Developing AI/ML Based Predictive Capabilities for a Compression Ignition Engine Using Pseudo Dynamometer Data

The compression ignition and small displacement spark-ignition engines play an important role in the urban air pollution. In particular, vehicles equipped with compression ignition engines are widely used because of their higher performance and fuel efficiency with respect to the spark-ignition ones. Nevertheless, spark-ignition engines with low displacements are even more wide-spreading because of the lower fuel consumption and emissions. They are also used for two-wheeled vehicles, whose easier navigation makes them widely used in heavily congested areas. Their contribution on urban pollution is worsened by the fact that these vehicles have to comply with the Euro 3 standard; light vehicles have, instead, to fulfill the more restrictive Euro 6, which for the compression ignition and gasoline direct injection engines indicates for particle emissions also a number-based regulation. This article aims to characterize the effects of biofuels on engine emissions and performance of compression ignition and spark-ignition engines. The investigation was carried out on different class of engines. Direct injection and a port fuel injection spark-ignition engines fueled with ethanol and its blends, 10 v/v%, 50 v/v% and 85 v/v% of ethanol in gasoline. The compression ignition engine was equipped with a common rail injection system and was fueled with pure rapeseed methyl ester, representative of fatty acid methyl ester, and its blends in diesel, 20 v/v% and 50 v/v%. The gaseous emissions and the particle concentration were measured at the exhaust by means of conventional instruments. Particle size distribution function was measured in the range from 5.6 to 560 nm by means of an engine exhaust particle sizer. A comprehensive characterization of the particulate carbon was performed by means of optical diagnostics in the combustion chamber. In particular, two-dimensional images of flame evolution were detected and processed by two-color pyrometry technique to assess the in-cylinder soot formation and oxidation processes. For both the investigated spark-ignition and compression ignition engines, the use of biofuels shows a partial increase in the specific fuel consumption and a reduction of the soot particles emission. Nevertheless, a further effort on engine technology should be paid to balance the mass with the size and number of the particles.

ABSTRACTAchieving the new emission norms is a difficult task to today’s compression ignition (CI) engine without any exhaust gas after-treatment technologies. It is necessary to find the practical method which reduces the unsafe emission, with minor modifications of the CI engine. Dual fuel homogeneous charge compression ignition (HCCI) engine has been recognised as one of the solutions to minimise the emissions and achieve higher performance. In the present study the dual fuel HCCI engine mode of operation carried out by supply of ethanol fuel-air mixture to the engine cylinder through the carburetor and diesel fuel is directly injected at the end of compression for the initiation of ignition. Dual fuel HCCI engine is one of the most promising engines suitable for alternative fuels and lower NOX emissions. An experimental investigation is carried out on dual fuel HCCI engine. Fuel consumption and exhaust emissions such as NOX, CO2, CO, HC are measured and compared with conventional CI engine. The results show that NOX emission tends to decrease at low and moderate loads of the engine, but at full load condition it is slightly higher. Further, thermal efficiency is calculated and compared in CI engine; it is observed that there is a slight improvement in thermal efficiency at high load operation. In the dual fuel HCCI engine mode, there is a provision to use of ethanol or any other alternate fuel for better energy efficiencies and low NOX emission.

In a homogeneous Poisson process in R d , consider an arbitrary point X and let Y be its kth nearest neighbour. Denote by Rk the rank of X in the proximity order defined by Y, i.e., Rk = j if X is the jth nearest neighbour to Y. A representation for Rk in terms of a sum of independent random variables is obtained, and the limiting distribution of Rk, as k →∞, is shown to be normal. This result generalizes to mixtures of Poisson processes.

Two melanoma data sets, with and without an important attribute called TDS, were studied. The index TDS is a valuable diagnostic parameter used in the diagnosis of melanoma. Both data sets were partitioned into training data sets (250 cases) and testing (26 cases). Two classifiers were used: the well-known k-th Nearest Neighbor (KNN) algorithm and the algorithm LEM2, a part of the data mining system LERS. KNN algorithm is a typical statistical method based on voting among k training cases that are as close to the tested case as possible. On the other hand, LEM2 induces rules, and then classification is based on voting among all rules that match the case. Surprisingly, both methods yield similar results. However, the KNN algorithm may produce a smaller error rate, but LEM2 offers explanation of its results.

It is key index of air-jet texturing yarn quality such as air-jet texturing yarn strength and so on. It can well control air-jet texturing yarn quality by predicting yarn strength and so on. Generally, it is normal used to predict yarn strength such as Multiple Linear Regression (MLR), Support Vector Regression (SVR) and shallow Artificial Neural Network (ANN). Because the processing of air-jet texturing yarn production has time sequence, the paper proposes a new deep neural network, it is Long Short-Term Memory neural network (LSTM). It used 1800 sets of data to train LSTM, SVR and ANN. It tested LSTM, MLR, SVR and ANN with 200 sets of data. Experimental results show that LSTM neural network is the best accuracy among these four algorithms.

The study presents the results of the development and testing of deep learning neural network architectures, which demonstrate high accuracy rates in classifying neurophysiological data, in particular hemodynamic brain activation patterns obtained by functional near-infrared spectroscopy, during solving mathematical problems on spatial-numerical associations. The analyzed signal represents a multidimensional time series of oxyhemoglobin and deoxyhemoglobin dynamics. Taking the specificity of the fNIRS signal into account, a comparative analysis of 2 types of neural network architectures was carried out: (1) architectures based on recurrent neural networks: recurrent neural network with long short-term memory, recurrent neural network with long short-term memory with fully connected layers, bidirectional recurrent neural network with long short-term memory, convolutional recurrent neural network with long short-term memory; (2) architectures based on convolutional neural networks with 1D convolutions: convolutional neural network, fully convolutional neural network, residual neural network. Trained long short-term memory recurrent neural network architectures showed worse results in accuracy in comparison with 1D convolutional neural network architectures. Residual neural network (model_Resnet) demonstrated the highest accuracy rates in three experimental conditions more than 88% in detecting age-related differences in brain activation during spatial-numerical association tasks considering the individual characteristics of the respondents’ signal.

In this paper, the locomotion and steering control of a simulated Mini Cheetah quadruped robot was investigated in the presence of terrain characterised by low friction. Low-level locomotion and steering control were implemented via a central pattern generator approach, whereas high-level steering control manoeuvres were implemented by comparing a neural network and a linear model predictive controller in a dynamic simulation environment. A data-driven approach was adopted to identify the robot model using both a linear transfer function and a shallow artificial neural network. The results demonstrate that, whereas the linear approach showed good performance in high-friction terrain, in the presence of slippery conditions, the application of a neural network predictive controller improved trajectory accuracy and preserved robot safety with different steering manoeuvres. A comparative analysis was carried out using several performance indices.

paper, we focus on exploring the relationship between weave patterns and their mechanical properties in woven fiber composites through Machine Learning. Specifically, we explore the interactions between woven architectures and in-plane stiffness properties through Deep Convolutional Neural Network (DCNN) and Generative Adversarial Network (GAN). Our research is important for exploring how woven composite’s pattern is related to its mechanical properties and accelerating woven composite design as well as optimization. We focus on two tasks: (1) Stiffness prediction: Predicting in-plane stiffness properties for given weave patterns. Our DCNN extracts high-level features through several convolutional and fully connected layers to determine the final predictions. (2) Weave pattern prediction: Predicting weave patterns for target stiffness properties, which can be treated as the reverse task of the first one. Due to many-to-one mapping between weave patterns and the composite properties, we utilize a Decoder Neural Network as our baseline model and compare its performance with GAN and Genetic Algorithm. We represent the weave patterns as 2D checkerboard models and use finite element analysis (FEA) to determine in-plane stiffness properties, which serve as input data for our ML framework. We show that: (1) for stiffness prediction, DCNN can predict stiffness values for a given weave pattern with relatively high accuracy (above 93%); (2) for weave pattern prediction, the GAN model gives the best prediction accuracy (above 92%) while Decoder Neural Network has the best time efficiency. HAOTIAN FENG

In the study, the authors consider classical classification methods for the applied problem of image recognition. Accuracy, computation time, classifier size, training time for the following methods are considered: "Fully Connected Neural Network", "Convolutional Neural Network", "Recurrent Neural Network", "Decision Tree", "Gradient Boosted Trees", "Logistic Regression", "Markov", "Naive Bayes", "Nearest Neighbors", "Random Forest", "Support Vector Machine". A new approach "Neural Network with Hash" is proposed, which represents image preprocessing using polynomial hashing. Collision resolution is performed by a fast method of open addressing. A computer experiment on classification by 10 classes was carried out on a dataset of 600 animal images using the Wolfram mathematical package. For the proposed approach with preprocessing, the results showed the same classification accuracy as the classical methods, and a higher training and computational speed than the "Convolutional Neural Network" and "Recurrent Neural Network".

Peptide-based vaccine development needs accurate prediction of the binding affinity between major histocompatibility complex I (MHC I) proteins and their peptide ligands. Nowadays more and more machine learning methods have been developed to predict binding affinity and some of them have become the popular tools. However most of them are designed by the shallow neural networks. Bengio said that deep neural networks can learn better fits with less data than shallow neural networks. In our case, some of the alleles only have dozens of peptide data. In addition, we transform each peptide into a characteristic matrix and input it into the model. As we know when dealing with the problem that the input is a matrix, convolutional neural network (CNN) can find the most critical features by itself. Obviously, compared with the traditional neural network model, CNN is more suitable for predicting binding affinity. Different from the previous studies which are based on blocks substitution matrix (BLOSUM), we used novel feature to do the prediction. Since we consider that the order of the sequence, hydropathy index, polarity and the length of the peptide could affect the binding affinity and the properties of these amino acids are key factors for their binding to MHC, we extracted these information from each peptide. In order to make full use of the data we have obtained, we have integrated different lengths of peptides into 15mer based on the binding mode of peptide to MHC I. In order to demonstrate that our method is reliable to predict peptide-MHC binding, we compared our method with several popular methods. The experiments show the superiority of our method.

Ceramic Material for Thermal Barrier Coatings in Compression Ignition Engine for its Performance Evaluation with Biodiesel

The importance of biodiesel as a renewable and economically viable alternative to fossil diesel for applications in compression ignition (CI) engines has led to intense research in the field over the last two decades. This is predominantly due to the depletion of petroleum resources, and increasing awareness of environmental and health impacts from the combustion of fossil diesel. Biodiesel is favoured over other biofuels because of its compatibility with present day CI engines, with no further adjustments required to the core engine configurations when used in either neat or blended forms. Studies conducted to date on various CI engines fuelled with varying biodiesel types and blends under numerous test cycles have shown that key tailpipe pollutants, such as carbon monoxide, aromatics, sulphur oxides, unburnt hydrocarbons and particulate matters are potentially reduced. The effects of biodiesel on nitrogen oxides emission require further tests and validations. The improvement in most of the diesel emission species comes with a trade-off in a reduction of brake power and an increase in fuel consumption. Biodiesel’s lubricating properties are generally better than those of its fossil diesel counterpart, which result in an increased engine life. These substantial differences in engine-out responses between biodiesel and fossil diesel combustion are mainly attributed to the physical properties and chemical composition of the fuels. Despite the purported benefits, widespread adoption of biodiesel usage in CI engines is hindered by outstanding technical challenges, such as low temperature inoperability, storage instabilities, in-cylinder carbon deposition and fuel line corrosion. It is imperative that these issues are addressed appropriately to ensure that long-term biodiesel usage in CI engines does not negatively affect the overall engine durability. Possible solutions range from biodiesel fuel reformulation through feedstock choice and production technique, to the simple addition of fuel additives. This calls for a more strategic and comprehensive research effort internationally, with an overarching approach for co-ordinating sustainable exploitation and utilisation of biodiesel. This review examines the combustion quality, exhaust emissions and tribological impacts of biodiesel on CI engines, with specific focus on the influence of biodiesel’s physico-chemical properties. Ongoing efforts in mitigating problems related to engine operations due to biodiesel usage are addressed. Present day biodiesel production methods and emerging trends are also identified, with specific focus on the conventional transesterification process wherein factors affecting its yield are discussed.

As for the problem of too long training time of convolution neural network (CNN), this paper proposes a fast training method for CNN in SAR automatic target recognition (ATR). The CNN is divided into two parts: one that contains all the convolution layers and sub-sampling layers is considered as convolutional auto-encoder (CAE) for unsupervised training to extract high-level features; the other that contains fully connected layers is regarded as shallow neural network (SNN) to work as a classifier. The experiment based on MSATR database shows that the proposed method can tremendously reduce the training time with little loss of recognition rate.

Automatic eyeglasses detection plays a major role in many facial analysis systems. To improve the robustness of these systems and cope with real-world applications, a high-speed eyeglasses detector that can achieve high accuracy is needed. Recent studies indicate that the features extracted from convolutional neural networks are compelling. Therefore, this paper presents an effective and efficient method for eyeglasses detection in facial images based on extracting deep features from a well-designed shallow convolutional neural network (CNN). The main contribution of this paper is to address the two essential aspects of CNN: (1) the size of the training dataset required and (2) the depth of the network architecture. To this end, we initialize the learning parameters of the shallow CNN by the parameters of a deep CNN which is fine-tuned on a small dataset. The depth of the neural network is then decreased by removing some convolutional layers after testing its performance on the validation dataset. As a result, a significantly more accurate shallow CNN architecture, Shallow-GlassNet, is obtained, which achieves not only high accuracy but also high speed in eyeglasses detection. Evaluation experiments have been conducted on two large unconstrained facial image databases, LFW and Celeb Faces. The results have demonstrated the superior performance of the proposed framework which achieves a mean accuracy of 99.73%.

Evaluating historical simulations from global climate models (GCMs) remains an important exercise for better understanding future projections of climate change and variability in rapidly warming regions, such as the Arctic. As an alternative approach for comparing climate models and observations, we set up a machine learning classification task using a shallow artificial neural network (ANN). Specifically, we train an ANN on maps of annual mean near‐surface temperature in the Arctic from a multi‐model large ensemble archive in order to classify which GCM produced each temperature map. After training our ANN on data from the large ensembles, we input annual mean maps of Arctic temperature from observational reanalysis and sort the prediction output according to increasing values of the ANN's confidence for each GCM class. To attempt to understand how the ANN is classifying each temperature map with a GCM, we leverage a feature attribution method from explainable artificial intelligence. By comparing composites from the attribution method for every GCM classification, we find that the ANN is learning regional temperature patterns in the Arctic that are unique to each GCM relative to the multi‐model mean ensemble. In agreement with recent studies, we show that ANNs can be useful tools for extracting regional climate signals in GCMs and observations.