Accurate prediction of price-incentive heating, ventilation and air conditioning (HVAC) loads is an important prerequisite for their participation in demand response (DR). However, traditional physics-based prediction models exhibit poor scenario adaptability that leads to modeling bias, while data-driven models struggle to characterize the physical relationships between inputs and output and thus suffer from poor interpretability. To address these limitations, this paper proposes a physics-informed long short-term memory network (PILSTM) for price-incentive HVAC loads prediction. The model dynamically adjusts feature weights to capture the time-varying effects of temperature and electricity prices on load, and optimizes the loss function of the model by incorporating physical monotonic consistency and load boundary constraints, thereby concurrently enhancing both the model prediction accuracy and interpretability. Experimental results across different scenarios show that: in real-time pricing scenario, compared to LSTM, the proposed model reduces the mean absolute error (MAE), root mean square error (RMSE), and mean absolute percentage error (MAPE) by 13.57%, 12.45%, and 12.42%, respectively, while its predictions are more consistent with the actual response characteristics of HVAC; in time-of-use pricing scenario, the R2 of PILSTM is 10.18% higher than that of LSTM, demonstrating superior prediction accuracy; seasonal load forecasts reveal that the performance indicators of PILSTM exhibit better values and smaller fluctuations, indicating its adaptability to diverse HVAC operating conditions and better prediction stability. By dynamically adjusting feature weights and incorporating physical constraints, the proposed model accurately predicts HVAC loads under diverse operating conditions, thereby providing reliable data support for energy system optimization.

Multiple uncertainties in the park-level integrated energy system (PIES) can affect the optimal operation of system. Information gap decision theory (IGDT) is a commonly used method of dealing with uncertainty by developing risky strategies to avoid risks or seek risky returns. However, there is no unified method or process for the selection of risk strategies and the setting of related parameters, leading to a certain blindness in the application of IGDT method. A comprehensive risk strategy (CRS)- IGDT approach is proposed for scheduling of PIES considering uncertainties of heat load, photovoltaic output and electric load. Risk averse strategy (RAS) and Risk seek strategy (RSS) scheduling models are constructed. Then an optimized solution method based on adaptive steps ratio (ASR) is proposed to solve the above two models. The CRS and comprehensive risk cost function are proposed from a risk equalization perspective. The target deviation coefficient and steps ratio in the IGDT model are automatically optimized with the objective of minimizing the comprehensive risk cost. Combining the two cases, the average cost reduction is 6.6% compared to Risk-neutral (RN), 11 % compared to RAS-IGDT, and 4.1 % compared to RSS-IGDT. Moreover, the average costs of CRS-IGDT are lower compared to stochastic programming and robust optimization methods. The experiments verify the generalization and superiority of the proposed method in coping with different information gap situations caused by uncertainty. CRSIGDT provides new research ideas for dealing with uncertainty problems in PIES and other fields.

Partial discharge (PD) diagnosis in gas insulated switchgear (GIS) is crucial for the safe and stable operation of substations. Pattern recognition and severity assessment are two major areas of research in GIS partial discharge diagnosis. Currently, studies on these topics are relatively independent, lacking a unified model, which results in poor model generalization and high consumption of hardware and software resources. To address this issue, this paper proposes the Gradient Balanced Selective Mixture-of-Experts (GB-SMoE) for GIS partial discharge pattern recognition and severity assessment. Firstly, a phase-wise fusion method for partial discharge phase-resolved pulse sequence (PRPS) data is designed, which combines acoustic emission and ultra-high frequency data of GIS partial discharges as input for the multi-task model. Secondly, the GB-SMoE multi-task model is proposed, optimizing the multi-task model in terms of parameter sharing and task weighting. Finally, the proposed method is trained and evaluated on the constructed dataset. Experimental results demonstrate that phase-wise fused PRPS data facilitates more accurate pattern recognition and severity assessment of partial discharges. The proposed GB-SMoE model achieves a classification accuracy of 94.97% and a severity assessment accuracy of 91.31%, significantly improving the accuracy of both tasks compared to single-task and basic multi-task models.

Path planning for mobile robots can be divided into two categories based on how much information is available apriori to planning - global path planning and local path planning. While in global path planning, details about the environment are known to the robot, in local path planning almost all that information is not known in advance. Several methodologies have been proposed for path planning optimization for mobile robots in a known environment. While many of these methods use grid-based and sampling-based algorithms such as A-star (A*) and rapidly exploring random trees (RRT), more recently the application of metaheuristic search-based optimization techniques such as genetic algorithms (GA), particle swarm optimization (PSO) and differential evolutions (DE) have been explored. However, in metaheuristic search-based optimization methods, the fitness function to guide the search is based on rewarding solutions or agents that find the shortest path irrespective of whether the shortest path involves irregular and erratic motion that promotes excessive wear and tear of the robot over time. Papers that have investigated smooth plan planning algorithms almost always use curve smoothing techniques such as higher degree splines (Bezier or B-splines) to construct smooth and continuous trajectories, rather than using kinematics based objective function optimization during path planning. In this paper, we investigate the use of multi-objective kinematics-based performance criteria for global path planning that reward two competing objectives: (1) minimizing twists and turns, and (2) achieving the shortest possible path. The path planning is implemented using four variants of the traditional PSO and DE algorithms. Simulation results are compared with path planning methods using traditional methods based on RRT. Preliminary results show the proposed method converges as quickly as the single objective path planning and provides more realistic and smooth paths (minimizes twists and turns but does not necessarily result in the shortest path). Amongst the algorithms used, the DE and its variants presented in this work perform better than the PSO.

End-to-end neural network models, often seen as black boxes, have been widely used in photovoltaic (PV) power forecasting. However, they face challenges regarding poor model adaptability, transferability, and interpretability. To address these issues, this paper proposes a physical-encoded PV forecasting model, which decomposes the end-to-end network into a data-driven external parameter forecasting model and a physics-driven power calculation model. The power calculation model, with explicit physical meanings, enhances the model's interpretability. A continual learning mechanism is designed to enable the model to quickly adapt to environmental changes, mitigating the impact of model drift and improving adaptability and transferability. A multi-digital twins synergistic operation mechanism is introduced to incorporate the strengths of other models, further enhancing forecasting accuracy. Model drift can be categorized into concept drift and data drift. This paper designs two scenario experiments to test these drifts. Scenario 1 focuses on concept drift, and the experimental results show that the proposed method in this paper achieves improvements of 30.5 %, 16.5 %, and 1.9 % in the nMAE, nRMSE, and R2 metrics, respectively, compared to the best results of the comparison models. In Scenario 2, the model is transferred to other power plants for data drift tests. Results show that when transferred to Plant 4, its accuracy improves by 45.8 %, 21 %, and 2.1 % compared to the best comparison method; for Plant 5, the improvements are 34.1 %, 18.3 %, and 2.5 %.

Despite the current advances in additive manufacturing, subtractive manufacturing methods such as machining still account for the major share of modern manufacturing. Surface roughness of a machined product is a crucial parameter that impacts the functionality, assembly, and service life of the product. Surface finish texture of machined components is too complex for accurate prediction when using analytical or computer simulation techniques. This is because there are numerous parameters relating to the material of the workpiece, cutting tool, and machining process conditions. Therefore, machine learning (ML) techniques are becoming more popular in creating model-based methods that are capable of providing more reliable real-time surface quality prediction. The aim of this study is to develop ML models to predict the surface roughness in shoulder milling of steel parts using audible acoustic emission data produced during machining. Microphones are used to pick up acoustic data that is highly correlated to acoustic waves produced by the machining process. These sound measuring devices are non-invasive and can be easily integrated within the machining envelope without disrupting or stopping the machining process. Features are then extracted from the acoustic data that include averaged wavelet decomposition quantities, statistical quantities, and filtered time signatures of the sound waves. These are then (1) used to training classifiers using important features or a combination of features that have highly corelated to surface finish, and (2) develop a learning model to use these features to predict the surface roughness in shoulder milling. In this study, we use a variety of dimensionality reduction algorithms in the training phase. In the learning model, we use classification algorithms and develop suitable classifiers. The overall objective is to develop a reliable and robust predictive tool with potential for practical implementation in a real-time industrial machine tool installation for process monitoring.

Density-based clustering methods such as DBSCAN are known to be robust against outliers in data; however, they are sensitive to user-specified parameters, the selection of which are not trivial. In this paper, the user-defined parameters of DBSCAN are evolved using a quantum-inspired genetic algorithm (QGA). The quantum-bit or Q-bit representation of a partition is an improvement over the more popular binary label-based representations and real-coded representation of partition cluster centers. A resulting algorithm called DBSCAN-QGA in the relational data space is proposed, and three different fitness functions are devised to evaluate partitions both in terms of cluster compactness and separation, and the relative number of entities classified as noise. The performance of the proposed algorithm is compared to synthetic and benchmark datasets from the UCI machine learning repository with encouraging results.

Surface roughness quality has implications on the functionally, assembly, service life, and appearance of the machined product. Considering the complex nature of metal cutting processes, computer simulation models may not provide the needed accuracy to predict surface conditions under all cutting conditions. Therefore, machine learning (ML) techniques can provide more reliable predication models that are based on real time cutting process sensory data. The implementation of artificial intelligence (AI) techniques in the monitoring of manufacturing processes has been gaining momentum. The focus of this study is to predict the surface roughness using acoustic emissions (AE) signals during the dry end milling of stainless steel. AE sensors have been widely used to monitor the condition of structures and manufacturing processes. Furthermore, acoustic sensors are non-invasive and can be used at any location without disrupting or stopping the machining process. Features extracted from the AE signals are used as surface roughness quality indicators. These features include frequency bands averaged amplitudes, statistical quantities of the wavelet decompositions, raw signal RMS values, and crest factor. In this work, several machine learning algorithms are used to process the extracted AE features for surface roughness characterization. The total AE features are first processed for feature set reduction since many of the features are highly correlated. This is done using both supervised and unsupervised feature reduction and subset selection methods. The features extracted from supervised feature reduction methods are used to train three supervised classifiers — k-nearest neighbor (kNN) classifier, a radial-basis function support vector machine (RBF-SVM), and a random forest (RF) classifier. The reduced feature set from the unsupervised feature reduction methods are used as input to two unsupervised clustering methods — K-Means and DBSCAN. The classifier models are trained using multi-fold cross-validated mix of subsets of the reduced features. In this study we have used ten models using two-fold cross validation for training and validation for the supervised learning methods. The results of supervised classification are compared to unsupervised clustering and are reported for an average of the ten models (or ten runs with distinct initializations of the clustering algorithm), along with a detailed nonparametric testing to verify statistical significance in performance level between pairs of algorithms.

The need to accurately estimate wind power is essential to the design and deployment of individual wind turbines and wind farms. The estimation problem is framed as wind power curve modeling. Lately, machine learning techniques have been used to model power curves and provide power estimates. Such models rely on the fact that all outliers are removed from the raw wind data before they are used in modeling and estimation since outliers can adversely affect performance. However, generating outlier-free data is not always possible. Robust models and robust objective functions can be two effective ways to obtain accurate power curves in the presence of outliers. In this paper, a robust density-based clustering technique (DBSCAN) to first identify outliers in the dataset is proposed, followed by artificial neural network (ANN) models that are trained using the outlier-free data to obtain accurate power curve estimates. ANNs are trained using a range of optimization methods and are compared in this study. Preliminary results show the proposed method is superior to probabilistic models that use error-functions to generate accurate power curves and that the proposed hybrid model can generate more accurate power output estimations in the presence of outliers compared to deterministic models such as integrated curve fitting models that are known to be robust.

In this work, vibration response of a rolling element bearing under the influence of static radial loading is investigated. Radial loading results in a periodically varying stiffness (or compliance) which causes a cyclic dynamic response of the bearing assembly even under perfect balancing and other operating load conditions. These loads cause high stresses to develop in bearing elements and may cause fatigue, cracks, and spalls that limits the life of these components. A special bearing test rig was designed and manufactured to apply varying levels of radial load and measure the vibration response of the loaded roller bearing. The test is focused on new bearings free from any faults or defects. The radial load is varied in steps and the vibration signal is collected and analyzed at each level for different rotor speeds. The spectral components are analyzed using Fast Fourier Transform (FFT) and time-frequency wavelet transform. Statistical techniques are applied to both the vibration signature obtained using a piezoelectric accelerometer sensor and the wavelet decomposed approximations and details of the original vibration signals. The statistical measures, wavelet approximation and details are first processed for feature set reduction since many of the features are highly correlated. This is done using three feature reduction and subset selection methods — ReliefF, Recursive Feature Extraction (RFE) and Multi-Cluster Feature Selection (MCFS). These features and the original extracted features are used as features to train two classifiers. The classification is used to estimate high and low thresholds for both radial load and running speed. The classifiers used are (1) radial-basis function support vector machine (RBF-SVM), and (2) k-nearest neighbor (kNN). Performance of machine learning algorithms depends on the training data and physical collected datasets are often limited to specific operating conditions, necessitating the use of training with many models using multi-fold cross-validated subsets. In this study we have used ten models using two-fold cross validation for training and validation. The classification results reported are average of these models. In limited experimentation, the RBF-SVM outperforms the kNN classifier and among the feature sets used, the ReliefF set seems marginally superior to the other sets. However, the accuracy, precision, and recall (combined as an F-score) of the original extracted feature set are better than the reduced feature sets; the downside being the relatively high run time in the training phase.

Fuzzy c-means (FCM), the fuzzy variant of the popular k-means, has been used for data clustering when cluster boundaries are not well defined. The choice of initial cluster prototypes (or the initialization of cluster memberships), and the fact that the number of clusters needs to be defined a priori are two major factors that can affect the performance of FCM. In this paper, we review algorithms and methods used to overcome these two specific drawbacks. We propose a new cooperative multi-population differential evolution method with elitism to identify near-optimal initial cluster prototypes and also determine the most optimal number of clusters in the data. The differential evolution populations use a smaller subset of the dataset, one that captures the same structure of the dataset. We compare the proposed methodology to newer methods proposed in the literature, with simulations performed on standard benchmark data from the UCI machine learning repository. Finally, we present a case study for clustering time-series patterns from sensor data related to real-time machine health monitoring using the proposed method. Simulation results are promising and show that the proposed methodology can be effective in clustering a wide range of datasets.

One of the widely used aerospace alloys is 7075 aluminum alloy (AA) because of its high tensile and compressive strength and good response to exfoliation corrosion. A solution heat treatment followed by artificial ageing is known as T6 temper designation was initially used to obtain peak strength for 7075 AA. The artificial aging, done at 115°C to 130°C (T6 temper), increases strength of 7075 AA to a peak level then decreases, however, resistance to stress-corrosion cracking is decreased. Recent trend shows that the strength can be increased more by applying multi-stage aging process. This research focuses on the effect of multiple aging temperature, time on the mechanical properties of 7075 aluminum alloys. ASTM standard coupons were machined from an as received aluminum plate and applied different combination of RRA age treatment. The initial results on mechanical properties are reported. The maximum tensile strength obtained was over 100 ksi for a double RRA at 200°C for 10 min. The hardness was measured using a micro hardness tester. The Vickers hardness numbers were within the range in literatures.

Surface roughness measurements of machined parts are usually performed off-line after the completion of the machining operation. The objective of this work is to develop a surface roughness prediction method based on the processing of vibration signals during steel end milling operation performed on a vertical CNC machining center. The milling cuts were run under varying conditions (such as the spindle speed, feed rate, and depth of cut). This is a first step in the attempt to develop an online milling process monitoring system. The study presented here involves the analysis of vibration signals using statistical time parameters, frequency spectrum, and time-frequency wavelet decomposition. The analysis resulted in the extraction of 245 features that were used in the evolutionary optimization study to determine optimal cutting conditions based on the measured surface roughness of the milled specimen. Three feature selection methods were used to reduce the extracted feature set to smaller subsets, followed by binarization using two binarization methods. Three evolutionary algorithms—a genetic algorithm, particle swarm optimization and two variants, differential evolution and one of its variants, have been used to identify features that relate to the “best” surface finish measurements. These optimal features can then be related to cutting conditions (cutting speed, feed rate, and axial depth of cut). It is shown that the differential evolution and its variant performed better than the particle swarm optimization and its variants, and both differential evolution and particle swarm optimization perform better than the canonical genetic algorithm. Significant differences are found in the feature selection methods too, but no difference in performance was found between the two binarization methods.

This paper investigates deception in the context of motion using a simulated mobile robot. We analyze some previously designed deceptive strategies on a mobile robot simulator. We then present a novel approach to adaptively choose target-oriented deceptive trajectories to deceive humans for multiple interactions. Additionally, we propose a new metric to evaluate deception on data collected from the users when interacting with the mobile robot simulator. We performed a user study to test our proposed adaptive deceptive algorithm, which shows that our algorithm deceives humans even for multiple interactions and it is more effective than random choice of deceptive strategies.

In metal-cutting processes, the interaction between the tool and workpiece is highly nonlinear and is very sensitive to small variations in the process parameters. This causes difficulties in controlling and predicting the resulting surface finish quality of the machined surface. In this work, vibration signals along the major cutting force direction in the turning process are measured at different combinations of cutting speeds, feeds, and depths of cut using a piezoelectric accelerometer. The signals are processed to extract features in the time and frequency domains. These include statistical quantities, Fast Fourier spectral signatures, and various wavelet analysis extracts. Various feature selection methods are applied to the extracted features for dimensionality reduction, followed by applying several outlier-resistant unsupervised clustering algorithms on the reduced feature set. The objective is to ascertain if partitions created by the clustering algorithms correspond to experimentally obtained surface roughness data for specific combinations of cutting conditions. We find 75% accuracy in predicting surface finish from the Noise Clustering Fuzzy C-Means (NC-FCM) and the Density-Based Spatial Clustering Applications with Noise (DBSCAN) algorithms, and upwards of 80% accuracy in identifying outliers. In general, wrapper methods used for feature selection had better partitioning efficacy than filter methods for feature selection. These results are useful when considering real-time steel turning process monitoring systems.

Model-based design of manufacturing processes have been gaining popularity since the advent of machine learning algorithms such as evolutionary algorithms and artificial neural networks (ANN). The problem of selecting the best machining parameters can be cast an optimization problem given a cost function and by utilizing an input-output connectionist framework using as ANNs. In this paper, we present a comparison of various evolutionary algorithms for parameter optimization of an end-milling operation based on a well-known cost function from literature. We propose a modification to the cost function for milling and include an additional objective of minimizing surface roughness and by using NSGA-II, a multi-objective optimization algorithm. We also present comparison of several population-based evolutionary search algorithms such as variants of particle swarm optimization, differential evolution and NSGA-II.

Bearing faults in machinery are among the most critical faults that require attention by maintenance personnel at early stages of fault initiation. In many cases it is difficult to directly and accurately identify the fault type and its extent under varying operating conditions. This work demonstrates a novel procedure for bearing fault detection and identification in an experimental set-up. Three seeded faults, in the rotating machinery supported by the test ball bearing, include inner race fault, outer race fault and one roller fault. The rotor is run at different speeds and with small level of rotating mass unbalance. Accelerometer based vibration signals are analyzed for the different bearing faults’ signatures using statistical features, frequency spectra and wavelet coefficients. The composite differential evolution technique is proposed for parameter estimation when the system response is known a-priori. The algorithm is compared to five other differential evolution algorithms using conventional crossover and mutation operators. The objective is to correlate bearing faults to the extracted vibration features. The results of this analysis will be extended for applications in real time bearing condition monitoring system.

Automatic monitoring of group-housed pigs in real time through porcine acoustic signals has played a crucial role in automated farming. In the process of data collection and transmission, acoustic signals are generally interfered with noise. In this paper, an effective porcine acoustic signal denoising technique based on ensemble empirical mode decomposition (EEMD), independent component analysis (ICA), and wavelet threshold denoising (WTD) is proposed. Firstly, the porcine acoustic signal is decomposed into intrinsic mode functions (IMFs) by EEMD. In addition, permutation entropy (PE) is adopted to distinguish noise-dominant IMFs from the IMFs. Secondly, ICA is employed to extract the independent components (ICs) of the noise-dominant IMFs. The correlation coefficients of ICs and the first IMF are calculated to recognize noise ICs. The noise ICs will be removed. Then, WTD is applied to the other ICs. Finally, the porcine acoustic signal is reconstructed by the processed components. Experimental results show that the proposed method can effectively improve the denoising performance of porcine acoustic signal.

Abstract. With the rapid development of large-scale breeding, manual long-term monitoring of the daily activities and health of livestock is costly and time-consuming. Therefore, the application of bio-acoustics to automatic monitoring has received increasing attention. Although bio-acoustical techniques have been applied to the recognition of animal sounds in many studies, there is a dearth of studies on the automatic recognition of abnormal sounds from farm animals. In this study, an automatic detection and recognition system based on bio-acoustics is proposed to hierarchically recognize abnormal animal states in a large-scale pig breeding environment. In this system, we extracted the mel-frequency cepstral coefficients (MFCC) and subband spectrum centroid (SSC) as composite feature parameters. At the first level, support vector data description (SVDD) is used to detect abnormal sounds in the acoustic data. At the second level, a back-propagation neural network (BPNN) is used to classify five kinds of abnormal sounds in pigs. Furthermore, improved spectral subtraction is developed to reduce the noise interference as much as possible. Experimental results show that the average detection accuracy and the average recognition accuracy of the proposed system are 94.2% and 95.4%, respectively. The effectiveness of the proposed sound detection and recognition system was also verified through tests at a pig farm. Keywords: Abnormal sounds, MFCC, SSC, States of pigs, SVDD.

The present methods of data preservation and representation for barrel finishing processes which include paper and electronic documents have several disadvantages such as restrictions in size and complexity, and limitations on query and updation speed. Aiming at these disadvantages, a new database platform for barrel finishing data has been constructed by using database technology and case-based reasoning. The design procedure of the database platform is expounded in detail, covering analysis of database platform requirements, establishment for conceptual model of database data structure, designs for logical model of database data structure, determination for physical model of database data structure, choice for network structure of database platform, data management and storage method. The application results demonstrate that the database platform can ensure the safe and convenient storage as well as the sharing of experimental data of the barrel finishing process. It can also provide guidance and technical information for scientific researchers, experts, technicians, and production site operators to choose the processing technology and the processing parameters reasonably.