Timeseries forecasting is applied to many areas of smart factories, including machine health monitoring (MHM), predictive maintenance, and production scheduling. In smart factories, machine speed prediction can be used to dynamically adjust production processes based on different system conditions, optimise production throughput, and minimise energy consumption. However, making accurate data-driven machine speed forecasts is challenging. Given the complex nature of industrial manufacturing processes, predictive models that are robust to noise and can capture the temporal and spatial distributions of the input timeseries are prerequisites for accurate forecasting. Motivated by recent deep learning studies in smart manufacturing, this paper proposes an end-to-end model for multi-step-ahead machine speed prediction. The model, known as 2D-Convolutional LSTM Autoencoder (2DConvLSTMAE), comprises a deep convolutional LSTM (ConvLSTM) encoder-decoder architecture. Extensive empirical analyses using real-world data obtained from a metal packaging plant in the United Kingdom demonstrate the value of the proposed method when compared to state-of-the-art deep learning models.

The growing adoption of decentralised renewable energy generation (such as solar photovoltaic panels and wind turbines) and low-carbon technologies will increase the strain experienced by the distribution networks in the near future. In such a scenario, energy storage is becoming a key alternative to traditional expensive reinforcements to network infrastructure, due to its flexibility, decreasing costs and fast deployment capabilities. In this work, an end-to-end data-driven solution to optimally design the control of a battery unit with the aim of reducing the peak electricity demand is presented. The proposed solution uses state-of-the-art machine learning methods for forecasting electricity demand and PV generation, combined with an optimisation strategy to maximise the use of photovoltaic energy to charge the energy storage unit. To this end, historical demand, weather, and solar energy generation data collected at the Stentaway Primary substation near Plymouth, UK, and at other six locations were employed.

An optimal component feeder arrangement and robotic placement sequence are both important for improving assembly efficiency. Both problems are combinatorial in nature and known to be NP-hard. This article presents a novel discrete hybrid bat-inspired algorithm for solving the feeder slot assignment and placement sequence problem encountered when planning robotic assembly of electronic components. In our method, we use the concepts of swap operators and swap sequence to redefine position and velocity operators from the basic bat algorithm (BA). Furthermore, we propose an improved local search method based on genetic operators of crossover and mutation enhanced by the 2-opt search procedure. The algorithm is formulated with the objective of minimizing the total traveling distance of the pick-and-place device. Through numerical experiments, using a real printed circuit board (PCB) assembly scenario, we demonstrate the considerable effectiveness of the proposed discrete BA to improve the selection of feeder arrangement and placement sequence in PCB assembly operations and achieve high throughput production. The results also highlighted that even though the algorithms outperformed traditional lead through programming techniques, the programmer must consider the influence of different robot motions and speeds.

Smart factories are intelligent, fully-connected and flexible systems that can continuously monitor and analyse data streams from interconnected systems to make decisions and dynamically adapt to new circumstances. The implementation of smart factories represents a leap forward compared to traditional automation. It is underpinned by the deployment of cyberphysical systems that, through the application of Artificial Intelligence, integrate predictive capabilities and foster rapid decision-making. Deep Learning (DL) is a key enabler for the development of smart factories. However, the implementation of DL in smart factories is hindered by its reliance on large amounts of data and extreme computational demand. To address this challenge, Transfer Learning (TL) has been proposed to promote the efficient training of models by enabling the reuse of previously trained models. In this paper, by means of a specific example in aluminium can manufacturing, an empirical study is presented, which demonstrates the potential of TL to achieve fast deployment of scalable and reusable predictive models for Cyber Manufacturing Systems. Through extensive experiments, the value of TL is demonstrated to achieve better generalisation and model performance, especially with limited datasets. This research provides a pragmatic approach towards predictive model building for cyber twins, paving the way towards the realisation of smart factories.

The hot metal silicon content is a key indicator of the thermal state in the blast furnace and it needs to be kept within a pre-defined range in order to ensure efficient operations. Effective monitoring of silicon content is challenging due to the harsh environment in the furnace and irregularly sampled measurements. Data-driven approaches have been proposed in the literature to predict silicon content using process data and overcome the sparsity of silicon content measurements. However, these approaches rely on the selection of hand-crafted features and ad hoc interpolation methods to deal with irregular sampling of the process variables, adding complexity to model training and optimisation, and requiring significant effort when tuning the model over time to keep it to the required level of accuracy. This paper proposes an improved framework for the prediction of silicon content using a novel deep learning approach based on Phased LSTM. The model has been trained using 3 years of data and validated over a 1-year period using a robust walk-forward validation method, therefore providing confidence in the model performance over time. The Phased LSTM model outperforms competing approaches due to its in-built ability to learn from event-based sequences and scalability for real-world deployments. This is the first time that Phased LSTM has been applied to real-world datasets and results suggest that the ability to learn from event-based data can be beneficial for the process industry where event-driven signals from multiple sensors are common.

Width-related defects are a common occurrence in the Hot Strip Mill process which can lead to extra processing, concessions, or scrapping. The detection and Root Cause Analysis of these defects is a largely manual process and is vulnerable to several negative factors including human error, late feedback, and knock-on effects in successive steel strip products. Automated tools which utilize Artificial Intelligence and Machine Learning for defect detection and Root Cause Analysis in hot rolling have not yet been adopted outside of surface defect detection and roller force optimization. In this paper, we propose an automated tool for the detection and Root Cause Analysis of width-related defects in the hot rolling process which utilizes a combination of expert knowledge and several Machine Learning models. Through this, we aim to increase the scope, and encourage further development, of Machine Learning applications within the Hot Strip Mill process. Both classical algorithms and Computer Vision methods were used for the Machine Learning component of the tool, namely, classification trees and pre-trained convolutional neural networks. The tool is trained and validated using data from an existing hot rolling mill and thus the challenges of collecting and processing real-world legacy data are highlighted and discussed. The Machine Learning models used are shown to perform optimally by validation performance metrics. The tool is found to be suitable for the specified purpose and would be further improved with more training data.