Repetitive Learning Research Articles

Runout rejection in end milling through two-dimensional repetitive force control

Cutter runout presents an impediment to the attainable productivity and part quality in machining processes. This study establishes a mechatronic approach to compensate for cutter runout effect during multi-flute machining in real-time. It is accomplished through measuring the runout-related cutting force component, formulating repetitive controlling commands, and manipulating the workpiece position to counteract the variation of chip load during the course of machining. In an end milling configuration, the approach has been realized with a two-axis test bed attached to the machining table. The test bed operates independently to the computer numerical control (CNC) servo loops of the machine tool, and it provides a fine range and high frequency motion of the workpiece. In this paper the development of the test bed is discussed in terms of the hardware component design and the software controller formulation. Numerical simulation of the cutting force in response to repetitive learning control at the presence of runout disturbance is examined. To evaluate the feed-back system performance, experimental study based on the implementation of both single- and two-dimensional controls is presented in the context of cutting force regulation and part surface finish improvement.

In-Process Compensation for Milling Cutter Runout via Chip Load Manipulation

This paper discusses a real-time chip load compensation methodology for the elimination of cutting force oscillation and machined surface scalloping due to cutter runout so as to gain better utilization of machine tools. The concept and implementation of the methodology is illustrated using end milling as a process of example. In this work a force feedback system was discussed in the angle domain based upon a proportional-integral control strategy and a repetitive learning control strategy to actively manipulate the chip load during end milling. Numerical simulations based on experimentally identified machining dynamics were presented to compare the performance of the two control schemes. Experimental investigation under various cutting conditions was performed to assess the viability of the feedback compensation system in the context of cutting force response as well as machined surface finish. It has been shown that a proportional-integral control has limited effectiveness in eliminating the runout-induced cutting force oscillation due to the constraints of system stability and dynamic performance. On the other hand, the learning control system based on the internal model principal successfully yields a cutting force free of oscillatory components at the spindle frequency and significantly improves the quality of machined surfaces by cancelling the nonasymptotically stable dynamics of cutter runout.

Implementating a hybrid learning force control scheme

The theory and implementation of a repetitive learning control algorithm for hybrid position and force control of robotic manipulators is presented here. The complete control system involves learning position control for translational motion tangent to an unknown surface, learning force control normal to the surface, and learning orientation control using torque feedback to maintain tangential motion relative to the surface. An IBM 7545 robot equipped with a wrist force/torque sensor is used to evaluate the performance of the proposed controller. >

適応型繰り返し制御を用いた光ディスクのトラッキング制御システム

In response to the recent demand for high-density recording on an optical disk, the industry is now “going narrow track”. We have developed an adaptive learing control method capable of changing its learning according to track correlation, thus making it possible to avoid unnecessary learning during equipment vibration to avoid the lack of stability that is a problem for learning control, and to increase follow-up ability. As a result we have confirmed that track-error can be reduced to the equivalent noise level (0.1μm) and that stabilized operation with repetitive learning control system is possible.

Remedial program for diagnostic radiology residents.

We describe a 1-year remedial program for radiology residents with deficiencies in basic medical sciences that cause difficulty in the understanding and application of basic radiologic principles. Six of 107 residents participated in the remedial program at separate times during an 8-year period. The design of the remedial program was based on the following: 1) reassignment to first year basic radiology rotations with 1-year delayed advancement to the subspecialty rotations; 2) comprehensive radiology texts that contain a detailed review of related embryology, anatomy, physiology and pathology; 3) instruction in the study of the texts; and 4) repetitive learning accomplished by intermittent reassignment of rotations in which difficulty was demonstrated. Resident performance was evaluated by correlation of monthly written faculty evaluations, the results of the American College of Radiology Intraining Examination, and the American Board of Radiology (ABR) written and oral examinations. During the period of remediation, faculty evaluation revealed improved performance in all six residents from below-average ratings (mean = 2.3) to above-average and superior ratings (mean = 8.3). These results were correlated with the American College of Radiology Intraining Examination scores, which improved from the lowest quartile (range 0% to 28%) to the highest (range 78% to 99%). Each resident successfully passed the ABR written and oral boards on the first attempt. Two additional residents who were offered but did not accept remediation eventually dropped out of the program or repeatedly failed the written ABR examination. Ultimately, strengthening basic medical science knowledge facilitated interpretation of radiographs by enabling the integration of basic radiologic and scientific principles and understanding the underlying basic medical, radiologic, and clinical significance.

Suppression control of speed variation in an induction motor with fluctuating load by repetitive learning control

AbstractInduction motors are robust and inexpensive machines and are used widely for variable speed control because of the recent development of electronic technique. In the case where loads of the motors are compressors, pumps, and so on, the constant V/F control of the induction motors usually is employed because it is difficult to install speed sensors and accurate speed control is not required. In such loads, the rotational speed of the motors fluctuates considerably because the load torque is pulsated. When the frequency of the torque pulsation is close to the resonant frequency of the mechanical system, large vibration and acoustic noise are produced especially in the low‐frequency region.This paper proposes a method to suppress the variation of the rotational speed of the V/F controlled induction motor with a fluctuated load by feedforward compensation using a timing sensor of 1 pulse/rev, considering that the load torque varies periodically. The feedforward data by a period for the compensation is obtained by the learning control based on the repetitive control in which the motor speed is controlled by periodically reflecting the past speed error on the present V/F input to the inverter. Effectiveness of the proposed method is confirmed by approximate analysis, simulations and experiments.

Suppression Control of Speed Variation of the Induction Motor with Fluctuating Load by Repetitive Learning Control.

Induction motors are robust and inexpensive machines and are widely used for variable speed control because of recent development of electronic technique. In the case that loads of the motors are compressors, pumps, and so on, the constant V/F control of the induction motors is usually employed because it is difficult to install speed sensors and accurate speed control is not required. In such loads, the rotational speed of the motors is considerably fluctuated because the load torque is pulsated. When frequency of the torque pulsation is close to the resonant frequency of the mechanical system, large vibration and acoustic noise are produced especially in low frequency region.In this paper, authors propose a method to suppress the variation of the rotational speed of the V/F controlled induction motor with a fluctuated load by feedforward compensation using a timing sensor of 1 pulse/rev. considering that the load torque varies periodically. The feedforward data by a period for the compensation is obtained by the learning control based on the repetitive control in which the motor speed is controlled by periodically reflecting the past speed error on the present V/F input to the inverter. Effectiveness of the proposed method is confirmed by approximate analysis, simulations and experiments.

A new repetitive controller for mechanical manipulators

AbstractA new repetitive learning controller for motion control of mechanical manipulators undergoing periodic tasks is developed. This controller does not require exact knowledge of the manipulator dynamic structure or its parameters, and is computationally efficient. In addition, no actual joint accelerations or any matrix inversions are needed in the control law. The global asymptotic stability of the ideal and the robust stability of the nonideal control system is proven, taking into account the full nonlinear dynamics of the manipulator. Simulation results of this algorithm applied to a realistic Scara type manipulator, which includes dry friction, pay‐load inertia variations, actuator/sensor noise, and unmodelled dynamics are also presented.

反復学習に見られる視覚記憶と聴覚記憶の相互作用

A serial recall experiment under same- and cross-modality conditions was carried out according to Hebb's (1961) paradigm. Subjects memorized strings of nine digits in 18 trials, which were divided into early trials and late ones. Under the same-modality condition the strings used in the 18 trials were presented through the same modality, i.e., visual or auditory, whereas under the cross-modality condition the strings used in the early and late trials were presented through two different modalities. A repeated string and unrepeated strings were presented 3 and 6 times respectively in both early and late trials. The results showed that (1) repetitive learning in the same-modality occurred for both visual and auditory; (2) there was a non-additivity of cross-modal repetitive learning from auditory to visual, whereas there was an additivity of cross-modal repetitive learning from visual to auditory. We inferred that the visual inputs received both visual coding and auditory coding through an audio-transformation so as to be separately memorized in visual and auditory memories, whereas the auditory inputs were memorized in the auditory memory through the auditory coding alone.

Learning control theory for robotic motion

AbstractA new concept of learning control for the improvement of robot motions is proposed, which can be referred to a mathematical modelling of learning and generation of motor programmes in the central nervous system. It differs from conventional classical and modern control techniques. It stands for the repeatability of operating a given robotic system and the possibility of improving the command input on the basis of actual measurement data acquired at the previous operation. Hence adequate conditions on the repeatability and invariance of the system dynamics are assumed, but no precise description of the dynamics is required for construction of the learning algorithms. Two types of iterative learning algorithm are proposed: one uses a PD‐type (proportional and differential) update of input commands and the other a PI‐type (proportional and integral) update where velocity signals are regarded as outputs. It is shown that in both types a better performance is realized at every attempt of operation, provided a desired motion is given a priori and the actual motion (velocity signals) can be measured at every operation. Further, the robustness of such learning control algorithms with respect to the existence of perturbed errors of initialization of the robot, disturbances and measurement noise during operation is analysed in detail. It is shown that in PD‐type learning laws such errors are neither amplified nor aggregated in later consecutive trials of operation. In the case of PI‐type learning laws it is shown that such a robustness property is assured if a forgetting factor is adequately introduced into the repetitive learning law.

Paradoxical fall in repetitive learning of a random series of digits

In this study, on each trial subjects were presented visually a different random series of nine digits, and required to memorize and recall them immediately. Among a sequence of trials there were three repetitive learning trials, a few trials apart one from another. On these trials an identical series of nine digits was presented repeatedly. Recall rates of the series of digits on repetitive learning trials rose from the first trial to the second, and, by contraries, fell on the third. We call this fall "paradoxical fall," since it is contrary to the expectation of standard theories of learning. It was pointed out that we should investigate the detailed cognitive process, i.e., the microcognitive process, in learning.memory of simple series of digits, because it involves the important basic cognitive process of learning.memory.

Voter Mobilization and Party Competition in a Volatile Electorate

Following a model suggested by McPhee and Smith, the institutionalization of voting behavior is analyzed as a repetitive learning process. Using the Weimar Republic as an example, volatile partisan attachments among groups of supporters of the more ideologically moderate political parties are examined. A formal model comprised of a system of four interdependent differential equations is used to characterize the aggregate voter shifts over time of various groups in the population. The results show that a large wave of new voters disrupted emerging electoral patterns in 1930. In the subsequent realigning election of July 1932, electoral support for non-Catholic moderate parties was transferred to the extreme rightist parties, notably the Nazis.

Duration of exposure to a novel environment affects retention in aging mice

Aging mice were previously found not to retain information acquired in a single exposure to a novel environment as well as young mice. In an attempt to eliminate this difference, mice were given repeated daily 3-min exposures to a shuttlebox, and daily changes in exploratory activity reflected their retention of prior exposures. Activity levels progressively declined over days in all age groups (2, 12, and 24 months). However, activity in young mice fell to a lower level than that of the older mice, indicating that brief but repetitive learning sessions did not eliminate the age differences. Increasing each exposure session to 12 min eliminated the age differences in retention when the whole test session was examined. These data suggest that older mice may require more time to acquire the same amount of information as young mice. However, analysis of minute-by-minute acitivity over days indicated that older mice had forgotten their prior exposure at the first minute of each new session, but over the course of the next few minutes remembered their previous experience as well as young animals. Therefore, older mice also appear to be subject to retrieval failures even after they have received prolonged training sessions.