Flexible Machine Research Articles

Energy forecasting with robust, flexible, and explainable machine learning algorithms

AbstractEnergy forecasting is crucial in scheduling and planning future electric load, so as to improve the reliability and safeness of the power grid. Despite recent developments of forecasting algorithms in the machine learning community, there is a lack of general and advanced algorithms specifically considering requirements from the power industry perspective. In this paper, we present eForecaster, a unified AI platform including robust, flexible, and explainable machine learning algorithms for diversified energy forecasting applications. Since October 2021, multiple commercial bus load, system load, and renewable energy forecasting systems built upon eForecaster have been deployed in seven provinces of China. The deployed systems consistently reduce the average Mean Absolute Error (MAE) by 39.8% to 77.0%, with reduced manual work and explainable guidance. In particular, eForecaster also integrates multiple interpretation methods to uncover the working mechanism of the predictive models, which significantly improves forecasts adoption and user satisfaction.

A method for understanding and digitizing manipulation activities using programming by demonstration in robotic applications

Robots are flexible machines, where the flexibility is achieved, mainly, by the re-programming of the robotic system. To fully exploit the potential of robotic systems, an easy, fast, and intuitive programming methodology is desired. By applying such methodology, robots will be open to a wider audience of potential users (i.e. SMEs, etc.) since the need for a robotic expert in charge of programming the robot will not be needed anymore. This paper presents a Programming by Demonstration approach dealing with high-level tasks taking advantage of the ROS standard. The system identifies the different processes associated to a single-arm human manipulation activity and generates an action plan for future interpretation by the robot. The system is composed of five modules, all of them containerized and interconnected by ROS. Three of these modules are in charge of processing the manipulation data gathered by the sensors system, and converting it from the lowest level to the highest manipulation processes. In order to do this transformation, a module is used to train the system. This module generates, for each operation, an Optimized Multiorder Multivariate Markov Model, that later will be used for the operations recognition and process segmentation. Finally, the fifth module is used to interface and calibrate the system. The system was implemented and tested using a dataglove and a hand position tracker to capture the operator’s data during the manipulation. Four users and five different object types were used to train and test the system both for operations recognition and process segmentation and classification, including also the detection of the locations where the operations are performed.

Integrating a tailored recurrent neural network with Bayesian experimental design to optimize microbial community functions.

Microbiomes interact dynamically with their environment to perform exploitable functions such as production of valuable metabolites and degradation of toxic metabolites for a wide range of applications in human health, agriculture, and environmental cleanup. Developing computational models to predict the key bacterial species and environmental factors to build and optimize such functions are crucial to accelerate microbial community engineering. However, there is an unknown web of interactions that determine the highly complex and dynamic behavior of these systems, which precludes the development of models based on known mechanisms. By contrast, entirely data-driven machine learning models can produce physically unrealistic predictions and often require significant amounts of experimental data to learn system behavior. We develop a physically-constrained recurrent neural network that preserves model flexibility but is constrained to produce physically consistent predictions and show that it can outperform existing machine learning methods in the prediction of certain experimentally measured species abundance and metabolite concentrations. Further, we present a closed-loop, Bayesian experimental design algorithm to guide data collection by selecting experimental conditions that simultaneously maximize information gain and target microbial community functions. Using a bioreactor case study, we demonstrate how the proposed framework can be used to efficiently navigate a large design space to identify optimal operating conditions. The proposed methodology offers a flexible machine learning approach specifically tailored to optimize microbiome target functions through the sequential design of informative experiments that seek to explore and exploit community functions.

Treating Semiempirical Hamiltonians as Flexible Machine Learning Models Yields Accurate and Interpretable Results.

Quantum chemistry provides chemists with invaluable information, but the high computational cost limits the size and type of systems that can be studied. Machine learning (ML) has emerged as a means to dramatically lower the cost while maintaining high accuracy. However, ML models often sacrifice interpretability by using components such as the artificial neural networks of deep learning that function as black boxes. These components impart the flexibility needed to learn from large volumes of data but make it difficult to gain insight into the physical or chemical basis for the predictions. Here, we demonstrate that semiempirical quantum chemical (SEQC) models can learn from large volumes of data without sacrificing interpretability. The SEQC model is that of density-functional-based tight binding (DFTB) with fixed atomic orbital energies and interactions that are one-dimensional functions of the interatomic distance. This model is trained to ab initio data in a manner that is analogous to that used to train deep learning models. Using benchmarks that reflect the accuracy of the training data, we show that the resulting model maintains a physically reasonable functional form while achieving an accuracy, relative to coupled cluster energies with a complete basis set extrapolation (CCSD(T)*/CBS), that is comparable to that of density functional theory (DFT). This suggests that trained SEQC models can achieve a low computational cost and high accuracy without sacrificing interpretability. Use of a physically motivated model form also substantially reduces the amount of ab initio data needed to train the model compared to that required for deep learning models.

Optimal Framework for Level Based Access Control for VM Auditing on Cloud

The growth in the cloud computing have motivated and enable lot of application developer to deploy the applications on cloud. The major challenge of hosting on cloud is the service provider or the application provider must comply to a good number of rules. These compliance reports are time to time validated and checked by external auditors. The auditing process for the cloud services are critical and the access controls must be enabled. Due to the higher complexity and less flexibility of the virtual machines, most of the cases this access control mechanism is compromised. This work proposes four algorithms to identify and enhance the LBAC mechanism for cloud services with access updates based on time variant characteristics analysis and predictive analysis with selective cryptographic methods. The proposed model produces significantly improved results to overcome three major issues in the cloud service management as selective LBAC, static privileges and open access control for the auditors.

Research on Optimization Algorithm of AGV Scheduling for Intelligent Manufacturing Company: Taking the Machining Shop as an Example

Intelligent manufacturing workshop uses automatic guided vehicles as an important logistics and transportation carrier, and most of the existing research adopts the intelligent manufacturing workshop layout and Automated Guided Vehicle (AGV) path step-by-step optimization, which leads to problems such as low AGV operation efficiency and inability to achieve the optimal layout. For this reason, a smart manufacturing assembly line layout optimization model considering AGV path planning with the objective of minimizing the amount of material flow and the shortest AGV path is designed for the machining shop of a discrete manufacturing enterprise of a smart manufacturing company. Firstly, the information of the current node, the next node and the target node is added to the heuristic information, and the dynamic adjustment factor is added to make the heuristic information guiding in the early stage and the pheromone guiding in the later stage of iteration; secondly, the Laplace distribution is introduced to regulate the volatilization of the pheromone in the pheromone updating of the ant colony algorithm, which speeds up the speed of convergence; the path obtained by the ant colony algorithm is subjected to the deletion of the bi-directional redundant nodes, which enhances the path smoothing degree; and finally, the improved ant colony algorithm is fused with the improved dynamic window algorithm, so as to enable the robots to arrive at the end point safely. Simulation shows that in the same map environment, the ant colony algorithm compared with the basic ant colony algorithm reduces the path length by 40% to 67% compared to the basic ant colony algorithm and reduces the path inflection points by 34% to 60%, which is more suitable for complex environments. It also verifies the feasibility and superiority of the conflict-free path optimization strategy in solving the production scheduling problem of the flexible machining operation shop.

Untapping Industrial Flexibility via Waste Heat-Driven Pumped Thermal Energy Storage Systems

Pumped thermal energy storage (PTES) is a promising long-duration energy storage technology. Nevertheless, PTES shows intermediate round-trip efficiency (RTE—0.5 ÷ 0.7) and significant CAPEX. sCO2 heat pumps and power cycles could reduce PTES CAPEX, particularly via reversible and flexible machines. Furthermore, the possibility to exploit freely available heat sources (such as waste heat and/or CSP inputs) could increase RTE, making the system capable of an apparent RTE > 100% as well as reducing CAPEX, avoiding the need for two TES systems. This paper analyses the potential valorization of industrial waste heat (WH) to enhance PTES thermodynamic performance as well as increase industrial energy efficiency, valorizing different levels of WH sources in the 100–400 °C temperature range. In fact, the use of additional heat, otherwise dumped into ambient surroundings, may contribute to avoiding the need for a second TES, thus enhancing plant competitiveness. Starting from an assessment of the most relevant industrial sectors to apply the proposed solution (looking at available WH and electric flexibility needed), this paper analyses the feasibility of a specific sCO2-based PTES case study, where the cycle is integrated into a cement production plant with a WH temperature of around 350 °C. It is demonstrated that the CAPEX of the proposed systems are still relevant and only a robust exploitation of the PTES in the ancillary service market could attract industrial customers’ interest in sCO2 PTES.

A Low-Power ECG Processor ASIC Based on an Artificial Neural Network for Arrhythmia Detection

The early detection of arrhythmia can effectively reduce the risk of serious heart diseases and save time for treatment. Many healthcare devices have been widely used for electrocardiogram (ECG) monitoring. However, most of them can only complete simple two-classes detection and have unacceptable hardware overhead and energy consumption. For achieving accurate and low-power arrhythmia detection, a novel ECG processor application specific integrated circuit (ASIC) is proposed in this paper, which can perform the prediction of five types of cardiac arrhythmias and heart rate monitoring. To realize hardware-efficient R-peak detection, an ECG pre-processing engine based on a first derivative and moving average comparison method is proposed. Efficient arrhythmia detection is realized by the proposed low-power classification engine, which is based on a carefully designed lightweight artificial neural network (ANN) with good prediction accuracy. The hardware reuse strategy is used to implement the hardware logic of ANN, where computations are executed by only one processing unit (PU), which is controlled by a flexible finite state machine (FSM). Also, the weights of ANN are configurable to facilitate model updates. We validate the functionality of the design using real-world ECG data. The proposed ECG processor is implemented using 55 nm CMOS technology, occupying an area of 0.33 mm2. This design consumes 12.88 μW at a 100 kHz clock frequency, achieving a classification accuracy of 96.69%. The comparison results with previous work indicate that our design has advantages in detection performance and power consumption, providing a good solution for low-power and low-cost ECG monitoring.

Video Fire Detection Methods Based on Deep Learning: Datasets, Methods, and Future Directions

Among various calamities, conflagrations stand out as one of the most-prevalent and -menacing adversities, posing significant perils to public safety and societal progress. Traditional fire-detection systems primarily rely on sensor-based detection techniques, which have inherent limitations in accurately and promptly detecting fires, especially in complex environments. In recent years, with the advancement of computer vision technology, video-oriented fire detection techniques, owing to their non-contact sensing, adaptability to diverse environments, and comprehensive information acquisition, have progressively emerged as a novel solution. However, approaches based on handcrafted feature extraction struggle to cope with variations in smoke or flame caused by different combustibles, lighting conditions, and other factors. As a powerful and flexible machine learning framework, deep learning has demonstrated significant advantages in video fire detection. This paper summarizes deep-learning-based video-fire-detection methods, focusing on recent advances in deep learning approaches and commonly used datasets for fire recognition, fire object detection, and fire segmentation. Furthermore, this paper provides a review and outlook on the development prospects of this field.

A Pareto-based two-stage evolutionary algorithm for flexible job shop scheduling problem with worker cooperation flexibility

The previous studies on the flexible job shop scheduling problems (FJSP) with machine flexibility and worker flexibility normally assume that each machine is operated by one worker at any time. However, it is not accurate in many cases because many workers may be required for machines in processing complex operations. Hence, this paper studies a universal version, i.e., FJSP with worker cooperation flexibility (FJSPWC), which defines that each machine can be used only if their required workers are prepared. A mixed-integer linear programming model tuned by CPLEX is established for the problem aiming to collaboratively minimize the makespan, maximum workload of machines and maximum workload of workers. To solve the problem efficiently, a Pareto-based two-stage evolutionary algorithm (PTEA) is proposed. In the PTEA, a well-tailored initialization operator and the NSGA-II structure are designed for global exploration in the first stage, and a competitive objective-based local search operator is developed to improve its local search ability and accelerate the convergence in the second stage. Extensive experiments based on fifty-eight newly formulated benchmarks are carried out to validate the effectiveness of the well-designed initialization operator and two-stage architecture. Comprehensive experiments are performed to evaluate the proposed PTEA, and the results reveal that the PTEA is superior to four comparison algorithms concerning the distribution, convergence, and overall performance.

Flexible Pressure Sensors and Machine Learning Algorithms for Human Walking Phase Monitoring.

Soft sensors are attracting much attention from researchers worldwide due to their versatility in practical projects. There are already many applications of soft sensors in aspects of life, consisting of human-robot interfaces, flexible electronics, medical monitoring, and healthcare. However, most of these studies have focused on a specific area, such as fabrication, data analysis, or experimentation. This approach can lead to challenges regarding the reliability, accuracy, or connectivity of the components. Therefore, there is a pressing need to consider the sensor's placement in an overall system and find ways to maximize the efficiency of such flexible sensors. This paper proposes a fabrication method for soft capacitive pressure sensors with spacer fabric, conductive inks, and encapsulation glue. The sensor exhibits a good sensitivity of 0.04 kPa-1, a fast recovery time of 7 milliseconds, and stability of 10,000 cycles. We also evaluate how to connect the sensor to other traditional sensors or hardware components. Some machine learning models are applied to these built-in soft sensors. As expected, the embedded wearables achieve a high accuracy of 96% when recognizing human walking phases.

HiCognition: a visual exploration and hypothesis testing tool for 3D genomics

Genome browsers facilitate integrated analysis of multiple genomics datasets yet visualize only a few regions at a time and lack statistical functions for extracting meaningful information. We present HiCognition, a visual exploration and machine-learning tool based on a new genomic region set concept, enabling detection of patterns and associations between 3D chromosome conformation and collections of 1D genomics profiles of any type. By revealing how transcription and cohesion subunit isoforms contribute to chromosome conformation, we showcase how the flexible user interface and machine learning tools of HiCognition help to understand the relationship between the structure and function of the genome.

Efficient and flexible mediation analysis with time-varying mediators, treatments, and confounders

AbstractUnderstanding the mechanisms of action of interventions is a major general goal of scientific inquiry. The collection of statistical methods that use data to achieve this goal is referred to asmediation analysis. Natural direct and indirect effects provide a definition of mediation that matches scientific intuition, but they are not identified in the presence of time-varying confounding. Interventional effects have been proposed as a solution to this problem, but existing estimation methods are limited to assuming simple (e.g., linear) and unrealistic relations between the mediators, treatments, and confounders. We present an identification result for interventional effects in a general longitudinal data structure that allows flexibility in the specification of treatment-outcome, treatment-mediator, and mediator-outcome relationships. Identification is achieved under the standard no-unmeasured-confounders and positivity assumptions. In this article, we study semi-parametric efficiency theory for the functional identifying the mediation parameter, including the non-parametric efficiency bound, and was used to propose non-parametrically efficient estimators. Implementation of our estimators only relies on the availability of regression algorithms, and the estimators in a general framework that allows the analyst to use arbitrary regression machinery were developed. The estimators are doubly robust,n\sqrt{n}-consistent, asymptotically Gaussian, under slow convergence rates for the regression algorithms used. This allows the use of flexible machine learning for regression while permitting uncertainty quantification through confidence intervals andpp-values. A free and open-sourceRpackage implementing the methods is available on GitHub. The proposed estimator to a motivating example from a trial of two medications for opioid-use disorder was applied, where we estimate the extent to which differences between the two treatments on risk of opioid use are mediated by craving symptoms.

A novel approach to predicting customer lifetime value in B2B SaaS companies

AbstractIn this paper, we propose a flexible machine learning framework to predict customer lifetime value (CLV) in the Business-to-Business (B2B) Software-as-a-Service (SaaS) setting. The substantive and modeling challenges that surface in this context relate to more nuanced customer relationships, highly heterogeneous populations, multiple product offerings, and temporal data constraints. To tackle these issues, we treat the CLV estimation as a lump sum prediction problem across multiple products and develop a hierarchical ensembled CLV model. Lump sum prediction enables the use of a wide range of supervised machine learning techniques, which provide additional flexibility, richer features and exhibit an improvement over more conventional forecasting methods. The hierarchical approach is well suited to constrained temporal data and a customer segment model ensembling strategy is introduced as a hyperparameter model-tuning step. The proposed model framework is implemented on data from a B2B SaaS company and empirical results demonstrate its advantages in tackling a practical CLV prediction problem over simpler heuristics and traditional CLV approaches. Finally, several business applications are described where CLV predictions are employed to optimize marketing spend, ROI, and drive critical managerial insights in this context.

EForecaster: Unifying Electricity Forecasting with Robust, Flexible, and Explainable Machine Learning Algorithms

eForecaster: Unifying Electricity Forecasting with Robust, Flexible, and Explainable Machine Learning Algorithms

Deep reinforcement learning for optimal planning of assembly line maintenance

Discovering the optimal maintenance planning strategy can have a substantial impact on production efficiency, yet this aspect is often overlooked in favor of production planning. This is a missed opportunity as maintenance and production activities are deeply intertwined. Our study sheds light on the significance of maintenance planning, particularly in the dynamic setting of an assembly line. By maximizing the average production rate and incorporating flexible planning windows, buffer content, and machine production states, a unique problem is addressed in which a policy for planning maintenance on the final machine of a serial assembly line is developed. To achieve this, novel average-reward deep reinforcement learning techniques are employed and pitted against generic dispatching methods. Using a digital twin with real-world data, experiments demonstrate the immense potential of this new deep reinforcement learning technique, producing policies that outperform generic dispatching strategies and practitioner policies.

A multi-objective joint optimisation method for simultaneous part family formation and configuration design in delayed reconfigurable manufacturing system (D-RMS)

ABSTRACT In the era of Industry 4.0, the demand fluctuation has become fiercer due to the characteristics of diversification, customisation, and uncertainty. Reconfigurability of manufacturing systems has been proven to be a useful and necessary feature when it comes to handling demand uncertainty. This feature can be achieved through the implementation of reconfigurable manufacturing system (RMS) and delayed reconfigurable manufacturing system (D-RMS). D-RMS is a subclass of RMS that focuses primarily on improving the convertibility of the manufacturing system. The two main phases involved in implementing D-RMS are part family formation and configuration design. Therefore, we proposed a multi-objective joint optimisation method of part family formation and configuration design according to the philosophy of D-RMS. Firstly, we develop a multi-objective joint optimisation model that takes into account investment cost, reconfiguration cost, similarity coefficient, and delayed reconfiguration to optimise the part family and configuration of D-RMS simultaneously. Three types of machine tools namely dedicated machine tools, flexible machine tools, and reconfigurable machine tools are considered in the optimisation model. Secondly, the non-dominated sorting genetic algorithm-III (NSGA-III) is adopted to solve the proposed multi-objective integer programming problem. Finally, numerical experiments are presented to demonstrate the effectiveness of the proposed multi-objective joint optimisation method.

Labour-saving detection of hybrid rice rows at the pollination stage based on a multi-perturbed semi-supervised model

Autonomous navigation of the hybrid rice pollination process is essential to increasing seed production. Detecting crop rows in real-time with flexible and cost-effective machine vision equipment is crucial to achieving autonomous navigation. The use of deep learning-based models has proven effective in crop row detection. However, training such models is highly dependent on manually finely annotated image data, which becomes a bottleneck that hinders their practical application and improvement of crop row detection accuracy. This study proposed a multi-perturbed semi-supervised learning-based model to enhance the semantic segmentation performance of hybrid rice regions by mining valuable information from easily accessible unlabelled images. The input data perturbation and network perturbation developed effectively alleviated the overfitting and teacher-student weight coupling issues when applying semi-supervised models to the hybrid rice image dataset with high content similarity. Navigation centrelines were obtained by post-processing the segmentation masks of the crop regions. Under a 1/2 partition protocol of labelled and unlabelled images, the semantic segmentation performance of the proposed approach achieved a mean intersection over union (mIoU) of 0.887, outperformed its supervised baseline (DeepLabv3+ ) and original model (mean teacher) by 1.8% and 4.1%, respectively. 82.38% of the crop row were accurately detected using the proposed model, demonstrating an improvement of 3.04% and 14.06% over its baseline and the original model, respectively.

Case study of vibration in a mobile milling machine

Many industries rely on heavy-duty machinery such as cranes, pipelines, motors, which are subject to wear and tear. Milling of such parts using conventional machine tools presents many drawbacks and is impossible in some cases. In such situations, mobile machine tools can be used to machine directly on site. However, mobile machines are flexible, and are prone to detrimental vibrations during cutting operations. Vibrational stability in the surfacing process has been extensively studied in conventional milling. However, the focus of this article is an original study on a mobile milling machine used for on-site surfacing operations. An experimental investigation into the vibrational behavior of the machine is carried out and a sensorless method is used to detect the onset of chatter. Machine flexibility is discussed using a tool-point frequency response function and references from the literature. Milling operations are carried out over a range machining parameter to determine an experimental stability limit. Internal sensors for spindle motor consumption and spindle motor encoder are monitored during the machining process. A threshold is established for determining the appearance of chatter using only the machine’s internal sensors.

Sensor information‐based crop recommendation system using machine learning for the fertile regions of Maharashtra

SummaryAgriculture is the major backbone of India. Therefore, the crop recommendation system is important for the farmers and the country as it reflects the country's economic growth. The crop yield rate is affected due to various parameters such as climatic changes, soil properties, temperature, humidity and so forth. An effective prediction system is required to monitor the field and suggest suitable crops that can provide a maximum yield rate. Therefore, a prediction system is developed in the proposed framework for crop recommendation using the sensor information collected from Maharashtra, India. The dataset has been built with the information collected by 250 sensors located in different Maharashtra places. Initially, the gathered dataset is subjected to preprocessing steps like data cleaning, removing duplicate values, and filling up the missing values. Then, robust and flexible machine learning models like decision tree, random forest, and support vector machine are used, which analyze the preprocessed data and predict the suitable crops that show high yield for a particular area. The implementation of the proposed framework is done using Python. The RF classifier achieved the highest accuracy rate of 95% among the three classifiers.