Industrial Scenarios Research Articles

Diverse local calibration approaches for chemometric predictive analysis of large near-infrared spectroscopy (NIRS) multi-product datasets

This study aimed to assess the predictive accuracy of Near-Infrared Spectroscopy (NIRS) across a large multi-product library, employing novel local calibration methodologies. Three local strategies were examined: LOCAL Algorithm, Locally Weighted Regression predicted on k-nearest neighbor selection (kNN-LWPLSR), along with a newly proposed algorithm within this study called Hybrid Local. These strategies were applied to an extensive multi-product dataset. When compared with Global PLS models, the results exhibited significant reductions in RMSEP values for all local strategies. Particularly, the kNN-LWPLSR demonstrated proficient prediction for the constituents of ADF and DM. The newly proposed method [Hybrid Local] exhibits comparable performance to the LOCAL Algorithm; however, it notably reduces the prediction time by half compared to the latter, representing a significant advancement for the practical implementation of NIRS technology within industrial processing scenarios.

An intelligent compound fault diagnosis method using generalized zero-shot model of bearing

Abstract Compound fault occurrence has been unpredictable, especially in industrial scenarios where it is difficult to collect a large number of labeled samples for compound fault. Based on this, this paper proposes a generative generalized zero-shot learning(GZSL) model aimed at synthesizing compound fault features through training with single fault samples. These synthesized features are then used for the recognition of compound fault. Firstly, in order to construct an accurate and effective semantic vector, the semantic generation module and discriminator are utilized to generate the semantics of compound fault. Secondly, a feature extraction module based on CNN is designed to extract various fault features from the two-dimensional time-frequency diagram. Finally, a fault semantic matching module is designed to match the feature vectors of compound faults with the generated fault semantic vectors. This enables the identification of unseen compound fault by computing their maximum similarity. The experimental results demonstrate that the proposed method achieved H scores of 75.83 and 69.24 on two real fault datasets, ensuring the correct classification of compound fault to the greatest extent possible.

BearingFM: Towards a foundation model for bearing fault diagnosis by domain knowledge and contrastive learning

Monitoring bearing failures in production equipment can effectively prevent finished product quality issues and unplanned factory downtime, thereby reducing supply chain uncertainties and risk. Therefore, monitoring bearing failures in production equipment is important for improving supply chain sustainability. Due to the generalization limitations of neural network models, specific models must be trained for specific tasks. However, in real industrial scenarios, there is a severe lack of labeled samples, making it difficult to deploy fault diagnosis models across massive amounts of equipment in workshops. In order to solve the above issue, this paper proposes a cloud-edge-end collaborative semi-supervised learning framework, which provides multi-level computing power and data support for building a foundation model. A data augmentation method based on the bearing fault mechanism is proposed, which effectively preserves the inherent essential characteristics in vibration signals by normalizing frequency and adding noise in specific frequency bands. A novel contrastive learning model is designed, which narrows the distances between positive samples and widens the distances between negative samples in the high-dimensional space through cross comparisons in the time dimension and knowledge dimension, thereby extracting the most essential characteristics from the unlabeled signals. Multiple sets of experiments conducted on four datasets demonstrate that the proposed approach achieves an approximately 98% fault classification accuracy with only 1.2% labeled samples.

Efficient Mixed-Type Wafer Defect Pattern Recognition Based on Light-Weight Neural Network.

Wafer defect pattern recognition can help engineers improve the production process of semiconductor chips. In real industrial scenarios, the recognition of mixed-type wafer defects is difficult and the production scale of semiconductor wafers is large, which requires high accuracy and speed in wafer defect pattern recognition. This study proposes a light-weight neural network model to efficiently recognize mixed-type wafer defects. The proposed model is constructed via inverted residual convolution blocks with attention mechanisms and large kernel convolution downsampling layers. The inference speed of the inverted residual convolution block is fast, and the attention mechanism can enhance feature extraction capabilities. Large kernel convolutions help the network retain more important feature information during downsampling operations. The experimental results on the real Mixed-type WM38 dataset show that the proposed model achieves a recognition accuracy of 98.69% with only 1.01 M parameters. Compared with some popular high-performance models and light-weight models, our model has advantages in both recognition accuracy and inference speed. Finally, we deploy the model as a TensorRT engine, which significantly improves the inference speed of the model, enabling it to process more than 1300 wafer maps per second.

Adaptive energy management strategy for optimal integration of wind/PV system with hybrid gravity/battery energy storage using forecast models

Hybrid energy systems, including hybrid power generation and hybrid energy storage, have attracted considerable attention as eco-friendly solutions to meet the increasing global energy demands while minimizing environmental impacts. The economic viability and resilience of hybrid energy system solutions depend on careful consideration of economic and reliability factors during the design phase. This paper explores the optimization and design of a wind turbine (WT)/photovoltaic (PV) system coupled with a hybrid energy storage system combining mechanical gravity energy storage (GES) and an electrochemical battery system. An adaptive energy management strategy linked to an optimization process has been proposed for the optimal integration of the WT/PV system with the hybrid Gravity/Battery storage system. Forecast models have been employed to predict solar and wind generation. Simulation models are developed for each component of the multi-source power plant to predict energy flow behavior based on real-world industrial load demand scenarios. The renewables contribution to 100 % reliability is approximately 18 % for photovoltaic and 92 % for wind turbine, primarily attributed to the great wind energy potential of the selected site. The optimal cost of energy is 0.28 €/kWh for a reliability of 100 %, which reduces to 0.03€/kWh when reliability level is decreased to 20 %. The incorporation of forecast models into the study and the hybridization of storage options improve the design accuracy and optimal operation of the multi-source power system leading to reliable and efficient use of renewable energy.

Evaluation and application of kinetic models for Cu-catalyzed acetylene hydrochlorination

Evaluation and application of kinetic models for Cu-catalyzed acetylene hydrochlorination

A Novel Unsupervised Domain Adaptation Transformation Reconstructed Gru Framework Considering Prediction Uncertainty for Machinery Prognostics Under Variable Lubrication Conditions

Abstract As critical components in industrial application scenarios, high-precision and high-confidence health assessment of rolling bearings attract more and more attention. Currently, predictive maintenance obtains outstanding achievements under the same object and working conditions. However, evaluation performances under variable working conditions and different specifications still need to be improved. This study zeroes in on the cross-domain prognostics of rotating machinery under oil and grease lubrication conditions. It proposes an unsupervised domain adaptation transform reconstruction GRU (UDATrGRU) prognostics framework, which captures the common degradation characteristics under different lubrication conditions through the designed second-order statistical quantity, facilitating the following high-precision predictions. To be specific, the vibration degradation features are first extracted through signal preprocessing and then input into UDATrGRU. The developed domain adaptation layer calculates high dimensional projections between diverse data sets, and then corresponding degradation features are statistically aligned under the pressure of the designed quantity. Subsequently, time series modeling and Bootstrap-based uncertainty estimations are carried out. Finally, life-cycle accelerated tests of the rolling bearing from PRONOSITA and ABLT-1A cross-validate the feasibility ' effectiveness of the proposed machinery prognostics framework. The results are pretty promising: compared to existing methods, our UDATrGRU framework has achieved an improvement of at least 5.65% in R2 and a reduction of at least 21.5% in RMSE.

Optimization of target detection model based on Yolov5 network with small samples

As an important part of automotive shock absorber, the columnar parts in automotive shock absorber will inevitably have machining defects during the process, which will not only degrade the performance of the parts, but also degrade or even fail the performance of the final shock absorber after assembly. Yolov5, as a target detection algorithm, has received much attention due to its high accuracy and fast operation speed. However, the algorithm faces some challenges when applied in a practical industrial environment. In this paper, improvement measures are proposed to address the limitations of sample collection and the high speed of pipeline recognition in industrial environments. The network model is optimized and designed. Firstly, the ASPP module is replaced by the SPP module thus improving the viewability throughout the process providing recognition accuracy. Secondly, the Conv and C3 layers of Yolov5s are replaced by Transformer to obtain higher recognition accuracy. By improving and optimizing the above methods, we can better cope with the improvement of detection accuracy under small sample conditions. Experiments show that the method can significantly improve the detection accuracy and operation speed of Yolov5s under the hardware condition of lower computing power, which is more suitable for industrial scenario application scenarios.

Environmental Impact, Mechanical Properties, and Productivity: Considerations on Filler Wire and Scanning Strategy in Laser Welding

Abstract Sustainability, as well as high-quality outcomes, pose significant challenges within the context of current manufacturing cycles, in alignment with European strategies aimed at decarbonization. This framework encourages a systematic evaluation of manufacturing processes in terms of their performance and carbon footprint. One sector where this is particularly relevant is the production of batteries for electric mobility, thanks to its exponential growth. Out of all the processes involved, laser welding stands out as being a critical step since it offers potential energy savings through optimization. With the dual goals of achieving mechanical strength and environmental sustainability, this study investigates alternative solutions for laser welding of aluminum sheets. Different laser welding configurations are tested to evaluate the effect of process setups on weld quality and carbon emissions across different productivity scenarios. The key findings can be summarized as follows: (1) the selection of welding setup significantly influences both quality and sustainability requirements; (2) the optimal conditions for meeting strength requirements may diverge from those aimed at minimizing environmental impact; (3) the choice of the final solution is influenced by the specific industrial scenario. The study specifically demonstrated that aluminum alloys can be welded with higher quality (porosity below 1% and equivalent ultimate strength up to 204 MPa) when filler wire is introduced alongside an active wobbling scanning strategy. Conversely, filler wire can be omitted in scenarios prioritizing high-productivity and low-carbon emissions, such as when employing a linear scanning strategy, resulting in a reduction of equivalent carbon emissions by up to 140%.

Working fluid pair selection of thermally integrated pumped thermal electricity storage system for waste heat recovery and energy storage

Global issues such as the energy crisis and carbon emissions impulse the development of waste heat recovery and energy storage technologies. In most practical industrial scenarios, the electricity supply and consumption cannot be perfectly matched and effective utilization of waste heat is in urgent need. In the present study, we develop a mathematical model to evaluate the thermally integrated pumped thermal electricity storage (TI-PTES) system to achieve off-peak electricity storage along with low-grade waste heat recovery. A double-layer optimization for screening working fluid pairs with high round-trip efficiency is carried out from 24 fluids of the heat pump and 21 fluids of the Organic Rankine cycle (ORC). In the first-layer multi-objective optimization, 3 types of working fluid pair combination strategies are compared and the great improvement of round-trip efficiency by using zeotropic fluids is proved. Among 7 energy storage temperatures covering from 393.15 K to 423.15 K with an increment interval of 5 K, the highest round-trip efficiency of 101.29% is achieved by adopting the zeotropic fluid pair [90Diethyl ether_10Pentane - 80Butane_20Pentane] at 398.15 K. Furthermore, in the second-layer single-objective optimization, the thermo-economic performance indicators of TI-PTES is evaluated and compared under different designing weighting factor groups, which effectively contributes to the screening of working fluids according to designer's trade-off. Finally, through varying energy storage temperatures and designing weighting factors, optimal working fluid pair recommendations including pure fluids and zeotropic ones were proposed to the fluid selection of TI-PTES.

Acoustic signal recovering for rubbing in a dual-rotor system based on diffusion probabilistic models

Inverse problems that recovering target data from measured ones remains a challenging task. This study proposed a novel diffusion probabilistic method for the inverse problem that recovers target acoustic data from distorted ones, in particular industrial scenarios. Specifically, the distribution of acoustic signals was incorporated into the forward process where a parameterized score function is introduced. For the reverse process, a novel strategy is proposed to approximate the posterior distribution to recover acoustic signals from distorted ones through a scheduled Langevine Sampling. The effectiveness of this method was validated on a public dataset as well as signals obtained from a realistic dual-rotor test rig. Experimental results indicated that this method achieved an average improvement of 1.42 dB and boosted the performance of existing state-of-the-art studies in the quality of recovered data. This implies that the proposed approach seems proficient in mitigating distortions arising from Gaussian noise encountered in industrial condition diagnosis scenarios.

Damage identification technique by model enrichment for structural elastodynamic problems

Structural Health Monitoring (SHM) techniques are key to monitor the health state of engineering structures, where damage type, location and severity are to be estimated by applying sophisticated techniques to signals measured by sensors. However, very localized damage detection algorithms applied to dynamics problems when dealing with rigid structures at low-frequency range remains still a big challenge. The last due to the low influence of very localized damage on the overall response of the structure (Saint-Venant principle). In this context, in the present work, we propose a methodology for locally correcting the models from collected data for elastodynamics problems at low-frequency range which is able to predict very localized damage. The proposed technique consists in enriching the structural model in a sparse way and solving the identification problem in the frequency domain, where the influence of damage over a large frequency band is exploited to improve the prediction of the damage location. The advantages and potential of the proposed technique are illustrated for the damage detection in a plate problem, demonstrating the advantages of the method in detecting very localized damage. The proposed technique is limited to a methodological description, and further developments should be considered to approach its applicability in an industrial scenario.

STATNet: One-stage coal-gangue detector based on deep learning algorithm for real industrial application

Coal-gangue object detection has attracted substantial attention because it is the core of realizing vision-based intelligent and green coal separation. However, most existing studies have been focused on laboratory datasets and prioritized model lightweight. This makes the coal-gangue object detection challenging to adapt to the complex and harsh scenes of real production environments. Therefore, our project collected and labeled image datasets of coal and gangue under real production conditions from a coal preparation plant. We then designed a one-stage object model, named STATNet, following the “backbone-neck-head” architecture with the aim of enhancing the detection accuracy under industrial coal preparation scenarios. The proposed model utilizes Swin Transformer as backbone module to extract multi-scale features, improved path augmentation feature pyramid network (iPAFPN) as neck module to enrich feature fusion, and task-aligned head (TAH) as head module to mitigate conflicts and misalignments between classification and localization tasks. Experimental results on a real-world industrial dataset demonstrate that the proposed STATNet model achieves an impressive AP50 of 89.27 %, significantly surpassing several state-of-the-art baseline models by 2.02 % to 5.58 %. Additionally, it exhibits stronger robustness in resisting image corruption and perturbation. These findings demonstrate its promising prospects in practical coal and gangue separation applications.

Integrating intrinsic information: A novel open set domain adaptation network for cross-domain fault diagnosis with multiple unknown faults

Existing popular domain adaptation approaches typically assume that the source and target domains share the same label set. However, in industrial scenarios, the equipment may encounter unknown fault modes because of the harsh operating environment, which limits the application of existing diagnostic methods. To address the above problem, an intrinsic information-guided open set domain adaptation network is proposed for cross-domain fault diagnosis with unknown faults. First, a similarity-based discrimination framework is constructed to enhance the robustness of the model for unknown samples, which learns the similarity between samples and fault prototypes to enhance the classification performance. Then, a multi-information integrated weighting module is designed to quantify the transferability of samples through enhanced domain similarity information learning and prediction uncertainty information learning methods. Additionally, a self-supervised neighborhood clustering learning method is constructed, which enables the model to learn structural information about the target domain and encourages the target samples to cluster closely for better separability. Finally, the weighted open set adversarial training framework effectively facilitates diagnostic knowledge transfer and unknown fault recognition. Comprehensive experimental results on two datasets demonstrate the effectiveness of the proposed method in addressing the open set cross-domain diagnosis problem, which achieves promising performance over the comparison methods.

Evaluation of batch and fed-batch rotating drum biological contactor using immobilized Trametes hirsuta EDN082 for non-sterile real textile wastewater treatment

White rot fungi, in both free and immobilized forms, excel in degrading dyes through adsorption and enzyme degradation. However, existing studies often focus on synthetic wastewater within sterile lab conditions. This study extends the application of Trametes hirsuta EDN082 immobilized in light-expanded clay aggregates (myco-LECAs) packed in a rotating drum biological contactor (RDBC) for treating real textile wastewater under non-sterile conditions to simulate industrial treatment scenarios. Experiments included batch and fed-batch RDBC to assess the impact of additional glucose and stepwise dilution on quality indicators like pH, chemical oxygen demand (COD), ammonia nitrogen (NH4+-N), color, and solids. In fed-batch RDBC with 0.5 % additional glucose, myco-LECAs achieved a maximum of 94 % COD removal (day 15), 99 % NH4+-N reduction (day 33), and 39 % decolorization without additional glucose. In comparison, myco-LECAs in batch RDBC highlighted 89 % NH4+-N reduction in 9 days without pH adjustment or additional nutrients. The pH maintained between 6−9, with no toxicity in Artemia salina, and 97−100 % removal of E. coli. Compared to Indonesian textile wastewater discharge limits, the technology achieved effective ammonia removal below 8 mg/L. This suggests that immobilized T. hirsuta EDN082 in LECAs presents a viable, non-sterile treatment for real textile wastewater.

Deep transfer learning strategy in intelligent fault diagnosis of rotating machinery

Rotating machinery plays an essential part in many engineering fields. It needs prompt solutions to the prognosis and health management to ensure the system reliability. Facilitated by big data and artificial intelligence, intelligent fault diagnosis provides a new approach. As for the insufficient faulty data and complex conditions, deep transfer learning (DTL) presents a possible approach for cross-domain and cross-machine diagnosis. The published reviews thus far mainly emphasize on the analysis of fault diagnosis based on common classes of DTL or industrial application scenarios. This review concentrates on the applications of DTL in rotating machinery. Moreover, present relevant reviews were mainly till the end of 2021. The latest researches are analyzed from then until now. A special main line based on input types is chosen that distinguishes it from other reviews. From this perspective, it is therefore valuable to comprehensively investigate the fault diagnosis of rotating machinery. This survey firstly outlines the fundamental principle and conventional categories of DTL. The primary applications of DTL in fault diagnosis of rotating machinery are then summarized, and more than 100 relative studies have been analyzed. The special perspective of input types is selected and evaluated, including one-dimensional and two-dimensional, on the DTL framework as applied to the rotary machines discussed. Finally, the existing challenges are pointed out and potential future trends of DTL are prospected. This review helps readers to understand the research status and development trends of transfer intelligent fault diagnosis. It serves to the innovative exploration from multiple different scales.

Advances in 2D Materials Based Gas Sensors for Industrial Machine Olfactory Applications.

The escalating development and improvement of gas sensing ability in industrial equipment, or "machine olfactory", propels the evolution of gas sensors toward enhanced sensitivity, selectivity, stability, power efficiency, cost-effectiveness, and longevity. Two-dimensional (2D) materials, distinguished by their atomic-thin profile, expansive specific surface area, remarkable mechanical strength, and surface tunability, hold significant potential for addressing the intricate challenges in gas sensing. However, a comprehensive review of 2D materials-based gas sensors for specific industrial applications is absent. This review delves into the recent advances in this field and highlights the potential applications in industrial machine olfaction. The main content encompasses industrial scenario characteristics, fundamental classification, enhancement methods, underlying mechanisms, and diverse gas sensing applications. Additionally, the challenges associated with transitioning 2D material gas sensors from laboratory development to industrialization and commercialization are addressed, and future-looking viewpoints on the evolution of next-generation intelligent gas sensory systems in the industrial sector are prospected.

Domain Adaptation With Self-Supervised Learning and Feature Clustering for Intelligent Fault Diagnosis.

Domain adaptation indeed promotes the progress of intelligent fault diagnosis in industrial scenarios. The abundant labeled samples are not necessary. The identical distribution between the training and testing datasets is not any more the prerequisite for intelligent fault diagnosis working. However, two issues arise subsequently: Feature learning in domain adaptation framework tends to be biased to the source domain, and unreliable pseudolabeling seriously impacts on the conditional domain adaptation. In this article, a new domain adaptation approach with self-supervised learning and feature clustering (DASSL-FC) is proposed, trying to alleviate the issues by unbiased feature learning and pseudolabels updating strategy. Taking different transformation methods as pretext, the transformed data and its pretext train a neural network in an SSL way. As to pseudolabeling, clusters are taken as the auxiliary information to correct the network predicted labels in terms of the "strong cluster" rule. Then, the updated pseudolabels and their confidence are enforced to further estimate the conditional distribution discrepancy and its confidence weight. To verify the effectiveness of the proposed method, the experiments are implemented on intraplatform and interplatforms for simulating the practical scenarios.

Visual evaluation and information improvement method of smart factory layout based on information value added

Abstract The process of Industry 4.0 is constantly promoted. In order to realize the high-quality connotation development of enterprises, manufacturing enterprises need to carry out information transformation and upgrading. The technology of information flow analysis can consistently assess the worth of the factory, overcome existing technical challenges, and achieve intelligent modernization and enhancement of the system. This study presents a visualization and evaluation of the factory’s layout, production lines, and workstation from three perspectives: macro, meso, and micro. To begin, a value added heat map (VAHM) is created by examining the visualization technique for information value added information (IVA). This analysis aims to assess the extent of information wastage in terms of the factory’s physical space, resources, and staff. Additionally, a method for analyzing the information gain of a production line, which is based on the concept of distributed cognition, is introduced. This method evaluates the production line by integrating the economic principle of information processing (EPIP). Furthermore, when assessing complexity, the volume of information in workstations is measured and enhanced by applying the principle of information processing economy. Ultimately, the method’s validity and feasibility are confirmed by applying it to real-world manufacturing industry scenarios.

Learning to generalize with latent embedding optimization for few- and zero-shot cross domain fault diagnosis

Ensuring the safety and reliability of rotating machinery in modern industrial production and intelligent manufacturing is of paramount importance. While deep learning-based fault diagnosis methods offer promise, the scarcity of fault samples and the variations in distributions between training and test data due to variable working conditions make it difficult for these methods to be applied in industrial scenarios. To surmount these obstacles, we present a novel solution: learning to generalize with latent embedding optimization. Our proposed method, tailored for few-shot and zero-shot cross domain fault diagnosis, shows promise in addressing the industrial fault diagnosis problems under small samples and various working condition. The proposed method builds upon the latent embedding optimization algorithm which capitalizes on the essence of meta-learning, effectively addressing few-shot challenges. Additionally, we harness an efficient pretraining model, enhancing feature extraction and domain adaptation, effectively handling the scarcity of fault data. In tackling the cross domain issue, we introduce an innovative meta-task organization and amplify the episode training strategy within meta-learning. These enhancements empower the model to develop the ability to generalize effectively. The proposed approach is substantiated through comprehensive case studies in bearing and gearbox fault diagnosis. The results demonstrate its exceptional efficacy in both few-shot and zero-shot cross domain fault diagnosis scenarios.