In the increasingly volatile and competitive environment of the mobility industry, introducing new innovative products faster than competitors forms a key success factor for OEMs. To increase strategic flexibility, it is critical to speed up the development process while at the same time reducing the cost per vehicle project. Despite the increasing use of virtual methods in product development and production system design, hardware prototypes are indispensable for the series development today and will remain necessary in the foreseeable future. As these prototypes are very costly and their production tedious, companies are trying to reduce the number of pre-series vehicles required and optimize their production. Therefore, the authors developed a worker assistance system specifically designed for the pre-series production, that aims at increasing efficiency and maximizing the benefit for the series development. This paper reports on a case study that evaluated the application of this assistance system in a real industry environment. The results show that the assessed approach offers a significant potential to speed up the prototype assembly (up to 40% in the examined case), while at the same time contributing to the series assembly system development by systematically testing, correcting and evaluating planned series procedures.
The aim of the paper is to review the trends on low-cost open-source Internet of Thing (IoT) in the Framework of Industry 4.0. We hypothesize that open source IoTs benefits small and medium sized (SME) manufacturing companies to help them tackle economical and technical barriers for technology adoption at factory floors. This research reviews the state of the art of open-source hardware and software and identifies the challenges of open-source IoT by compiling scientific literature and relevant online resources. The methodology is twofold: (i) we use a grey literature review including online information, and (ii) we complement the analysis by a structured keyword search using IEEE scientific repository. The study finds that open source is and will be an essential part of developing Industry 4.0 solutions and its integral constituent, Industrial Internet of Things systems (IIoT). However, there are challenges to overcome, such as interoperability and reliability.
Typically, before constructing an object with an additive manufacturing system, the 3D object must be sent through a process called slicing. Slicing converts a 3D object commonly in the form of an STL file into a set of layers by horizontally intersecting a plane with the object at various heights. At each height, called a layer, multiple 2D polygons can be generated. Each polygon represents a boundary for solid geometry and is called an island. Each island is then comprised of multiple path types in an attempt to optimally fill the polygon. To move between each island and each islands’ paths, travels are inserted. Travels are simply motion by the system to move from one area of construction to another. Travels do not contribute to the construction of the object, and so, are considered wasted motion. In large-scale additive manufacturing, objects can be quite large and the distance between islands can be large as well. As a result, these travels can waste a significant amount of time. Ideally, travels would be as short as possible, however, computing global minimal travel paths is computationally expensive. To combat this problem, researchers at Oak Ridge National Lab developed a GPU-based approach to travel insertion based on a unique factoradic representation. This representation was then utilized by the GPU to solve the Traveling Salesman Problem (TSP). This algorithm was able to compute global minimal travel paths quickly resulting in faster object construction. A general investigation was also carried out to determine when a GPU vs CPU implementation would be beneficial.
The goal of this work was to determine which standard particle size metric derived from optical analysis most closely approximates the sieved weight percent of irregularly shaped powder intended to be used for directed energy deposition. In this investigation, equivalent circle diameter, maximum diameter, minimum diameter, and perimeter were used as metrics to “virtually sieve” the particles in samples of irregularly shaped powder into the following particle size bins: <45 µm, 45 µm – 150 µm, and >150 µm. The percentage in the 45 µm – 150 µm bin were then compared to the weight percent of the powder mechanically sieved into this size range. The absolute difference between the virtually sieved percentage and the mechanically sieved percentage was assessed for 81 samples of mechanically-generated stainless steel 316L powder, all produced under different processing conditions in an oscillation ball mill. This difference was found to be on average, the least with the minimum diameter assessed as an area percentage, followed by the equivalent diameter assessed as an area percentage, and then the maximum diameter assessed as a percentage of the total number of particles ranked third. These findings and the methodology used to obtain them may be used by powder production process engineers and quality assurance personnel to assist in process control as more diverse additive manufacturing feedstocks become utilized.
Although Metal Powder Bed Fusion Additive Manufacturing (PBFAM) process has emerged as an important industrial process, part distortion due to repeated heating and cooling is a major barrier that stands in the way of using this process for mass commercial production. Laser hatch pattern is a critical factor that influences the part distortion and has been widely studied through experimental investigations. In this paper, we describe an approach for hatch pattern optimization to minimize part GD&T errors using Genetic algorithms (GA). A pre-trained Artificial Neural Network (ANN) was adopted to predict the inherent strain of any hatch pattern, while a Backward Interpolation (BI) approach was used to predict the distortion of the sample part based on the inherent strain obtained from ANN. A Genetic Algorithm approach was then used to optimize the hatch angle of each layer to minimize the flatness form error of a flat feature of a sample part. Three types of island strategies were investigated for laying hatch patterns, and the results show that the increased number of scan islands in the hatch pattern results in less distortion and minimization of flatness error. The flatness values of four other benchmark hatch patterns were also investigated and compared with the optimal hatch pattern results. The comparison showed that the flatness of the sample part with the optimized hatch pattern performs better than the benchmark hatch patterns. The overall computational time for hatch pattern optimization was found to be reasonable, considering the large number of distortion simulations performed during the optimization process.
Additive Manufacturing technologies enable the fabrication of structures with multiscale complexities without increasing their cost. To take advantage of the unique design freedom enabled by additive manufacturing processes, this paper reports a new design method that can generate a fully customized porous shoe sole. The proposed design method contains five major steps and can generate a fully customized shoe sole with customized features on both macro and mesoscale. Specifically, compared to a conventional flat shoe sole, the top surface of a customized sole can fully conform to the bottom surface of patient’s feet, which can significantly reduce the peak plantar pressure. In addition to that, the strut diameters of designed lattice structures can also be customized based on the proposed data-driven design optimization algorithm. By varying the struts’ diameters on the different regions of the designed lattice shoe sole, the peak plantar pressure can be further minimized. The case study provided in this paper shows the proposed method that can significantly improve the performance of customized shoe sole. This promising result indicates that additive manufacturing fabricated lattice shoe soles can be a potential solution to prevent or treat diabetic ulcers.
Advances in digital technologies, amongst others, present process innovation opportunities in manufacturing which if appropriately exploited will increase performance and productivity. Central to successful implementation of process innovation initiatives is adequate preparation during pre-implementation phase and ensuring that the business is ready prior to the deployment of their process innovation initiatives. Akin to digital transformation, key elements of process innovation deployment include people, process, and technology. An understanding of these key elements of process innovation deployment readiness will help towards achieving successful implementation outcome. This paper explores manufacturing process innovation deployment readiness from an extended people, process, and technology framework perspective. The extension adds to the traditional PPT framework the context of the deployment. Attributes of manufacturing process innovation deployment readiness were obtained from the literature and the derived attributes form the basis for discussing the extended PPT view of deployment readiness. It is concluded that failure to either consider or grossly underestimate the role of people, processes, technology, and the context of the deployment will undermine implementations of process innovation in manufacturing.
In-process machining data (e.g., cutting forces and vibrations) have been typically collected and structured as time-referenced measurements (i.e., time-series data) and utilized in this structure to develop statistical data models used in process monitoring and control methods. This paper argues that a time-only-referenced representation overlooks the 3D nature of the physical process generating the data, and that machining data can be represented alternatively as functions of the tool-workpiece relative position resulting in a spatial point cloud data structure. High-density measurements of such spatially refenced data could be highly correlated to surrounding measurements, resulting in spatial correlation structures that could be of physical meaning and value to preserve and leverage. Using a simulated data study, this paper shows that preserving the spatial correlation structure of the data clearly improves the relative modeling performance when utilizing machining data point clouds versus the traditional time-referenced data structure. Specifically, this simulation study investigated the hypothesis that “considering the Gaussian process model class, the best model among all possible models developed using the spatial point cloud data structure has smaller/equal modeling and prediction errors compared to the best model among all possible models developed using the time-referenced data structure.” While this investigation was limited to considering the case of stationary isotropic processes, it demonstrated that the performance gap was relatively large. This encourages further investigations using real-world data to better understand the types of spatial correlations that exist in machining data and the specific machining regimes and process variables that would benefit the most from the spatial point cloud representation of the data.
Laser shock peening (LSP) is a non-contact surface treatment method that has been experimentally found to help increase fracture toughness, induce near-surface compressive residual stress and increase hardness in ceramic materials. Numerous experiments, with associated costs and challenges, are needed to identify the application-specific LSP parameters. Physics-based computational models provide a less expensive and more flexible alternative to performing the requisite experiments, yet the trade-off between computational cost and accuracy of different models needs to be considered. In this work, LSP treatment finite element simulations are executed using a calibrated Drucker-Prager (DP) plasticity model combined with a Mie-Grüneisen equation of state as well as Johnson-Cook rate dependence and a damage initiation criterion. Unlike existing Johnson-Holmquist (JH) methods, the calibrated DP constitutive model does not require damage model parameters to be continually updated with deformation, resulting in lower computational time. Near-surface compressive residual stress, predicted by the model, is compared to measurements reported in recent experimental studies. At the lowest pulse energy tested (1 J) an approximate 5% difference exists between the simulated RS and that measured by Raman spectroscopy in the experiments. This difference appears to increase with greater pulse energies. A parametric comparison of computational time for differences in laser pulse energy, constitutive models, and materials is also performed. Results reveal that the demonstrated model at the lowest pulse energy on alumina is up to 37.5% faster, and when applied to silicon carbide, is up to 15.7% faster in comparison to existing JH methods.
Creasing and scoring are methods used in the production of carton blanks in package manufacturing. The objective of this paper was to compare the creasing and scoring processes to investigate the effect on the performance in folding and in manufactured packages. A series of converting experiments and measurements were performed to investigate the differences between the processes. The results show differences between the creased and scored samples. The blanks die-cut with flatbed die cutting machine required the lowest folding force, while the packages manufactured with laser scoring had the highest compression strength. The results indicate that all methods are suitable to be used in the manufacturing of carton blanks.
The five-parameter Johnson-Cook (J-C) material model represents the behavior of a material under extreme mechanical loading, including high temperatures, strains, and strain rates. The goal of this study is to estimate five J-C material parameters and chip thickness jointly for a given set of force components, power, and temperature. The approach uses two neural network models on a dataset simulated by finite element analysis for orthogonal cutting of aluminum 6061-T6. The first model develops a function approximator to predict the force components, power, and temperature using a given set of J-C parameters and chip thickness for aluminum 6061-T6. The second model searches the input space of the first model to estimate the J-C parameter values and chip thickness, given a set of targeted force components, power, and temperature of interest. The performance of both neural network models is evaluated using mean absolute percentage error. The results suggest that the developed neural networks-based approach is capable of estimating multiple J-C parameters and chip thickness that will result in a targeted force components, power, and temperatures of interest, given starting ‘educated guesses’ about these values.
Majority of methods currently used for quality assessment of tissue engineered medical products (TEMPs) are offline and destructive in nature, which is one of the factors impeding the scale up and translation of these technologies. In this study, we investigate quality assessment of TEMP via dielectric impedance spectroscopy (DIS) and supervised machine learning (ML) as a non-destructive alternative that requires minimal human intervention. 3D printed, NaOH-treated polycaprolactone (PCL) scaffolds seeded with human adipose-derived stem cells (hASC), NIH 3T3, MG63, and human chondrocyte cells were assessed via DIS over 4 days of in vitro culture. The results showed that the cell type and duration in culture had a significant effect on the delta permittivity (Δε, an important DIS metric. Five supervised ML algorithms – K Nearest Neighbors (KNN), Logistic Regression, Random Forest Classifiers, Support Vector Machines, and artificial neural network – were then used to analyze the comprehensive structured permittivity datasets to determine their ability to discern between different cell types and culture durations. The KNN algorithm demonstrated the best accuracy (99%). The outcomes of this study demonstrate the approach of using DIS and supervised ML in conjunction for assessment of TEMPs in an automated manufacturing system.
Developments in high degree-of-freedom(DOF) manufacturing processes such as 5-axis machining and additive manufacturing have greatly moderated the design constraint and brought unprecedented manufacturing capability for parts in complex geometry. The advancement in manufacturing processes, at the same time, leads to significant challenges for process planning due to the increasing decision complexity. A method is needed to enable full automated process planning for high DOF manufacturing processes in the foreseeable future. This work focuses on exploring an artificial neural network(ANN) based approach for machining process planning, specifically the toolpath planning for milling operations. The objective of this research is to construct a framework for automated machining process planning that leverages the advancement in ANN methodologies in an attempt to generate an optimized toolpath without any human logic input. In this proposed framework, the voxel model is used as part design and stock geometry representations. An evolving ANN method NeuralEvolution of Augmenting Topologies(NEAT) is applied as the solution algorithm. A prototype implementation of the proposed framework is created and experimented with reasonably simplified machining scenarios and basic part geometries. Initial experiments demonstrate optimistic results supporting the feasibility of creating such an ANN through an evolutionary method to accomplish specific manufacturing requirements on different geometries. The work also revealed that the geometric input is a critical factor for successfully training an ANN model. Further work is needed to encode the part design geometric information as input. Additionally, an improved evolutionary ANN algorithm needs to be created to accelerate the model training.
Proper functioning of rolling element bearings is critical to ensuring reliable and safe power transmission. The ability to automatically recognize fault-related characteristics is key to enabling intelligent bearing fault recognition. While many techniques have been developed, effective bearing fault recognition under non-stationary conditions remains a challenge. In this paper, a hybrid method that integrates generalized demodulation and artificial neural network is presented that has shown to improve the fault recognition accuracy. Based on the modulation characteristics of bearing vibration signals, a phase function is designed, which allows the mapping of the time-varying modulation rotating frequencies and fault characteristic frequencies into constant frequency components in the demodulation spectrums, thereby eliminating the effect of non-stationarity and facilitating physics-based feature extraction. The features are subsequently classified by an artificial neural network for fault recognition. The physical nature of the features provides the basis for the network to generalize well for unseen non-stationary conditions, and the method has shown to outperform a variety of existing bearing fault recognition techniques in experimental evaluations.
This paper describes an uncertainty evaluation for axial location-dependent cutting edge radius and angle values extracted from structured light scans of a variable pitch endmill. Two cases are evaluated: 1) a single scan of the endmill is performed and that scan is manually fit five times to record the cutting edge geometry for all teeth on the endmill; and 2) five scans of the endmill are performed and each scan is manually fit a single time. Both cases therefore include five manual fits. The standard deviations in radius and angle values are used to represent the statistical uncertainty and the two cases are compared. The mean standard deviations in radius are 0.005 mm and 0.007 mm; the mean standard deviations in angle are 0.066 deg and 0.092 deg. As expected, the uncertainties are higher with additional scans. The scan results are then used in a time-domain simulation to predict the cutting force profiles for the variable pitch endmill. It is shown that the two cases provide similar agreement between prediction and measurement due to the small measurement uncertainties.
The reduced cost of implementing pervasive industrial sensing networks enables universities to incorporate these tools in engineering curricula. They provide engineering students from increasingly computerized backgrounds, such as mechanical and automotive engineering, the opportunity to work alongside students from technical schools who bring different skill sets than what students may be used to, synthesize historical data, and drive the sensing system’s physical system design and implementation. This paper outlines this convergent curriculum’s initial implementation stage, including the wireless environmental sensing Internet of Things (IoT) network, focusing on laboratory environmental sensing. Students placing many sensors around the lab and on equipment generates a wealth of real-time and historical data for use in the classroom and provides them a tangible example of learning to measure the world around them. This setup parallels the current varied Industry 4.0 state of the manufacturing industry, where Big Data exists but is underutilized, and where additional sensors and intelligent machine data streams are added each year. Students in each class are given a defined portion of a broader roadmap to a fully instrumented and intelligent laboratory environment. In the first step, student-programmed environmental sensors were placed around the lab and provide temperature, humidity, pressure, and gas mixture measures every five minutes. Classroom use of the aggregated data includes visualizing the laboratory and essential equipment’s current status using a Microsoft PowerBI dashboard and historical data visualization and analysis through trend forecasting and outlier detection in Python JupyterLab notebooks. The IoT system’s installation also provided an infrastructure for further study of future student-designed IoT projects.
Multi-material lattice structures are used in a range of load-bearing applications for multiple conditions including mechanical and thermal loads. Additive manufacturing processes with multi-material capabilities are well suited to manufacture multi-material structures. In this paper, a multi-material topology optimization approach has been presented using variable-density lattice structures where the geometry of the lattice structure is pre-defined. The objective of the proposed topology optimization method is to design lightweight parts with minimized compliance and thermal energy or improve the heat transfer capability. To facilitate that, a novel interpolation scheme based on the stiffness matrices of the lattice structures has been proposed. This interpolation scheme, unlike the traditional Solid Isotropic Material Penalization (SIMP) interpolation, is observed to perform better in terms of approximating the structure’s load-bearing capacity, primarily due to its formulation on the lattice’s stiffness matrices. This cubic Hermite spline-based interpolation scheme makes it amenable for gradient-based optimization methods. A sequential linear programming method has been used to solve the weighted multi-objective optimization model. A Pareto-frontier study has also been carried out to fully characterize the trade-offs between the two objectives – compliance minimization and thermal energy minimization.
Production scheduling faces three challenges, two of which are trade-offs and the third is processing time uncertainty. The two sources of trade-offs are between inconsistent key performance indicators (KPIs), and between the expected return and the risk of KPI portfolios. Given the KPIs of total completion time (TCT) and variance of completion times (VCT) are inconsistent for one-stage production, we propose our trade-off balancing (ToB) heuristics. Based on comprehensive case studies, we show that our ToB heuristics efficiently and effectively balance the trade-offs from these two sources. Daniels and Kouvelis (DK) proposed a scheduling scheme to optimize the worst-case scenarios against processing time uncertainty, and they designed the endpoint product (EP) and endpoint sum (ES) heuristics for robust scheduling accordingly. Using 5 levels of coefficients of variation (CVs) to represent processing time uncertainty, we show that our ToB heuristics are robust as well, and even better than the EP and ES heuristics at high levels of processing time uncertainty. In addition, our ToB heuristics generate undominated solution spaces of KPIs, which provides a solid base in deciding control and specification limits for stochastic process control (SPC). Moreover, based on the normalized deviations from optima, our trade-off balancing scheme can be generalized to balance any inconsistent KPIs.
Additive manufacturing (AM) machines have developed more rapidly than standardized frameworks needed for the qualification of their geometric capabilities. While some manufacturer-specific methods exist to test capabilities and perform some calibration tasks, standardization efforts have only recently been undertaken in the form of ISO/ASTM 52902. In this study, the recommended methodology prescribed by the standard was implemented by building geometric artifacts with a laser powder bed fusion (LPBF) system and performing dimensional inspection with a coordinate measurement machine (CMM), amongst other methods. Typical dimensional capabilities of the LPBF system are identified and commentary is made on applying metrology methods, detecting geometric error, and diagnosing base causes in the LPBF system. In doing so, favored metrology practices and measurement analysis methods auxiliary to the standard are proposed. Artifact measurements were used to characterize beam positioning error and beam offset error. Methods for decoupling the effects of error sources are proposed. Difficulties in the inspection of AM components are identified, and the effects of various CMM measurement strategies are evaluated. Insights on the application of the new standard are presented, along with commentary as to its fitness for the LPBF process.
Biocompatible polymer fibers have garnered significant interest due to their unique properties. Applications range from absorbent media to tissue engineering and drug delivery products. Many manufacturing processes produce such fibers, but a gap exists in highly scalable processes for fibers loaded with thermolabile additives like pharmaceuticals. This study investigates preliminary process-structure-function relationships of solution blown poly(ethylene oxide) fibers loaded with doxycycline, a drug that has demonstrated antibiotic, anti-inflammatory, and anti-tumoral properties. After parameter screening, a factorial experiment mapped the solution blowing design space with a multi-nozzle apparatus. A 1 mm-thick mat was fabricated comprising doxycycline loaded polymer fibers with a mean diameter of 552 ± 200 nm. Study of release kinetics showed the doxycycline released with a significant burst effect over approximately 1 minute. This study highlights solution blowing as a scalable manufacturing platform for fabricating poly(ethylene oxide) fibers loaded with this impactful drug.
