Back Propagation Network Approach Research Articles

Improving image resolution using deabsorption prestack time migration with effective Q estimation: A back-propagation network approach

The conventional time migration method does not consider the attenuation caused by the viscoelasticity of the underground media during the imaging process. Therefore, the final imaging amplitude and phase include inaccuracies caused by attenuation. In this study, we develop a migration scheme to compensate for absorption and dispersion using an effective quality factor ( Q) estimation based on a back-propagation (BP) neural network. We use BP neural network technology to automatically estimate the effective Q value from stacked imaging data, thereby avoiding manual Q estimation using conventional methods. Our scheme can be incorporated into conventional seismic data-processing workflows. Furthermore, synthetic and field data sets are used to validate our scheme, which is used to acquire high-resolution images with low noise levels. In addition to developing a completely data-driven Q-value estimation strategy, this study demonstrates close integration of artificial intelligence, data mining, and conventional geophysics; our approach is appropriate for estimating the effective Q and has strong industrial application value and significance.

Fuzzified deep neural network ensemble approach for estimating cycle time range

Because of the high uncertainty associated with predicting the cycle time of a job in a complex manufacturing system, this task is a challenge for production planners. To replace an inaccurate cycle time forecast, the range of cycle time is estimated. To estimate the cycle time range of a job as precisely as possible, a fuzzified deep neural network (FDNN) ensemble approach is proposed in this paper. This approach involves the following steps. First, a deep neural network (DNN) is constructed to predict the cycle time of a job. The parameters of the DNN are then fuzzified to generate a fuzzy cycle time forecast that contains the actual cycle time. In contrast to existing methods that involve fuzzifying multiple parameters simultaneously, which is computationally intensive, the proposed methodology involves fuzzifying parameters independently to facilitate problem-solving. Subsequently, the cycle time ranges estimated by fuzzifying various parameters are aggregated through fuzzy intersection. The proposed FDNN ensemble approach was applied to a real case to validate its effectiveness. The experimental results indicated that the precision of cycle time range estimation was up to 38% higher with the FDNN ensemble approach than with the fuzzy backpropagation network approach.

A Selectively Fuzzified Back Propagation Network Approach for Precisely Estimating the Cycle Time Range in Wafer Fabrication

Forecasting the cycle time of each job is a critical task for a factory. However, recent studies have shown that it is a challenging task, even with state-of-the-art deep learning techniques. To address this challenge, a selectively fuzzified back propagation network (SFBPN) approach is proposed to estimate the range of a cycle time, the results of which provide valuable information for many managerial purposes. The SFBPN approach is distinct from existing methods, because the thresholds on both the hidden and output layers of a back propagation network are fuzzified to tighten the range of a cycle time, while most of the existing methods only fuzzify the threshold on the output node. In addition, a random search and local optimization algorithm is also proposed to derive the optimal values of the fuzzy thresholds. The proposed methodology is applied to a real case from the literature. The experimental results show that the proposed methodology improved the forecasting precision by up to 65%.

Fuzzy back-propagation network approach for estimating the simulation workload

Estimating the time required for simulating a factory online is a crucial topic in manufacturing. However, this topic has rarely been discussed. A fuzzy back-propagation network (FBPN) approach for estimating the workload of a simulation task according to the required simulation time is proposed in this paper. In the proposed FBPN approach, tasks for which simulation times cannot be effectively estimated are considered as outliers and excluded, and therefore, more appropriate lower and upper bounds on the simulation time can be set. Thus, the ranges of all simulation times can be efficiently narrowed; however, this decrease is achieved at the expense of a slight decrease in the hit rate, which is still tolerable. A real case containing data of 90 simulation tasks was used to validate the proposed methodology. In addition, two existing FBPN methods, the adaptive-network-based fuzzy inference system method and the Chen (Comput Ind Eng 66:834---848, [2013]) method, were applied to these tasks for comparison. The experimental results showed that the proposed methodology was superior to the two existing FBPN methods in estimating precision and accuracy.

A fuzzy back-propagation network approach for planning actions to shorten the cycle time of a job in dynamic random access memory manufacturing

Reducing cycle time is an essential task that enables dynamic random access memory (DRAM) manufactures to maintain sustainability and gain competitiveness. However, the uncertainty of cycle times makes this a challenging task. Overestimating or underestimating cycle times leads to an incorrect assessment of the effects of an action performed to reduce the cycle time. To address this problem, in this study, the uncertainty of the cycle time was considered and modeled using a fuzzy value. A fuzzy back-propagation network (FBPN) approach is proposed to estimate the cycle time of a job based on its attributes and factory conditions. The lower and upper bounds of the cycle time established using the FBPN approach are tight. In addition, an FBPN can be applied to assess the effects of a cycle time reduction action that shortens the cycle time of a job by improving control over factory conditions. The control action is flexible because the attitude of the managers is considered. Furthermore, a control mechanism for multiple consecutive jobs is established. To illustrate the applicability of the proposed methodology, a DRAM factory simulator was used to generate data. According to the experimental results, the ranges of cycle times, including the ±3? range, determined using the FBPN approach were narrower than those determined using four existing methods. In addition, the relationship between the cycle time and attributes of a job was determined to be different when the upper (or lower) bound of the cycle time rather than the most likely value was considered.

CART–BPN approach for estimating cycle time in wafer fabrication

Cycle-time management plays a crucial role in improving the performance of a wafer-fabrication factory, beginning with the estimation of the cycle time of each job. Although this topic has been widely investigated, several problems still need to be addressed, such as how to classify jobs suitable for the same estimation mechanism into the same group. Most existing methods classify jobs by their attributes; however, the differences between the attributes of various jobs may not be reflected in their cycle times. The biobjective nature of a classification and regression tree (CART) makes it particularly suitable for resolving this problem. However, in a CART, the cycle times of jobs of a branch are estimated with the same value, which is inexact. Hence, this study proposes a joint use of a CART and back propagation network (BPN), in which the BPN is constructed to estimate the cycle times of jobs of a branch. A real case was used to evaluate the effectiveness of the proposed methodology. The experimental results supported the superiority of the proposed methodology over existing methods. In addition, the managerial implications of the proposed methodology are also discussed.

An Agent-Based Fuzzy Collaborative Intelligence Approach for Precise and Accurate Semiconductor Yield Forecasting

Yield forecasting is an important task for the manufacturer of semiconductors. Owing to the uncertainty in yield learning, it is, however, often difficult to make precise and accurate yield forecasts. To solve this problem, we propose an agent-based fuzzy collaborative intelligence approach that is modified from the fuzzy linear regression and back propagation network approach. In the proposed methodology, software agents rather than domain experts are used to improve the efficiency of collaboration. In addition, an agent decides the adjustable parameters by referencing to others so that the overall prediction performance can be improved in an effective way. In addition, we proposed a simple and effective way to aggregate the fuzzy forecasts by agents. Compared with the fuzzy linear regression and back propagation network approach, the proposed methodology reduced the average range and mean absolute percentage error by 18% and 99%, respectively.

A fuzzy rule for job dispatching in a wafer fabrication factory—a simulation study

Job scheduling in wafer fabrication factories is subject to many sources of uncertainty or randomness. To consider the uncertainty and improve the scheduling performance in a wafer fabrication factory, this paper proposes an innovative fuzzy rule that solves the problem of slack overlapping in a non-subjective way. The fuzzy rule considers the uncertainty in the remaining cycle time and is aimed at the simultaneous optimization of the average cycle time and cycle time standard deviation. Few existing publications discuss this issue. A systematic procedure has also been established to optimize the value of the adjustable parameter in the fuzzy rule. On the other hand, a novel fuzzy back propagation network approach is also proposed in this paper to estimate the remaining cycle time accurately. The performance of the proposed methodology is evaluated with a series of production simulation experiments.

A Fuzzy Collaborative Forecasting Approach for Forecasting the Productivity of a Factory

Productivity is always considered as one of the most basic and important factors to the competitiveness of a factory. For this reason, all factories have sought to enhance productivity. To achieve this goal, we first need to estimate the productivity. However, there is considerable degree of uncertainty in productivity. For this reason, a fuzzy collaborative forecasting approach is proposed in this study to forecast the productivity of a factory. First, a learning model is established to estimate the future productivity. Subsequently, the learning model is converted into three equivalent nonlinear programming problems to be solved from various viewpoints. The fuzzy productivity forecasts by different experts may not be equal and should therefore be aggregated. To this end, the fuzzy intersection and back propagation network approach is applied. The practical example of a dynamic random access memory (DRAM) factory is used to evaluate the effectiveness of the proposed methodology.

Forecasting the Long-Term Electricity Demand in Taiwan with a Hybrid FLR and BPN Approach

Precisely and accurately predict the electricity demand is an important task for the government in each country. In addition, establishing the lowest upper bound for the electricity demand also avoids unnecessary power plant investment. To this end, a hybrid fuzzy linear regression (FLR) and back propagation network (BPN) approach is proposed in this study. In the proposed methodology, multiple experts construct their own FLR equations to predict the future electricity demand from various viewpoints. Each FLR equation can be fitted by solving two equivalent nonlinear programming problems, based on the opinions of experts. In order to aggregate these fuzzy electricity demand forecasts, a two-step aggregation mechanism is used. First, fuzzy intersection is applied to aggregate the fuzzy electricity demand forecasts into a polygon-shaped fuzzy number, in order to improve the precision. Subsequently, a BPN is constructed to defuzzify the polygon-shaped fuzzy number and generate a representative/crisp value, in order to enhance the accuracy. The actual case of Taiwan is used to evaluate the effectiveness of the proposed methodology. According to the experimental results, the proposed methodology improved the precision and accuracy of the electricity demand forecasting by 33% and 99%, respectively.

Applying the hybrid fuzzy c-means-back propagation network approach to forecast the effective cost per die of a semiconductor product

Forecasting the unit cost of every product type in a factory is an important task to the factory. After the unit cost of every product type in a factory is accurately forecasted, several managerial goals (including pricing, cost down projecting, capacity planning, ordering decision support, and guiding subsequent operations) can be simultaneously achieved. However, it is not easy to deal with the uncertainty in the unit cost. In addition, most references in this field were focused on costing and seldom investigated the forecasting of the unit cost. To tackle these problems, the hybrid fuzzy linear regression (FLR) and back propagation network (BPN) approach is applied to forecast the unit cost of every product type in a wafer fabrication plant, which is usually referred to as the determination of the effective cost per die. In practical situations the long-term effective cost per die of a product type is usually approximated with a linear regression (LR) equation, according to the “continuous cost down” philosophy, which is prone to error. Conversely, the proposed FLR–BPN approach is more accurate and be able to deal with the uncertainty in the unit cost in a simple and intuitive way. For evaluating the effectiveness of the proposed methodology, a demonstrative case was used. Experimental results showed that the hybrid FLR–BPN approach was superior to some existing approaches in forecasting accuracy and precision.

A bi-criteria nonlinear fluctuation smoothing rule incorporating the SOM–FBPN remaining cycle time estimator for scheduling a wafer fab—a simulation study

This paper proposes a bi-criteria nonlinear fluctuation smoothing rule to further improve the performance of job scheduling in a wafer fabrication factory (wafer fab). The rule is based on the well-known fluctuation smoothing rules. First, the remaining cycle time of a job is estimated by applying the self-organization map–fuzzy back propagation network approach to improve the estimation accuracy. Second, two nonlinear forms of the fluctuation smoothing rules are obtained to enhance the balance and responsiveness. Third, the two nonlinear fluctuation smoothing rules are merged into a bi-criteria rule for considering two performance measures (average cycle time and cycle time variation) at the same time. Finally, the content of the bi-criteria rule can be tailored for the wafer fab and be scheduled with an adjustable factor. To evaluate the effectiveness of the proposed methodology, a production simulation was conducted. According to the experimental results, the proposed methodology outperformed some of the existing approaches by reducing the average cycle time and cycle time variation at the same time. In addition, the experimental results showed that the bi-criteria rule made it possible to improve one performance measure without raising the expense of another one.

An optimized tailored nonlinear fluctuation smoothing rule for scheduling a semiconductor manufacturing factory

An optimized tailored nonlinear fluctuation smoothing rule is proposed in this study to improve the performance of scheduling jobs in a semiconductor manufacturing factory. The tailored nonlinear fluctuation smoothing rule is modified from the well-known fluctuation smoothing rules with some innovative treatments. At first, the fuzzy c-means and fuzzy back propagation network approach is applied to estimate the remaining cycle time of every job. Then the nonlinear fluctuation smoothing rules are constructed and tailored to the semiconductor manufacturing factory to be scheduled. Finally, the back propagation network optimization approach is proposed to optimize the rule. To evaluate the effectiveness of the proposed methodology, production simulation is also applied in this study to generate some test data. According to experimental results, the proposed methodology outperformed nine existing approaches in reducing the cycle time averages and standard deviations.

A nonlinear scheduling rule incorporating fuzzy-neural remaining cycle time estimator for scheduling a semiconductor manufacturing factory—a simulation study

A nonlinear scheduling rule incorporating a fuzzy-neural remaining cycle time estimator is proposed in this study to improve scheduling performance in a semiconductor manufacturing factory. The proposed scheduling rule is modified from the well-known fluctuation smoothing rule with three treatments. At first, the look-ahead self-organization map-fuzzy back-propagation network approach in our previous study is used to estimate the remaining cycle time of every job in the semiconductor manufacturing factory. Subsequently, the release time and remaining cycle time of a job are both normalized to balance their importance in the fluctuation smoothing rule. Finally, the normalized release time is divided by the normalized remaining cycle time to obtain the slack. In this way, the proposed scheduling rule becomes a nonlinear one. To evaluate the effectiveness of the proposed methodology, production simulation is used to generate some test data. According to experimental results, the proposed methodology outperformed many existing approaches in reducing both the average cycle times and cycle time standard deviations. The advantage was up to 41% over the basis p-FS policy when the cycle time standard deviations were to be minimized.

Incorporating fuzzy c-means and a back-propagation network ensemble to job completion time prediction in a semiconductor fabrication factory

Job completion time prediction is a critical task to a semiconductor fabrication factory. To further enhance the effectiveness/accuracy of job completion time prediction in a semiconductor fabrication factory, a hybrid fuzzy c-means (FCM) and back propagation network (BPN) approach is proposed in this study. In the proposed FCM–BPN approach, input examples are firstly pre-classified with FCM before they are fed into the BPN. Then, examples belonging to different categories are learned with different BPNs but with the same topology. After learning, these BPNs form a BPN ensemble that can be applied to predict the completion time of a new job. The output of the BPN ensemble is derived by aggregating the outputs from the component BPNs with another BPN and determines the completion time forecast. To validate the effectiveness of the proposed methodology and to make comparison with some existing approaches, the actual data in a semiconductor fabrication factory were collected. According to experimental results, the prediction accuracy of the proposed methodology was significantly better than those of some existing approaches. Besides, applying the fuzzy set theory was shown to be very effective in forming job categories and in deriving a representative value from the BPN ensemble. Both contributed to the superiority of the proposed methodology.

Source localization of focal ventricular arrhythmias using linear estimation, correlation, and back propagation networks

Catheter-based approaches used in the localization and treatment of the source of heart rhythm disturbances (arrhythmias) have become popular, because they do not require highly invasive and risky open-chest operations. In most of the existing approaches, mapping of the outer surface (epicardium) is not possible even though arrhythmic substrates involving epicardial and subepicardial layers account for about 15% of the ventricular tachycardias. In this study, we report a feasibility study of a novel mapping approach targeting the epicardium which is based on the measurements of multielectrode catheters placed in the coronary veins. We investigated three methods in determining the most probable region of early activation, i.e., the region that contains the source of the abnormal activation on the heart, using only a set of sparse venous catheter recordings. The methods we proposed here were the linear estimation, correlation, and the back propagation networks. The linear estimation technique hypothesized the relationship between venous catheter measurements and unmeasured epicardial sites based on a previously recorded training data set. The correlation method included a comparative analysis between test and training epicardial activation time maps based on the measured values from the venous sites. In the back propagation method, the input layer consisted of the source data in the form of 42 nodes which were the activation time values from the intravenous catheter leads. We used two hidden layers with 600 and 500 nodes, respectively. The output layer consisted of 28 nodes in the output layer that corresponded to the manually selected early activation regions on the epicardium. The results of the linear estimation and the correlation methods showed that they could be used as a good predictor for the region of early activation, and thus, these approaches may be employed to direct a subsequent more focused electrophysiological study and curative radiofrequency (RF) ablation. The back propagation network approach performed relatively well for the right ventricularly paced beats and the results demonstrated its potential as a supporting technique to the estimation and correlation methods. The results of this study encourage further investigation and provides evidence that an epicardial mapping approach based on the venous catheter recordings is feasible and can provide adequate accuracy for clinical applications.

Determination of aquifer parameters using radial basis function network approach

In this paper, a radial basis function network (RBFN) approach to the determination of aquifer parameters is developed. The approach is based on the combination of an RBFN and the Theis or the Hantuch‐Jacob solution. The proposed RBFN approach has advantages over the existing back‐propagation network (BPN) approach. It avoids inappropriate setting of a trained range. It also determines the aquifer parameters more accurately and needs less training time. Testing the BPN and RBFN approaches by synthetic data also demonstrates these advantages. As to the comparison between the RBFN approach and the type‐curve graphical method, two applications to actual time‐drawdown data show that the RBFN approach determines the aquifer parameters more precisely for both nonleaky‐confined and leaky‐confined aquifers. The RBFN approach is recommended as an alternative to the type‐curve graphical method and the existing BPN approach.

A neural network architecture for noise prediction

This paper presents application of neural networks (NNs) to the problem of prediction of noise caused by urban traffic. The most representative physical variable quantifying noise emissions is the equivalent sound pressure level. Up to now it has been identified on the basis of semi-empirical models, typically regression analysis, which generally do not provide very accurate approximations of the trend followed by sound pressure level. The authors have attempted to overcome this difficulty by adopting a neural approach based on a Backpropagation Network (BPN). Results obtained by the comparison of the BPN approach with those provided by selected relationships found in relevant literature, show how good is the approach proposed. The neural solution to the problem has shown the necessity, in certain phases, of a set of acoustic measurements which is as free as possible of error. The complexity of error identification by means of classical approaches has led the authors to explore the possibility of a neural solution to this problem as well. The authors therefore propose use of a neural architecture made up of two cascading levels. At the first level a supervised classifying network, the learning vector quantization (LVQ) network, filters the data discarding all the wrong measurements, while at the second level the BPN predicts the sound pressure level.