Chen et al.  used a multiple regression analysis (MRA) model for thermalerror compensation of a horizontal machining centre. With their experimental results, the thermalerror was reduced from 196 to 8 l m. Yang et al.  also used the MRA model to form an error synthesis model which merges both the thermal and geometric errors of a lathe. With their exper- imental results, the error could be reduced from 60 to 14 l m. However, the thermal displacement usually changes with var- iation in the machining process and the environment; it is difficult to apply MRA to a multiple output variable model. In order to overcome the drawbacks of MRA models, more attention has subsequently been given to the Artificial Intelligence (AI) techniques such as Artificial Neural Networks (ANNs). Chen et al.  proposed an ANN model structured with 15 nodes in the input layer, 15 nodes in the hidden layer, and six nodes in the output layer in order to drive a thermalerror compen- sation of the spindle and lead-screws of a vertical machining centre. The ANN model was trained with 540 training data pairs and tested with a new cutting condition, which was not included within the training pairs. Test results showed that the ther- mal errors could be reduced from 40 to 5 l m after applying the compensation model, but no justification for the number of nodes or length of training data was provided. Wang  used a neural network trained by a hierarchy-genetic-algorithm (HGA) in order to map the temperature variation against the thermal drift of the machine tool. Wang  also proposed a thermal model merging Grey system theory GM(1,m) and an Adaptive Neuro-Fuzzy Inference System (ANFIS). A hybrid learning method, which is a combination of both steepest descent and least-squares estimator methods, was used in the learning algorithms. Experimental results indicated that the thermalerror compensation model could reduce the thermalerror to less than 9.2 l m under real cutting conditions. He used six inputs with three fuzzy sets per input, producing a com- plete rule set of 729 (3 6 ) rules in order to build an ANFIS model. Clearly, Wang’s model is practically limited to low dimen- sional modelling. Eskandari et al.  presented a method to compensate for positional, geometric, and thermally induced errors of three-axis CNC milling machineusing an offline technique. Thermal errors were modelled by three empirical meth- ods: MRA, ANN, and ANFIS. To build their models, the experimental data was collected every 10 min while the machine was running for 120 min. The experimental data was divided into training and checking sets. They found that ANFIS was a more accurate modelling method in comparison with ANN and MRA. Their test results on a free form shape show average improvement of 41% of the uncompensated errors. A common omission in the published research is discussion or scientific rigour regarding the selection of the number and location of thermal sensors.
The architecture and learning procedure of the Adaptive Neuro-Fuzzy System (ANFIS), have both been described by Jang . According to Jang, the ANFIS is a neural network that is functionally the same as a Takagi-Sugeno type inference model. The ANFIS is a hybrid intelligent system that takes the advantages of ANN and the fuzzy logic theory into a single system. By employing the ANN technique to update the parameters of the Takagi-Sugeno type inference model, the ANFIS is given the ability to learn from given training data, the same as ANN. The solutions mapped out onto the Takagi- Sugeno type inference model can therefore be described in linguistic terms. The efficiency of any ANFIS model depends on the success in partitioning the input and output variables space correctly. This can be achieved by using a number of methods such as grid partitioning (ANFIS-Grid partition model), the subtractive clustering method (ANFIS-Scatter partition model) and fuzzyc- meansclustering . The equivalent ANFIS network with two variables is shown in Figure 2:The first layer implements a fuzzification, the second layer executes the T-norm of the antecedent part of the fuzzy rules, the third layer normalizes the membership functions (MF), the fourth layer computes the consequent parameters, and finally the last layer calculates the overall output as the summation of all incoming signals .
Abdulshahed et al.  employed an adaptive neuro fuzzy inference system (ANFIS) to forecast thermalerror compensation on CNC machinetools. Two types of ANFIS model were built in this paper: using grid-partitioning and usingfuzzyc-meansclustering. According to the results, the ANFIS with fuzzyc-meansclustering produced better results, achieving up to 94 % improvement in error with a maximum residual error of ± 4 μm. In another work  they built a thermal model by integrating ANN and GMC(1, N) models. The thermal model can predict the Environmental Temperature Variation Error (ETVE) of a machine tool with reduction in error from over 20 μm to better than ± 3 μm. Nevertheless, robust solution for both principle-based and some of data driven models require the measurement of temperature and related thermalerror components that have to be obtained by time-consuming experiments. This is difficult to achieve in a working machine shop, because of the prohibitively costly downtime required to conduct the experiments.
Construction of the ANFIS model requires the division of the input-output data into rule patches. This can be achieved by using a number of methods such as grid partitioning, subtractive clustering method and fuzzyc-means (FCM) . According to Jang , grid partition is only suitable for problems with a small number of input variables (e.g. fewer than 6). A model with three inputs with three fuzzy sets per input produces a complete rule set of 27 rules, whereas a model with six inputs requires 729 (3 6 ) rules. Clearly standard ANFIS models are practically limited to low dimensional modelling. It is important to note that an effective partition of the input space can decrease the number of rules and thus increase the speed in both learning and application phases. In order to obtain a small number of fuzzy rules, a fuzzy rule generation technique that integrates ANFIS with FCM clustering will be applied in this paper, where the FCM is used to systematically create the fuzzy MFs and fuzzy rules base for ANFIS. In addition, it also helps to determine the initial parameters of the fuzzy model. This is important because an initial value, which is very close to the ﬁnal value, will eventually result in the quick convergence of the model towards its ﬁnal value during the training process .
Early work by Chen et al.  used both a multiple regression analysis (MRA) model and an artiﬁcial neural network (ANN) model for thermalerror compensation of a horizontal machining cen- tre. To build their models, 810 data sets were collected from ﬁve different tests; each test was run for 6 h for a heating cycle and then stopped for 10 h for a cooling down cycle. With their experi- mental results, the thermalerror was reduced from 196 to 8 mm. Wang  used a Hierarchy-Genetic-Algorithm (HGA) trained neu- ral network in order to map the temperature change against the thermal response of the machine tool. Wang  also proposed a thermal model by using an Adaptive Neuro Fuzzy Inference Sys- tem (ANFIS) and optimised the number of sensors by Grey system model GM(1,m). A hybrid learning method, which is a combination of both steepest descent and the least-squares estimator methods, was used in the learning algorithms. Experimental results indicated that the thermalerror compensation model could reduce the ther- mal error to less than 9 m under real cutting conditions. Wang in Refs. [10,8] used 150 min and 480 min of data acquisition in order to build HGA and ANFIS models, respectively. However, both mod- els require training cycles to calibrate the model how to respond to various changes in input conditions. Eskandari et al.  pre- sented a method by which to compensate for positional, geometric, and thermally induced errors of three-axis CNC milling machineusing an ofﬂine technique. Thermal errors are modelled by three empirical models: MRA, ANN, and ANFIS. To build their models, the experimental data were collected every 10 min while the machine was running for 120 min. The experimental data are divided into training and checking data sets. Their validated results on a free form, show signiﬁcant average improvement of 41% of the errors. Abdulshahed et al.  proposed a thermal model by using an ANFIS with fuzzyc-meansclustering. Different groups of key tem- perature points were identiﬁed from thermal images using a novel schema based on a GM (0, N) model and Fuzzyc-meansclustering. Experimental results indicated that the thermalerror compensa- tion model could reduce the thermalerror to less than 2 m. Also, similar works have been carried out by the same authors in Refs. [11,14,15] .
The choice of inputs to the thermal model is a non-trivial decision which is ultimately a compromise between the ability to obtain data that sufficiently correlates with the thermal distortion and the cost of implementation of the necessary feedback sensors. In this thesis, temperature measurement was supplemented by direct distortion measurement at accessible locations. The location of temperature measurement must also provide a representative measurement of the change in temperature that will affect the machine structure. The number of sensors and their locations are not always intuitive and the time required to identify the optimal locations is often prohibitive, resulting in compromise and poor results. In this thesis, a new intelligent system for reducing thermal errors of machinetoolsusing data obtained from thermography data is introduced. Different groups of key temperature points on a machine can be identified from thermal images using a novel schema based on a Grey system theory and FuzzyC-Means (FCM) clustering method. This novel method simplifies the modelling process, enhances the accuracy of the system and reduces the overall number of inputs to the model, since otherwise a much larger number of thermal sensors would be required to cover the entire structure.
Image analysis generally refers to preparing of images by computer with the objective of discovering what objects are exhibited in the image .Image segmentation is one of the fundamental and difficult tasks in many of the image and vision applications. It has been studied extensively over the past several decades with a huge number of segmentation algorithms being published in the literature. Those image segmentation approaches can be divided broadly into four categories: thresholding, clustering, edge detection and region extraction.
Seismic surveys may be broadly defined as “surface seismic” and “borehole seismic” also known as “Vertical Seismic Profiling” (VSP). In a surface seismic study usually both the source and the receivers (“geophones”; sensitive ground velocity sensors) are located at the surface and the reflected waves are analysed. In a Vertical Seismic Profiling borehole investigation the receivers are located within the borehole and the source is usually located at the surface or, less frequently, downhole. The borehole seismic data recorded can provide calibrated, high resolution data that can be used alone, or in conjunction with surface seismic data in order to make exploration decisions and thus is a valuable technique for well characterisation. An additional application for borehole seismic logging tools is the monitoring of hydraulic fracturing (“fracking”) sites.
The quality of workpiece is depends on by the thermo-elastic behavior of the machine tool during the production process. Machine tool deformations occur due to waste heat from motors and frictional heat from guides, joints and the tool, while coolants act to reduce this influx of heat. Additional thermal influ- ences come from the machine tool’s environment and foundation. This leads to inhomogeneous, transient temperature fields inside the machine tool which dis- place the tool center point (TCP) and thus reduce production accuracy and finally the product quality . Next to approximation strategies such as characteristic diagram based correction as in  and structure model based correction shown in , the most reliable way to predict the TCP displacement is via structure-me- chanical finite element (FE) simulation. A CAD model of a given machine tool serves as the basis for this approach. On it a FE mesh is created. After establishing the partial differential equations (PDEs) describing the heat transfer within the machine tool and with its surroundings, FE simulations are run in order to obtain the temperature fields of the machine tool for specified load regimes. Using lin- ear thermo-elastic expansion, the deformation can then be calculated from each temperature field and the displacement of the TCP read from this deformation field, see . The accuracy of this latter approach depends on the correct mod- elling of the heat flux within the machine tool and the exchange with its sur- roundings. In order to calculate the correct amount of heat being exchanged with the environment, one may use known parameters from well-established tables. However, if the surrounding air is in motion or otherwise changing, computa- tional fluid dynamics (CFD) simulations are required to accurately determine these transient parameters. This two-step approach makes realistic thermo-elas- tic simulations particularly complicated and time-consuming. Negative aspect of this approach is the very computing time intensive CFD simulation. Some meth- ods aiming at real-time thermo-elastic simulations based on model order reduc- tion must therefore rely on the inaccurate predetermined parameter sets . This could be supported if all the necessary CFD simulations could be run in ad- vance and supplied to the thermo-elastic models when they are needed. Nevertheless, the whole output of this CFD simulations is too much amount of data for an effective computation of the correction steps. Therefore, a reduction of this data is desirable wherefore the ideas of this paper comes up.
field through the interaction of attraction and exclusion is similar to objective of the fuzzyclustering. There is a greater repulsive force between different clusters that makes clusters more separate and data in the same cluster more compact, which can reduce the impact of weak noise points. In addition, the theory of data field based on the concept of physical field can help us to get the center of the potential field in the data space. The number and location of the potential centers offer a very important reference for determining the number of clusters and selecting the initial cluster centers. This can improve the defect of traditional FCM algorithm that the initial clustering centers is overly sensitive. Kumar etc.  proposed a hybrid approach to solve the problem of random initialization of cluster centers. They used BBO—a population based evolutionary algorithm which motivated by migration mechanism of ecosystems. Aldahdooh and Ashour  proposed a selection method for initial cluster centroid in K-meansclustering instead of the random selection method. In recent years, under the inspiration of social and natural laws, the researchers have combined fuzzyclustering algorithm with other algorithms such as genetic algorithm, ant colony optimization algorithm and particle swarm algorithm and achieved great success. The EFCM algorithm proposed in this paper improves the selection process of initial clustering centers and iterative updating process of traditional FCM algorithm based on the theory of charge interaction, Coulomb's law and data field theory, which makes the algorithm more realistic. Experiments can be used to prove the advantages of the improved EFCM algorithm compared to the traditional FCM algorithm.
Currently, the fuzzyc-means algorithm plays a certain role in remote sensing image classification. However, it is easy to fall into local optimal solution, which leads to poor classification. In order to improve the accuracy of classification, this paper, based on the improved marked watershed segmentation, puts forward a fuzzyc-meansclustering optimization algorithm. Because the watershed segmentation and fuzzyc-meansclustering are sensitive to the noise of the image, this paper uses the adaptive median filtering algorithm to eliminate the noise information. During this process, the classification numbers and initial cluster centers of fuzzyc-means are determined by the result of the fuzzy similar relation clustering. Through a series of comparative simulation experiments, the results show that the method proposed in this paper is more accurate than the ISODATA method, and it is a feasible training method.
pling would allow for measurements of the dynamic evolu- tion of plumes, and feature tracking could then be used as a means to determine gas emission rates. A more fundamen- tal limitation is the NE1T of the spectrally filtered channels. There are cameras available commercially with NE1T ’s of < 20 mK and 60 Hz frame rates that can provide retrieval er- rors in SCDs below 10 %. Many improvements to the system can be envisaged. By viewing a target using three cameras ar- ranged with an angular spacing of 120 ◦ , a three-dimensional image could be acquired and quantitative measures of plume dimensions and plume morphology derived. Addition of fil- ters centred at different wavelengths would also permit a range of other gases to be measured. The camera could also be used in atmospheric research for studies of the radiative effects of clouds on the Earth’s radiation balance (Smith and Toumi, 2008) and to image toxic gases from industrial acci- dents or from deliberate gas releases, where personal safety is a major issue.
Clustering can be considered as the most important unsupervised learning problem. Clustering algorithms try to partition a set of unlabeled input data into a number of clusters such that data in the same cluster are more similar to each other than to data in the other clusters . Clustering has been applied in a wide variety of fields ranging from engineering (machine learning, artificial intelligence, pattern recognition, mechanical engineering, electrical engineering) [2-4], computer sciences (web mining, spatial database, analysis, textual document collection, image segmentation) [5,6], life and medical sciences (genetics, biology, microbiology, paleontology, psychiatry, clinic, pathology) [7-9], to earth sciences (geography, geology, remote sensing) , social sciences (sociology, psychology, archeology, education) and economics (marketing, business) [2, 10]. A large number of clustering algorithms have been proposed for various applications. These may be roughly categorized into two
The current study focuses on a spindle system of a box- type precision CNC coordinate boring machine. Thermal balance experiments were performed using a temperature displacement acquisition system to measure the distribution of the temperature field and thermal deformation at different spindle speeds. The study analyzes how different spindle speeds affect thermal characteristics, then usingfuzzyclustering regression analysis method to optimize the temperature variables, selected the variables for thermalerror- sensitive, finally the MIMO artificial neural network approaches were established for spindle axial thermal elongation and radial thermal tilts. Subsequently, a new set of sample data is used to validate the model. The results indicate that the model has high prediction accuracy with perfect generalizations; one can obtain an exact model for subsequent thermalerror compensation that provides references for the characteristic parameter for thermal equilibrium.
The Electroencephalogram (EEG) signal is a voltage signal arising from synchronized neural activity. EEG can be used to classify different mental states and to find abnormalities in neural activity. To check the abnormality in neural activity, EEG signal is classified using classifiers. In this project k-meansclustering and fuzzycmeans (FCM) clustering is used to cluster the input data set to Neural network. NeuroIntelligence is a neural network tool used to classify unknown data points. The non linear time series (NLTS) data set is initially clustered into Normal or Abnormal categories using k-means or FCM clustering methods. This clustered data set is used to train neural network. When an unknown EEG signal is taken, first NLTS measurements are extracted and input to trained neural network to classify the EEG signal. This method of classification proposed is unique and is very easy to classify EEG signals.
slim blood stream vessels may result in little round spots that are regionally just like MAs, both in style. Vessel segments may be turned off from the general shrub, and appear as little, black things of various forms. Almost every state-of-the-art technique views some type of image preprocessing phase, which usually includes disturbance decrease, filtering or colour modification. Retinal pictures have the largest comparison in the natural channel; accordingly it is a common practice to use the natural route for segmentation reasons. For noise decrease, convolution with Gaussian covers and median filtering are commonly used techniques. The number of pixels to be prepared is considerably decreased by only considering the local maxima of the preprocessed picture. We implement optimum recognition on each information, and determine a set of principles that explain the size, size, and form of the main optimum. The fundus picture features are produced with the success as 99, 94 and 100% for hard drive localization, hard drive border recognition and fovea localization re-spectively. These designs can be enhanced in bigger databases and also used for medical reasons. The area growing segmentation technique gives the good segmentation result in order to specify the area with appropriate factors. It takes too lots of your energy and effort to complete the clustering process, so it is expensive. The region splitting and consolidating technique will divided the pictures until the appropriate quality is achieved . It is not suitable for more variety of pictures prepared simultaneously. Watershed is the edge based picture segmentation technique provides a huge variety of segmented pictures with high reliability which also experiences in over segmentation. Unclear C indicates (FCM) is a details clustering technique in which a details set is arranged into ‘n’ groups with every details point in the dataset which belongs to every group to a certain degree. A conventional FCM criteria does not incorporate the spatial details which makes it delicate to disturbance and other picture relics whereas Spatial Unclear Cmeansclustering criteria features the spatial information into the account function for clustering. The Customized Spatial Unclear C-Meansclustering method is used to identify glaucoma which is existing in the retina with various spatial harmonizes.
Many approaches have been proposed to segment masses from surrounding tissues in digital mammograms. Mahfuzah Mustafa and al , used Chan-Vese Active Contour and Localized Active Contour for segmenting lesions in digitized mammogram images, the effectiveness of these techniques are then compared, the results obtained by Chan-Vese Active Contour are proven to be better than the Localized Active Contour method. J. Quintanilla et al , proposed mathematical morphology to enhance potential MCs. Afterwards, three algorithms (FuzzyC-Means, K-Means, and Possibilistic Fuzzyc-Means) are used and compared in order to segment ROIs images, trying to improve the results of microcalcifications cluster detection.
Fuzzyclustering introduces the concept of membership into data partition, for this reason that membership can indicate the degree to which an object belongs to the clusters definitely, and actually represents the data partition more clearly.. Study of cluster or segmentation technique ,  based on clustering assembles a set of entities in a manner that entities in the identical cluster have a superior degree of alikeness to each compared to the other clusters. Clusters are defined as contiguous regions of more than one-dimensional space comprising comparative points of high density, alienated from other exemplary regions comprising moderate points of low density. In image breakdown, clustering is the order of arrangement of pixels conferring to more or less features like intensity. Under hard clustering, data elements fit into one cluster simply and the membership value of belongingness to a cluster is precisely one. Under soft clustering, elements of data fit into more prominent than the single cluster and the membership value of belongingness to the cluster varies from 0 to 1.
Abstract—The problem of mining a high dimensional data includes a high computational cost, a high dimensional dataset composed of thousands of attribute and or instances. The efficiency of an algorithm, specifically, its speed is oftentimes sacrificed when this kind of dataset is supplied to the algorithm. FuzzyC-Means algorithm is one which suffers from this problem. This clustering algorithm requires high computational resources as it processes whether low or high dimensional data. Netflix data rating, small round blue cell tumors (SRBCTs) and Colon Cancer (52,308, and 2,000 of attributes and 1500, 83 and 62 of instances respectively) dataset were identified as a high dimensional dataset. As such, the Manhattan distance measure employing the trigonometric function was used to enhance the fuzzyc-means algorithm. Results show an increase on the efficiency of processing large amount of data using the Netflix ,Colon cancer and SRCBT an (39,296, 38,952 and 85,774 milliseconds to complete the different clusters, respectively) average of 54,674 milliseconds while Manhattan distance measure took an average of (36,858, 36,501 and 82,86 milliseconds, respectively) 52,703 milliseconds for the entire dataset to cluster. On the other hand, the enhanced Manhattan distance measure took (33,216, 32,368 and 81,125 milliseconds, respectively) 48,903 seconds on clustering the datasets. Given the said result, the enhanced Manhattan distance measure is 11% more efficient compared to Euclidean distance measure and 7% more efficient than the Manhattan distance measure respectively.
To test the procedure the TCP-displacements and boundary conditions for a thermal load case during the operating time is measured. For a finishing operation we reproduced the compensation procedure on standard PC which has nearly the same power as available on a standard machine tool control. As time step, to update the compensation values, ten minutes was chosen, based on measurements previously carried out. In Figure 6 the calculated thermally induced TCP-displacements after six hours machining time with an augmentation factor of 6’000 are shown. These calculated values are compared with measurements at five locations in the work space and have shown a good correlation. After numerically applying the compensation scheme based on location and component errors the remaining maximum TCP-displacements have been reduced by a ratio of 50.