Real Manufacturing Processes Research Articles

32‐2: Multi AI Approaches for Improving OLED Display Pattern Repair in Manufacturing Processes

The increasing resolution and complexity of OLED (organic light-emitting diode) panels have led to a rise in defects during the manufacturing process. Repairing these defects is crucial for enhancing the competitiveness of OLED manufacturing. The skill level of inspection operators has traditionally been the foundation of the repair system. However, an Image Processing-based Rule Base Auto Repair System has partially replaced the reliance on inspection operators. Yet, it falls short of completely replacing them without additional verification and re-repair. To overcome the challenges of detecting overlap areas between layers and detecting defects in gap regions, which are major causes of reduced accuracy in repair area detection, we explored the use of DL (deep learning) techniques. In this proposal, we introduced segmentation techniques commonly used in autonomous driving and medical fields to improve the performance of layer detection for repair targets. By developing a DL Auto Repair Process based on accurate segmentation detection of repair targets, we established a stable system by combining classification DL and detection DL. The proposed technology has led to the development of an Auto Repair System that eliminates the need for inspection operators for the specific defects. The proposed method has been successfully applied in real manufacturing processes

TPMS-based auxetic structure for high-performance airless tires with variable stiffness depending on deformation.

A novel auxetic structure applicable to airless tire spokes is designed based on the primitive-type triply periodic minimal surface (P-TPMS) to have higher stiffness through deformation under compressive force. For becoming higher stiffness by deformation, an unit cell of auxetic structure is proposed and its characteristics according to design parameters are studied. Based on the parametric study, a rotated primitive-type auxetic structure (RPAS) is designed, and the deformative behaviors of an airless tire with the RPAS spokes are compared with a generally used honeycomb spoke. Simulation and experiment results show that the designed RPAS tire exhibits more stable behavior through higher rigidity depending on the deformation state when compressed on flat ground and obstacles. This variable stiffness characteristic of RPAS tires can be advantageous for shock absorption and prevention of large local deformations. Also, the manufacturability of the designed auxetic structure is evaluated using real rubber-based additive manufacturing processes for practical application in the tire manufacturing industry.

Grain Crystallinity, Anisotropy, and Boundaries Govern Microscale Hydrodynamic Transport in Semicrystalline Porous Media.

Polycrystallinity is often an unintended consequence of real manufacturing processes used to produce designer porous media with deterministic and periodic architectures. Porous media are widely employed as high-surface conduits for fluid transport; unfortunately, even small concentrations of defects in the long-range order become the dominant impediment to hydrodynamic transport. In this study, we isolate the effects of these defects using a microfluidic analogy to energy transport in atomic polycrystals by directly tracking capillary transport through polycrystalline inverse opals. We reveal─using high-fidelity florescent microscopy─the boundary-limited nature of flow motions, along with nonlinear impedance elements introduced by the presence of "grain boundaries" that are separating the well-ordered "crystalline grains". Coupled crystallinity, anisotropy, and linear defect density contribute to direction-dominated flow characteristics in a discretized manner rather than traditional diffusive-like flow patterns. Separating individual crystal grains' transport properties from polycrystals along with new probabilistic data sets enables demonstrating statistical predictive models. These results provide fundamental insight into transport phenomena in (poly)crystalline porous media beyond the deterministic properties of an idealized unit cell and bridge the gap between engineering models and the ubiquitous imperfections found in manufactured porous materials.

Modern educational trends and technologies for training engineers for Industry 4.0

Podlesny S. V. Modern educational trends and technologies for training engineers for Industry 4.0
 The fourth industrial revolution, also known as Industry 4.0, is transforming the manufacturing industry and creating a demand for highly skilled process engineers. To meet this demand, educational institutions must adapt to the changing needs of environmental science and provide education that prepares students for the challenges of Industry 4.0. This article discusses modern educational trends and technologies for training materials processing engineers for Industry 4.0. One of the trends in modern education is the adoption of interdisciplinary curricula. Interdisciplinary curricula also prepare materials processing engineers to work in cross-functional teams. Project-based learning, an approach to education that emphasizes learning through projects rather than traditional lectures, is an effective way to prepare material processing engineers for Industry 4.0 as it allows students to apply their knowledge and skills to real-world problems. Virtual and augmented reality technologies are becoming more and more popular in education. These technologies provide immersive learning. Artificial intelligence is another technology that is transforming the manufacturing industry. Materials processing engineers must have a basic understanding of artificial intelligence in order to work effectively in Industry 4.0. Educational institutions can incorporate artificial intelligence into their curricula by teaching courses in machine learning, data analysis, and robotics. Schools should partner with industry to provide students with hands-on experience and familiarity with real manufacturing processes. Educational institutions must adapt to the changing demands of industry and provide training that prepares students for the future of work.

Energy-efficient scheduling for flexible job shop under multi-resource constraints using non-dominated sorting teaching-learning-based optimization algorithm

Due to the fact that the real manufacturing processes are often constrained by many kinds of resources and the trend that the energy consumption of factories is regulated more and more strictly, this paper studies the energy-efficient multi-resource flexible job shop scheduling problem (EE-MRFJSP). The goal is to minimize the energy consumption and completion time for all of the jobs’ production. Firstly, a general mathematic model for EE-MRFJSP is set up, in which the unit energy consumptions of the main resource’s different states are varied, and a constraint formula to ensure no crossover working periods for any resource is included. Then, a non-dominated sorting teaching-learning-based optimization(NSTLBO) algorithm is proposed to solving the problem, the details of NSTLBO include the real encoding method, Giffler Thompson rule for decoding, non-dominated sorting rule to rank the pareto sets and crowding distance of solution for maintaining the population’s diversity, and the traditional two evolving stages: teacher education and student mutual study. Finally, comparative experiments are made based on some new designed instances, and the results verify our proposed NSTLBO algorithm can effectively solve the EE-MMFJSP, and has obvious advantages by comparing with NSGA-II, NRGA, and MOPSO.

Assessment of quality predictions achieved with machine learning using established measurement process capability procedures in manufacturing

Abstract The increasing amount of available process data from machining and other manufacturing processes together with machine learning methods provide new possibilities for quality control and condition monitoring. A prediction of the workpiece quality in an early machining stage can be used to alter current quality control strategies and could lead to savings in terms of time, cost and resources. However, most methods are tested under controlled lab conditions and few implementations in real manufacturing processes have been reported yet. The main reason for this slow uptake of this promising technology is the need to prove the capability of a machine learning method for quality prediction before it can be applied in serial production and supplement current quality control methods. This article introduces and compares approaches from the fields of machine learning and quality management in order to assess predictions. The comparison and adaption of the two approaches is carried out for an industrial use case at Bosch Rexroth AG where the diameter and the roundness of bores are predicted with machine learning based on process data.

The Three-Level Elastoviscoplastic Model and Its Application to Describing Complex Cyclic Loading of Materials with Different Stacking Fault Energies.

The development of new technologies for thethermomechanical processing of metals and the improvement of the existing ones would be unattainable without the use of mathematical models. The physical and mechanical properties of alloys and the performance characteristics of the products made of these alloys are generally determined by the microstructure of materials. In real manufacturing processes, the deformation of metals and alloys occurs when they undergo complex (non-proportional) loading. Under these conditions, the formation of defect substructures, which do not happen at simple (proportional) loading, can take place. This is due to the occurrence of a great number of slip systems activated under loading along complex strain paths, which leads, for instance, to the more intense formation of barriers of different types, including barriers on split dislocations. In these processes, the formation and annihilation of dislocations proceed actively. In this paper, we present a three-level mathematical model that is based on an explicit description of the evolution dislocations density and the formation of dislocations barriers. The model is intended for the description of arbitrary complex loads with an emphasis on complex cyclic deformation.The model is composed of macrolevel (a representative macrovolume of the material that can be considered as an integration point in the finite-element modeling of real constructions), and mesolevel-1 (description of the mechanical response of a crystallite) and mesolevel-2 (description of the defect structure evolution in a crystallite) submodels. Using the model, we have performed a series of numerical experiments on simple and complex, monotonic and cyclic deformations of materials with different stacking fault energies, analyzed the evolution of defect densities, and analyzed the challenges of a relationship between the complexity of loading processes at a macrolevel and the activation of slip systems at low scale levels.

Deep transfer Wasserstein adversarial network for wafer map defect recognition

Deep neural networks (DNNs) are capable of extracting effective features from data by using deep structure and multiple non-linear processing units. However, they dependent on large datasets from the same distribution. It is difficult to collect wafer maps with various defect patterns in semiconductor manufacturing process. A new deep transfer learning model, deep transfer Wasserstein adversarial network (DTWAN) is proposed to recognize wafer map defect. An adaptive transfer learning framework based on adversarial training is proposed for DTWAN, where multi-stage optimization based on the maximum mean discrepancy (MMD), cross entropy, and adversarial loss is performed. Finally, a generative adversarial algorithm is developed to guide the model to extract general features from source and target domain. DTWAN transfers the key knowledge of the source domain data to target domain, and effectively reduces cost of data collection and improve industrial applicability of the recognition model. The effectiveness of DTWAN is verified on the transfer learning tasks from simulation wafer maps to real wafer maps. The testing results indicate that DTWAN is superior to those typical transfer learning algorithms in WMDR. This study will provide effective guidance to transfer knowledge of simulation data that are easy to collect to real manufacturing processes.

Fault detection based on ICA-GLR for non-Gaussian industrial processes

ABSTRACT As the growth of Industry 4.0, online fault detection plays a crucial role in ensuring the manufacturing quality. Generally, the fault detection methods can be classified into model-based and data-driven methods. There are advantages/disadvantages between two methods. In this study, we integrated both methods in order to develop an efficient fault detection method for non-Gaussian industrial processes. The data-driven method, independent component analysis (ICA) is used to extract non-Gaussian information and dimensionality reduction. Meanwhile, the model-based method, generalized likelihood ratio (GLR) test is adopted as the charting statistic. The proposed ICA-GLR method has advantages of 1) detecting a wide range of process changes, 2) estimating the change points and 3) needless prior parameters to be specified by practitioner. The efficiency of the proposed ICA-GLR fault detection method will be verified via implementing one simulated non-Gaussian process and two real manufacturing processes: Tennessee Eastman process and semiconductor manufacturing process. Results demonstrate that the proposed ICA-GLR method has superior fault detectability when compared to traditional methods, such as principal component analysis and ICA.

KIcker: An Industrial Drive and Control Foosball System automated with Deep Reinforcement Learning

The majority of efforts in the field of sim-to-real Deep Reinforcement Learning focus on robot manipulators, which is justified by their importance for modern production plants. However, there are only a few studies for a more extensive use in manufacturing processes. In this paper, we contribute to this by automating a complex manufacturing-like process using simulation-based Deep Reinforcement Learning. The setup and workflow presented here are designed to mimic the characteristics of real manufacturing processes and proves that Deep Reinforcement Learning can be applied to physical systems built from industrial drive and control components by transferring policies learned in simulation to the real machine. Aided by domain randomization, training in a virtual environment is crucial due to the benefit of accelerated training speed and the desire for safe Reinforcement Learning. Our key contribution is to demonstrate the applicability of simulation-based Deep Reinforcement Learning in industrial automation technology. We introduce an industrial drive and control system, based on the classic pub game Foosball, from both an engineering and a simulation perspective, describing the strategies applied to increase transfer robustness. Our approach allowed us to train a self-learning agent to independently learn successful control policies for demanding Foosball tasks based on sparse reward signals. The promising results prove that state-of-the-art Deep Reinforcement Learning algorithms are able to produce models trained in simulation, which can successfully control industrial use cases without using the actual system for training beforehand.

Highly Sensitive Nondestructive Tunneling Magneto Resistive Imaging: Simulation and Experimental Validation

With the increased use of high-performance thin steels, detecting internal or external micro defects in a manufacturing process has become critical for the steel manufacturing process's cost and quality management. For detecting defects, magnetic-based nondestructive testing (NDT) methods have been researched widely due to their simplicity, ease of use, and contactless nature. However, magnetic-based approaches suffer from low sensitivity, thus limiting their use. Here, we present a highly sensitive tunneling magnetoresistive (TMR) imaging system to detect sub-surface defects in thin steel samples. Artificially created holes in the steel plates were magnetically imaged by measuring magnetic flux leakage (MFL). The experimental results were compared against the simulated results to confirm the visibility of the holes. We also confirmed the successful visualization of the sub-surface defects in the steel plates formed during the real manufacturing processes. Thus, based on these results, we believe the TMR method holds great potential as an NDT technique for steel manufacture management.

AdaPipe: A Recommender System for Adaptive Computation Pipelines in Cyber-Manufacturing Computation Services

The industrial cyber-physical systems (ICPS) will accelerate the transformation of offline data-driven modeling to fast computation services, such as computation pipelines for prediction, monitoring, prognosis, diagnosis, and control in factories. However, it is computationally intensive to adapt computation pipelines to heterogeneous contexts in ICPS in manufacturing. In this article, we propose to rank and select the best computation pipelines to match contexts and formulate the problem as a recommendation problem. The proposed method adaptive computation pipelines (AdaPipe) considers similarities of computation pipelines from word embedding, and features of contexts. Thus, without exploring all computation pipelines extensively in a trial-and-error manner, AdaPipe efficiently identifies top-ranked computation pipelines. We validated the proposed method with 60 bootstrapped datasets from three real manufacturing processes: thermal spray coating, printed electronics, and additive manufacturing. The results indicate that the proposed recommendation method outperforms the traditional matrix completion, tensor regression methods, and a state-of-the-art personalized recommendation model.

A study on a general cyber machine tools monitoring system in smart factories

The massive data produced in real manufacturing processes provides a micro observation mechanism for managers and producers, which helps to analyze and grasp the manufacturing details, to form a macro decision-making mechanism, and finally to improve the manufacturing quality and production efficiency. This paper deeply investigates the real-time acquisition method of machine tool processing data and presents a general method of data analysis in different actual application scenarios to solve the general problem of insufficient and inappropriate utilization of machine tools. A framework of general machine tool monitoring system with four layers is proposed and several key stumbling blocks are researched, such as heterogeneous machine data acquisition, data analysis, and related application. These achievements are applied in a general cyber machine tools monitoring system based on MTConnect, which has greatly improved the utilization of machine tools and become a valuable unit in the framework of intelligent manufacturing and smart factory.

Development of the post-form strength prediction model for a high-strength 6xxx aluminium alloy with pre-existing precipitates and residual dislocations

The applications of lightweight and high strength sheet aluminium alloys are increasing rapidly in the automotive industry due to the expanding global demand in this industrial cluster. Accurate prediction of the post-form strength and the microstructural evolutions of structural components made of Al-alloys has been a challenge, especially when the material undergoes complex processes involving ultra-fast heating and high temperature deformation, followed by multi-stage artificial ageing treatment. In this research, the effects of pre-existing precipitates induced during ultra-fast heating and residual dislocations generated through high temperature deformation on precipitation hardening behaviour have been investigated. A mechanism-based post-form strength (PFS) prediction model, incorporating the flow stress model and age-hardening model, was developed ab-initio to predict strength evolution during the whole process. To model the stress-strain viscoplastic behaviour and represent the evolution of dislocation density of the material in forming process, constitutive models were proposed and the related equations were formulated. The effect of pre-existing precipitates was considered in the age-hardening model via introducing the complex correlations of microstructural variables into the model. In addition, an alternative time-equivalent method was developed to link the different stages of ageing and hence the prediction of precipitation behaviours in multi-stage ageing was performed. Furthermore, forming tests of a U-shaped component were performed to verify the model. It was found that the model is able to accurately predict the post-form strength with excellent agreement with deviation of less than 5% when extensively validated by experimental data. Therefore, the model is considered to be competent for predicting the pre-empting material response as well as a powerful tool for optimising forming parameters to exploit age hardening to its maximum potential in real manufacturing processes.

MANUFACTURE OF SOIL-CEMENT BRICKS WITH THE ADDITION OF SUGARCANE BAGASSE ASH

ABSTRACT The objective of this study is to evaluate the effect of sugarcane bagasse ash on the compressive strength and water absorption index of soil-cement bricks. The ash was sieved without chemical and/or physical treatment to reflect the use of this material in real manufacturing processes. The addition of natural ash decreased the quality of the bricks, reducing the compressive strength and increasing the water absorption index.

Robust design of a VP-NCS chart for joint monitoring mean and variability in series systems under maintenance policy

Abstract Statistical process monitoring and maintenance policy are two important tools to improve system reliability and products quality. To simplify mathematical modeling, most existing papers in the literature suffer from four major drawbacks: (1) designing the control charts under maintenance action for the single-unit systems, in spite of the fact that most real manufacturing processes include multiple units; (2) using control charts with fixed parameters that are slow in detecting small and moderate shifts; (3) this assumption that the occurrence of assignable cause only changes the process mean, which means it doesn’t have any effect on variance of quality characteristic, and (4) emplying the point estimates of input parameters while the precise estimation of them are unavailable in the real industrial situations due to the existence of uncertainty. To overcome the mentioned disadvantages, this study integrates the economic-statistical design of an adaptive non-central chi-square chart with maintenance policy for joint monitoring mean and variability of two-unit series systems. In addition, a robust optimization method is presented to minimize model costs under uncertain parameters. Finally, a case study is applied to indicate the model efficiency. The results confirm that the suggested control chart has better performance in cost-saving in comparison with the other control charts. Furthermore, the results demonstrate that when the number of uncertain parameters increases, the model cost increases obviously.

Development of batch proportional-integral-derivative controller

Previous batch control methods, such as iterative learning control (ILC) or run-to-run (R2R) control, can significantly improve the control performance of the batch process. However, to guarantee the expected good control performance, a fairly accurate process model is required for these controllers. Also, the implementation is numerically complicated so that it is difficult to be applied to real manufacturing processes. To overcome these problems, a new batch proportional-integral-derivative (PID) control method is proposed, which borrows the concept of the conventional PID control method. Simulation studies confirm that the proposed method shows acceptable performance in tracking a setpoint trajectory, rejecting disturbances, and robustness to noises and variation of process dynamics. The application to the commercial batch process of a single crystal grower verifies that the proposed method can significantly contribute to improving the control performances of real batch processes.

Managerial model of business performance in industry 4.0 concept concentrated on effective manufacturing process

A mutual combination of various managerial conceptions claims a lot of new opportunities for production managers, such as how to influence effectiveness and efficiency of flexible production flows based on the industry 4.0 concept. This paper is concentrated on the research of actual conditions, active influencing real manufacturing processes and their special process stability as a basic assumption for flexible implementation of industry 4.0 in industrial enterprises. In the real production process, the horizontal managerial concept is confronted with the vertical (digitised) manufacturing concept. The results presented in this paper are based on the research and survey realised in various industrial companies. Keywords: Process, model, industry 4.0, flexibility.

Discrete Event Simulation Method as a Tool for Improvement of Manufacturing Systems

The problem of production flow in manufacturing systems is analyzed. The machines can be operated by workers or by robots, since breakdowns and human factors destabilize the production processes that robots are preferred to perform. The problem is how to determine the real difference in work efficiency between humans and robots. We present an analysis of the production efficiency and reliability of the press shop lines operated by human operators or industrial robots. This is a problem from the field of Operations Research for which the Discrete Event Simulation (DES) method has been used. Three models have been developed, including the manufacturing line before and after robotization, taking into account stochastic parameters of availability and reliability of the machines, operators, and robots. We apply the OEE (Overall Equipment Effectiveness) indicator to present how the availability, reliability, and quality parameters influence the performance of the workstations, especially in the short run and in the long run. In addition, the stability of the simulation model was analyzed. This approach enables a better representation of real manufacturing processes.

Optimal design of a synthetic chart for monitoring process dispersion with unknown in-control variance

For a stable manufacturing process, quality problems are often caused by changes in process dispersion. Although there have been plenty of research on the monitoring of process dispersion, the existing studies of synthetic charts for monitoring process dispersion only focus on the upward shift monitoring. However, the decrease shift monitoring is also necessary and important. In this paper, a synthetic S2 chart is proposed to simultaneously monitor both upward and downward shifts and it consists of a Shewhart-type two-sided S2 sub-chart and a conforming run length sub-chart. In the known in-control variance case, the conforming run length sub-chart only needs a lower control limit and the proposed synthetic S2 chart is shown to be average run length (ARL) unbiased. The effect of parameter estimation on the proposed synthetic S2 chart is also investigated as it is an important issue especially in the real manufacturing processes. Considering that the in-control variance is usually unknown and needs to be estimated by Phase I samples in practice, a new synthetic S2 chart in which conforming run length sub-chart also only needs a lower control limit is developed when the in-control variance is estimated. Furthermore, optimal designs for both known and unknown parameter cases are studied. The advantage of the proposed chart in performance is shown in the results of the comparison with the ARL-unbiased S2 chart. Also, an example illustrates the construction and application procedure of this proposed chart.