods not applicable. In order to fill the gap, this paper proposes a novel deep learning–based method for anomaly detection in mechanical equipment by combining two types of deep learning architectures, stacked autoencoders (SAE) and long short-term memory (LSTM) neural networks, to identify anomaly condition in a completely unsupervised manner. The proposed method focuses on the anomaly detection through multiple features sequence when the history data is unlabelled and the empirical knowledge about anomaly is absent. An experiment for anomaly detection in rotary machinery through wavelet packet decomposition (WPD) and data-driven models demonstrates the efficiency and stability of the proposed approach. The method can be divided into two stages: SAE-based multiple features sequence representation and LSTM-based anomaly identification. During the experiment, fivefold cross-validation has been applied to validate the performance and stability of the proposed approach. The results show that the proposed approach could detect anomaly working condition with 99% accuracy under a completely unsupervised learning environment and offer an alternative method to leverage and integrate features for anomaly detection without empirical knowledge. [ABSTRACT FROM AUTHOR]

Tool wear and faults will affect the quality of machined workpiece and damage the continuity of manufacturing. The accurate prediction of remaining useful life (RUL) is significant to guarantee the processing quality and improve the productivity of automatic system. At present, the most commonly used methods for tool RUL prediction are trained by history fault data. However, when researching on new types of tools or processing high value parts, fault datasets are difficult to acquire, which leads to RUL prediction a challenge under limited fault data. To overcome the shortcomings of above prediction methods, a deep transfer reinforcement learning (DTRL) network based on long short-term memory (LSTM) network is presented in this paper. Local features are extracted from consecutive sensor data to track the tool states, and the trained network size can be dynamically adjusted by controlling time sequence length. Then in DTRL network, LSTM network is employed to construct the value function approximation for smoothly processing temporal information and mining long-term dependencies. On this basis, a novel strategy of Q-function update and transfer is presented to transfer the deep reinforcement learning (DRL) network trained by historical fault data to a new tool for RUL prediction. Finally, tool wear experiments are performed to validate effectiveness of the DTRL model. The prediction results demonstrate that the proposed method has high accuracy and generalization for similar tools and cutting conditions. [ABSTRACT FROM AUTHOR]