阅读说明：本技术 一种遮光控制系统 (Shading control system ) 是由 马从国 翁润庭 崔家兴 丁晓红 王苏琪 杨艳 柏小颖 周恒瑞 张月红 李亚洲 刘 于 2021-01-08 设计创作，主要内容包括：本发明涉及自动化生产领域,公开了一种遮光控制系统,利用NARX神经网络模型1、NARX神经网络模型2和NARX神经网络模型3分别对升降框位移的误差、控制量和实际位移值进行预测,NARX神经网络通过引入延时模块及输出反馈建立模型的动态递归网络,将输入和输出向量延时反馈引入网络训练中,形成新的输入向量,具有良好的非线性映射能力,输入不仅包括原始升降框位移的误差、控制量和实际位移值的输入数据,还包含经过训练后的对应输出数据,网络的泛化能力得到提高,使其在升降框对应参数预测中较传统的静态神经网络具有更好的预测精度和自适应能力,该单片机控制器提高了该控制系统的精确度、鲁棒性和系统的可靠性。(The invention relates to the field of automatic production, and discloses a shading control system, which respectively predicts the displacement error, the control quantity and the actual displacement value of an elevating frame by utilizing an NARX neural network model 1, an NARX neural network model 2 and an NARX neural network model 3, the NARX neural network introduces the delay feedback of input and output vectors into network training by introducing a delay module and a dynamic recursive network of an output feedback building model to form a new input vector, and has good nonlinear mapping capability Robustness and reliability of the system.)

1. A shading control system is characterized by comprising an object conveying device (4), a shading device (3) and a scanning device (1) which are arranged on a frame (10); the object conveying device is used for conveying an object to be scanned into a scanning area, the shading device is used for shading the top and the periphery of the scanning area, and the scanning device is used for scanning the object to be scanned;

the shading device comprises a single chip microcomputer controller, a lifting frame (35) fixed with a displacement sensor and lifting shading cloth (34) fixed on the lifting frame (35), wherein the single chip microcomputer controller drives the lifting shading cloth (34) to be folded and unfolded by adjusting the lifting frame (35) to lift;

the single chip microcomputer controller comprises an STM32 single chip microcomputer, an L298 motor driving circuit and an intelligent controller in an STM32 single chip microcomputer, wherein the intelligent controller comprises 3 NARX neural network models, an ESN neural network model, a PID controller, an LSTM neural network model, a dynamic recursive wavelet neural network model, a plurality of Elman neural network models and 1 beat-by-beat delay line TDL; the shading cloth receiving and releasing adjusting platform is composed of the STM32 single chip microcomputer, the L298 motor driving circuit, the lifting frame, the lifting shading cloth and the displacement sensor, and intelligent adjustment of the lifting shading cloth receiving and releasing is achieved through an intelligent controller in the STM32 single chip microcomputer.

2. The shading control system according to claim 1, wherein in the intelligent controller, the 3 NARX neural network models are NARX neural network model 1, NARX neural network model 2, and NARX neural network model 3, the output of NARX neural network model 1 is the input corresponding to PID controller and LSTM neural network model, and the input corresponding to dynamic recursive wavelet neural network model, the 3 outputs of dynamic recursive wavelet neural network model are the proportional, derivative, and integral coefficients of PID controller input, the sum of PID controller output value and LSTM neural network model output value is the input of NARX neural network model 2, the output of NARX neural network model 2 is the input corresponding to dynamic recursive wavelet neural network model, the input of L298 motor driving circuit, and the input corresponding to LSTM neural network model, the displacement sensor detects the elevator frame displacement value as the input of TDL and NARX neural network model 3 according to delay line, the output of the NARX neural network model 3 is respectively the input corresponding to the dynamic recursive wavelet neural network model and the input corresponding to the LSTM neural network model, a plurality of displacement values of the lifting frame output by a beat delay line TDL are respectively used as the input of a plurality of Elman neural network models, the output of the plurality of Elman neural network models is used as the input of an ESN neural network model, the output value of the ESN neural network model is used as a lifting frame displacement feedback value, and the error change rate of the lifting frame displacement given value and the output value of the ESN neural network model are used as the input of the NARX neural network model 1; the NARX neural network model 2 realizes the prediction of the sum of the output value of the PID controller and the output value of the LSTM neural network model and the next prediction control of the displacement of the lifting frame, and the ESN neural network model realizes the fusion of a plurality of output values of the Elman neural network model and the next accurate prediction of the displacement of the lifting frame.

3. The shading control system according to claim 2, wherein in the shading cloth retraction and extension adjustment, the output of the NARX neural network model 2 of the intelligent controller in the STM32 single chip microcomputer is used as the input of the L298 motor driving circuit, the output of the L298 motor driving circuit is used as the input of the driving motor in the lifting frame, the lifting frame drives the lifting shading cloth to move, the displacement sensor measures the displacement of the lifting frame, and the outputs of the displacement sensor are respectively used as the input of the beat delay line TDL of the intelligent controller in the STM32 single chip microcomputer and the input of the NARX neural network model 3.

4. A shading control system according to claim 1, wherein in the shading device, a bracket (311) is fixed on the frame (10) and located around the scanning area, and the top shading cloth (31) and the lifting shading cloth (34) are respectively installed on the top and around the bracket (311) through a retracting mechanism; when the transparent placing plate (42) in the object conveying device (4) is positioned in the scanning area, the top shading cloth (31) covers the top of the scanning area, and the lifting shading cloth (34) covers the periphery of the scanning area.

5. The shading control system according to claim 4, wherein the retracting mechanism comprises a lifting frame (35) and a first rotating shaft (32) and a second rotating shaft (33) which are installed at both ends of the top surface of the bracket (311) in parallel, one end of the top surface shading cloth (31) is fixed on the first rotating shaft (32), the other end is connected with one end of a top surface traction rope (38), and the other end of the top surface traction rope (38) is fixed on the second rotating shaft (33); the lifting shading cloth (34) is fixed on the lifting frame (35), the lifting frame (35) is connected to the outer wall of the support (311) in a sliding mode, lifting pulling ropes (39) are connected to the four sides of the lifting frame (35) respectively, and the top ends of the four lifting pulling ropes (39) are fixed on the first rotating shaft (32) and the second rotating shaft (33) respectively.

6. The shading control system according to claim 5, further comprising four spacers (310), wherein the four spacers (310) are fixed to two ends of the first shaft (32) and the second shaft (33), respectively, and the lifting/lowering rope (39) and the top shading cloth (31) are located on two sides of each spacer (310).

7. The shading control system according to any one of claims 1 to 6, wherein in the object conveying device (4), a conveying guide rail (41) is horizontally fixed on the frame (10), a transparent placing plate (42) is movably connected with the conveying guide rail (41), and a placing plate driving mechanism provided on the frame (10) is used for driving the transparent placing plate (42) to horizontally move into or out of a scanning area along the conveying guide rail (41).

8. The shading control system according to claim 7, wherein in the placing plate driving mechanism, a first motor (45) is fixed to the frame (10), a rotating shaft of a gear (44) is fixed to an output shaft of the first motor (45), one side of a rack (43) is fixed to the transparent placing plate (42), and the other side thereof is engaged with the gear (44).

9. The shading control system according to claim 7, further comprising a scanning rendering lamp (13) mounted on the gantry (10), the scanning rendering lamp (13) being located between the upper scanning probe (11) and the transparent placing plate (42).

10. The shading control system according to claim 7, wherein in the scanning device (1), an upper scanning probe (11) and a lower scanning probe (12) are respectively mounted on the frame (10) above and below the scanning area and respectively above and below the transparent placement plate (42).

Technical Field

The invention relates to the technical field of automatic production, in particular to a shading control system.

Background

At present, most of scanners used for recording important materials of existing files are in contact type scanning, paper is conveyed by using a combined roller and a conveyor belt, because the physical characteristics of the material paper are very sensitive, one part of the paper is likely to be adsorbed on a pressing plate above the paper during scanning, and the other part of the paper is flatly laid on a transparent placing plate below the paper, so that the flatness of the paper during scanning is not uniform, and the problems of clear partial scanning and unclear partial scanning occur; in addition, incomplete scanning may be caused by paper scrap blocking the paper, or uneven rolling of the paper.

In addition, in the existing scanning equipment, the controller has poor control precision, stability, timeliness and adaptability to all parts during scanning, and the scanning effect is influenced.

Disclosure of Invention

The purpose of the invention is as follows: aiming at the problems in the prior art, the invention provides a shading control system which can improve the control precision, stability, timeliness and adaptability of a controller to a shading device during scanning and improve the scanning effect.

The technical scheme is as follows: the invention provides a shading control system, which comprises an object conveying device, a shading device and a scanning device, wherein the object conveying device, the shading device and the scanning device are arranged on a rack; the object conveying device is used for conveying an object to be scanned into a scanning area, the shading device is used for shading the top and the periphery of the scanning area, and the scanning device is used for scanning the object to be scanned; the shading device comprises a single chip microcomputer controller, a lifting frame fixed with a displacement sensor and lifting shading cloth fixed on the lifting frame, wherein the single chip microcomputer controller drives the lifting shading cloth to be folded and unfolded by adjusting the lifting frame to lift; the single chip microcomputer controller comprises an STM32 single chip microcomputer, an L298 motor driving circuit and an intelligent controller in an STM32 single chip microcomputer, wherein the intelligent controller comprises 3 NARX neural network models, an ESN neural network model, a PID controller, an LSTM neural network model, a dynamic recursive wavelet neural network model, a plurality of Elman neural network models and 1 beat-by-beat delay line TDL; the shading cloth receiving and releasing adjusting platform is composed of the STM32 single chip microcomputer, the L298 motor driving circuit, the lifting frame, the lifting shading cloth and the displacement sensor, and intelligent adjustment of the lifting shading cloth receiving and releasing is achieved through an intelligent controller in the STM32 single chip microcomputer.

Further, in the intelligent controller, 3 NARX neural network models are NARX neural network model 1, NARX neural network model 2 and NARX neural network model 3, respectively, the output of the NARX neural network model 1 is the input corresponding to the PID controller and LSTM neural network model and the input corresponding to the dynamic recursive wavelet neural network model, respectively, the 3 outputs of the dynamic recursive wavelet neural network model are the proportional, derivative and integral coefficients of the input of the PID controller, respectively, the sum of the output value of the PID controller and the output value of the LSTM neural network model is the input of the NARX neural network model 2, the outputs of the NARX neural network model 2 are the input corresponding to the dynamic recursive wavelet neural network model, the input of the L motor driving circuit and the input corresponding to the LSTM neural network model, respectively, the displacement sensor detects 298 the up-down frame displacement value as the input of the slapping delay line TDL and the NARX neural network model 3, respectively, the output of the NARX neural network model 3 is respectively the input corresponding to the dynamic recursive wavelet neural network model and the input corresponding to the LSTM neural network model, a plurality of displacement values of the lifting frame output by a beat delay line TDL are respectively used as the input of a plurality of Elman neural network models, the output of the plurality of Elman neural network models is used as the input of an ESN neural network model, the output value of the ESN neural network model is used as a lifting frame displacement feedback value, and the error change rate of the lifting frame displacement given value and the output value of the ESN neural network model are used as the input of the NARX neural network model 1; the NARX neural network model 2 realizes the prediction of the sum of the output value of the PID controller and the output value of the LSTM neural network model and the next prediction control of the displacement of the lifting frame, and the ESN neural network model realizes the fusion of a plurality of output values of the Elman neural network model and the next accurate prediction of the displacement of the lifting frame.

Further, in the shading cloth folding and unfolding adjustment, the output of an NARX neural network model 2 of an intelligent controller in an STM32 single chip microcomputer is used as the input of an L298 motor driving circuit, the output of the L298 motor driving circuit is used as the input of a driving motor in a lifting frame, the lifting frame drives the lifting shading cloth to move, a displacement sensor measures the displacement of the lifting frame, and the output of the displacement sensor is respectively used as the input of a beat-pressing delay line TDL of the intelligent controller in an STM32 single chip microcomputer and the input of an NARX neural network model 3.

Furthermore, in the shading device, a bracket is fixed on the rack and positioned around the scanning area, and top shading cloth and lifting shading cloth are respectively arranged at the top and the periphery of the bracket through a retracting mechanism; when the transparent placing plate in the object conveying device is positioned in the scanning area, the top surface shading cloth covers the top of the scanning area, and the lifting shading cloth covers the periphery of the scanning area.

Furthermore, the retraction mechanism comprises a lifting frame, a first rotating shaft and a second rotating shaft which are arranged at two ends of the top surface of the support and are parallel to each other, one end of the top surface shading cloth is fixed on the first rotating shaft, the other end of the top surface shading cloth is connected with one end of a top surface traction rope, and the other end of the top surface traction rope is fixed on the second rotating shaft; the lifting shading cloth is fixed on the lifting frame, the lifting frame is connected to the outer wall of the support in a sliding mode, lifting traction ropes are connected to four sides of the lifting frame respectively, and the top ends of the four lifting traction ropes are fixed on the first rotating shaft and the second rotating shaft respectively. The first rotating shaft and the second rotating shaft synchronously rotate, on one hand, the pulling force of the top surface shading cloth is matched, the phenomenon of deviation of the top surface shading cloth caused by friction and inconsistent pulling force is avoided, and on the other hand, the lifting speed or the descending speed of each lifting frame is matched.

Furthermore, the retraction mechanism further comprises four isolation blocks, the four isolation blocks are respectively fixed at two ends of the first rotating shaft and two ends of the second rotating shaft, and the lifting traction rope and the top surface shading cloth are located on two sides of each isolation block. The isolation block prevents the top surface shading cloth and the lifting frame traction rope from being mutually wound together to influence the use of the device.

Furthermore, in the object conveying device, a conveying guide rail is horizontally fixed on the rack, a transparent placing plate is movably connected with the conveying guide rail, and a placing plate driving mechanism arranged on the rack is used for driving the transparent placing plate to horizontally move in or out of a scanning area along the conveying guide rail.

Preferably, in the placing plate driving mechanism, a first motor is fixed on the frame, a rotating shaft of a gear is fixed on an output shaft of the first motor, one side of a rack is fixed with the transparent placing plate, and the other side of the rack is meshed with the gear. The first motor drives the gear to rotate, the transparent placing plate is driven by the gear and the rack to enter or exit from a scanning area along the conveying guide rail, and the defects that short-distance conveying is not stable and reliable, conveying precision is not well controlled and the like caused by the transmission of a conveying belt on the traditional market are overcome.

Furthermore, in the scanning device, an upper scanning probe and a lower scanning probe are respectively installed on the rack and above and below the scanning area, and are respectively located above and below the transparent placing plate. The cooperation of the upper scanning probe and the lower scanning probe realizes the simultaneous scanning of the front and back surfaces of the object to be scanned.

Further, the shading control system further comprises a scanning and rendering lamp installed on the rack, and the scanning and rendering lamp is located between the upper scanning probe and the transparent placing plate. Aiming at the old and the wrinkles of the paper, the rendering lamp can adjust different brightness and light positions, so that the problems of the old and the wrinkles are solved, and the scanning effect is better.

Has the advantages that: compared with the traditional controller, the intelligent controller in the single chip microcomputer has the advantages that:

1. the method utilizes the NARX neural network model 1, the NARX neural network model 2 and the NARX neural network model 3 to respectively predict the displacement error, the control quantity and the actual displacement value of the lifting frame, because the NARX neural network establishes the dynamic recursive network of the model by introducing the delay module and the output feedback, the NARX neural network introduces the delay feedback of the input vector and the output vector into the network training to form a new input vector, and has good nonlinear mapping capability.

2. The invention adopts a plurality of Elman neural network models to simultaneously realize the prediction of the displacement of the lifting frame, and improves the accuracy and robustness of the prediction of the displacement value of the lifting frame. The transfer function of the hidden layer unit can adopt a linear or non-linear function, and the accepting layer is also called a context layer or a state layer, and is used for memorizing the output value of the hidden layer unit at the previous moment, which can be regarded as a time delay operator. The Elman neural network model is characterized in that the output of the hidden layer is self-connected to the input of the hidden layer through the delay and storage of the carrying layer, the self-connection mode enables the output to have sensitivity to the data of the historical state of the displacement of the lifting frame, and the addition of the internal feedback network increases the capability of the network for processing dynamic information, so that the purpose of dynamic modeling is achieved. The Elman neural network model regression neural network is characterized in that the output of the hidden layer is self-connected to the input of the hidden layer through the delay and storage of the structural unit, the self-connection mode enables the hidden layer to have sensitivity to the data of the historical state of the displacement of the lifting frame, and the addition of the internal feedback network increases the capability of the network for processing dynamic information, thereby being beneficial to the modeling of the dynamic process of the displacement data of the lifting frame; the Elman neural network model utilizes feedback connection of dynamic neurons of the association layer to strengthen the memory of the network on time sequence characteristic information, thereby improving the accuracy and robustness of the prediction of the lifting frame displacement data.

3. The LSTM neural network model is similar to a standard network containing a recursion hidden layer, the only change is to use a memory module to replace an original hidden layer unit, the problems of gradient disappearance and sharp increase are solved by self-feedback of the internal state of a memory cell and truncation of errors of input and output, compared with a BP neural network and a common RNN, the LSTM is added with 1 state unit c and 3 control gates, the feature inclusion capacity and the memory capacity of the model are greatly increased, and under-fitting and gradient disappearance are avoided. The function of the LSTM aims at the correlation existing in the output data of a plurality of NARX neural network models 1,2 and 3, and remembers the relationship and the change of the relationship in time, so that a more accurate result of the LSTM neural network model for outputting the control quantity of the control lifting frame is obtained, and the accuracy of the control quantity of the control lifting frame is improved.

4. The LSTM neural network model has a chain-like repeating network structure similar to a standard RNN, the repeating network in the standard RNN is very simple, and the repeating network in the LSTM neural network model has 4 interaction layers including 3 gate layers and 1 tanh layer. Processor state is a key variable in the LSTM neural network model that carries information of the previous NARX neural network model 1, NARX neural network model 2, and NARX neural network model 3 output steps and steps through the entire LSTM neural network model. The gates in the interaction layer may partially delete the processor state of the previous step and add new output information of the NARX neural network model 1, the NARX neural network model 2, and the NARX neural network model 3 to the processor state of the current step based on the hidden state of the previous step and the input of the current step. The inputs to each repeating network include the hidden state and processor state of the previous step and the input of the current step. The processor state is updated according to the calculation results of the 4 interaction layers. The updated processor state and hidden state constitute the output and are passed on to the next step.

5. The LSTM neural network model is a recurrent neural network with 4 interaction layers in a repeating network. It can not only extract information from the NARX neural network model 1, NARX neural network model 2 and NARX neural network model 3 output sequence data like a standard recurrent neural network, but also retain information with long-term correlation from previous distant steps. Furthermore, because the sampling intervals of the output quantities of the NARX neural network model 1, the NARX neural network model 2 and the NARX neural network model 3 are relatively small, there is a long-term spatial correlation of the output quantities of the NARX neural network model 1, the NARX neural network model 2 and the NARX neural network model 3, while the LSTM neural network model has sufficient long-term memory to deal with this problem.

6. The ESN neural network model designs a network hidden layer into a sparse network consisting of a plurality of neurons, the function of memorizing the output data of a plurality of Elman neural network models is achieved by adjusting the characteristic of internal weights of the network, an internal dynamic reserve pool contains a large number of sparsely connected neurons and contains the running state of the system, the ESN neural network model has the function of memorizing the output values of a plurality of Elman neural network models in a short term, the stability of a recursion network in the reserve pool is ensured by presetting the spectrum radius of an internal connection weight matrix of the ESN neural network model, and the stability and the accuracy of the feedback value of an output lifting frame of the ESN neural network model are improved.

7. The ESN neural network model inherits the current time of the state of the reserve pool to the previous time of the state of the reserve pool, and has a transient memory characteristic for historical data output by the plurality of Elman neural network models, and research results show that the ESN neural network model with the historical memory has the characteristics of high precision, high accuracy, high timeliness and stability of the plurality of Elman neural network models; as a novel dynamic recurrent neural network, the ESN neural network model is established by adopting a linear regression method, the problems that the traditional neural network is low in convergence speed and easy to fall into local minimum are solved, the complexity of the training process is simplified, and the purpose of efficiently outputting fusion values of a plurality of Elman neural network models is realized.

8. The difference between the dynamic recursive wavelet neural network model and the common static wavelet neural network lies in that the dynamic recursive wavelet neural network model has two function associated layer nodes which play the role of storing the internal state of the network, a self-feedback loop with fixed gain is added on the two associated layer nodes, and the memory performance of time series characteristic information is enhanced, so that the precision of proportion, integral and differential coefficients for adjusting PID is enhanced to ensure the accuracy and stability of better shading cloth folding and unfolding control.

Drawings

Fig. 1 is a schematic view of the entire structure of a shading control system;

FIG. 2 is a schematic view of the structure of the object transfer device;

FIG. 3 is a schematic view of the light shielding device;

FIG. 4 is a schematic view of a partial structure of the light shielding device;

fig. 5 is a working flow chart of the shading cloth folding and unfolding adjusting platform and the intelligent controller.

Detailed Description

The present invention will be described in detail with reference to the accompanying drawings.

The present embodiment provides a shading control system, as shown in fig. 1, which mainly comprises an object conveying device 4, a shading device 3 and a scanning device 1, which are sequentially arranged on a rack 10; the object conveying device 4 is used for conveying an object to be scanned into a scanning area, the shading device 3 is used for shading the top and the periphery of the scanning area, and the scanning device 1 is used for scanning the object to be scanned.

As shown in fig. 1 and 2, the object conveying device 4 includes a transparent placing plate 42 and a conveying rail 41, the conveying rail 41 is horizontally fixed on the frame 10, the transparent placing plate 42 is movably connected with the conveying rail 41, and a placing plate driving mechanism provided on the frame 10 is used for driving the transparent placing plate 42 to horizontally move into or out of the scanning area along the conveying rail 41; in the placing plate driving mechanism, a first motor 45 is fixed on the frame 10, a rotating shaft of a gear 44 is fixed on an output shaft of the first motor 45, one side of a rack 43 is fixed with the transparent placing plate 42, and the other side is meshed with the gear 44.

As shown in fig. 3 to 5, the light shielding device 3 includes a single chip microcomputer controller, a first rotating shaft motor 36, a second rotating shaft motor 37, a top surface light shielding cloth 31, a lifting frame 35 fixed with a displacement sensor (preferably, AZ-7120), a lifting light shielding cloth 34 fixed on the lifting frame 35, a bracket 311, and a first rotating shaft 32 and a second rotating shaft 33 installed at two ends of the top surface of the bracket 311 in parallel, wherein the single chip microcomputer controller drives the lifting light shielding cloth 34 to retract and retract by adjusting the lifting frame 35 to lift. The first rotary shaft 32 is fixed to an output end of a first rotary shaft motor 36, and the second rotary shaft 33 is fixed to an output end of a second rotary shaft motor 37. One end of the top shading cloth 31 is fixed on the first rotating shaft 32, the other end is connected with one end of a top surface hauling rope 38, and the other end of the top surface hauling rope 38 is fixed on the second rotating shaft 33; the lifting shading cloth 34 is fixed on the lifting frame 35, the lifting frame 35 is slidably connected to the outer wall of the bracket 311, the four sides of the lifting frame 35 are respectively connected with lifting pulling ropes 39, and the top ends of the four lifting pulling ropes 39 are respectively fixed on the first rotating shaft 32 and the second rotating shaft 33. The four isolation blocks 310 are respectively fixed at two ends of the first rotating shaft 32 and the second rotating shaft 33, and the lifting traction rope 39 and the top shading cloth 31 are positioned at two sides of each isolation block 310.

In the scanning device 1, the upper scanning probe 11 and the lower scanning probe 12 are respectively installed on the rack 10 above and below the scanning area, and the upper scanning probe 11 is located above the transparent placing plate 42, the lower scanning probe 12 is located below the transparent placing plate 42, and after the transparent placing plate 42 enters the scanning area, the upper scanning probe 11 and the lower scanning probe 12 are respectively located above and below the transparent placing plate 42.

On the gantry 10, scanning and rendering lamps 13 are further mounted between the upper scanning probe 11 and the transparent placing plate 42 and between the lower scanning probe 12 and the transparent placing plate 42.

The operating principle of the light shielding control system in the present embodiment is as follows:

after the paper to be scanned is placed on the transparent placing plate 42, the first motor 45 is started to rotate forward to drive the gear 44 to rotate forward, the transparent placing plate 42 enters the scanning area along the conveying guide rail 41 through meshing transmission of the gear 44 and the rack 43, and after the paper to be scanned on the transparent placing plate 42 enters the scanning area, the first motor 45 stops rotating forward just between the upper scanning probe 11 and the lower scanning probe 12.

Then the single chip microcomputer controller starts to control the second rotating shaft motor 37 to drive the second rotating shaft 33 to rotate, the top surface traction rope 38 wound on the second rotating shaft 33 is wound, so that the top surface traction rope 38 pulls the top surface shading cloth 31 until the scanning area is completely covered, meanwhile, the single chip microcomputer controller controls the first rotating shaft motor 36 to synchronously rotate, on one hand, the pulling force of the top surface shading cloth 31 is matched to avoid the deviation phenomenon of the top surface shading cloth 31 caused by the inconsistency of the friction force and the pulling force, on the other hand, the lifting speed or the descending speed of all the lifting traction ropes 39 is matched, one end of each lifting traction rope 39 is respectively wound on the first rotating shaft 32, the second rotating shaft 33, the other end of each lifting traction rope 39 is fixedly connected on the lifting frame 35, and the lifting traction ropes 39 and the top surface shading cloth 31 are separated by the isolating blocks 310 arranged on the first rotating shaft 32 and the second rotating shaft 33, affecting the operation of the device. When the top surface shading cloth 31 shades the top surface of the scanning area, the first rotating shaft 32 and the second rotating shaft 33 also drive the lifting traction rope 39 to enable the lifting frame 35 to ascend along the outer wall of the support 311, namely the lifting shading cloth 34 fixedly connected with the bottom of the lifting frame 35 is gradually lifted up to shade the front side, the rear side, the left side and the right side of the scanning area, when the displacement sensor fixed on the lifting frame 35 detects that the displacement of the lifting frame 35 reaches a preset value, the lifting frame 35 rises to the highest point, meanwhile, the top surface shading cloth 31 also completely shades the scanning area, the displacement sensor transmits a signal to the single chip microcomputer controller, and the single chip microcomputer controller controls the first rotating shaft motor 36 and the second rotating shaft motor 37 to stop rotating; the lifting frame 35 is connected to the outer wall of the bracket 311 in a sliding manner; the transportation rail 41 is fixedly connected to the inner wall of the frame 311, and when the transparent placing plate 42 moves into the scanning area, the transparent placing plate 42 is completely located inside the frame 311, and the lifting frame 35 is located below the transportation rail 41 at the bottommost portion of the frame 311, so that the lifting frame 35 does not interfere with the transportation rail 41.

After the top surface shading cloth 31 shades the top surface of the scanning area, the lifting shading cloth 34 shades the front, the back, the left and the right sides of the scanning area, the upper placed scanning probe 11 and the lower placed scanning probe 12 start to carry out non-contact scanning on the front and the back surfaces of the paper to be scanned; after the scanning is finished, the upper scanning probe 11 and the lower scanning probe 12 stop scanning. Then the single chip microcomputer controller controls the first rotating shaft motor 36 to rotate reversely to drive the first rotating shaft 32 to retract the top shading cloth 31, meanwhile, the single chip microcomputer controller controls the second rotating shaft motor 37 to rotate reversely to drive the second rotating shaft 33 to drive the lifting frame 35 with the lifting shading cloth 34 to descend to the bottommost part of the support 311, at the moment, the displacement sensor detects that the displacement of the lifting frame 35 also reaches a preset value, the displacement sensor transmits a signal to the single chip microcomputer controller, the single chip microcomputer controller controls the first rotating shaft motor 36 and the second rotating shaft motor 37 to stop rotating, and at the moment, the lifting frame 35 is located below the conveying guide rail 41.

Then the first motor 45 drives the gear 44 to rotate reversely, the transparent placing plate 42 moves out of the scanning area along the conveying guide rail 41 through the engagement of the gear 44 and the rack 43, the first motor 45 stops rotating reversely, the scanning is finished, and the paper to be scanned on the transparent placing plate 42 is taken away.

Aiming at old or wrinkled paper, the scanning rendering lamp 13 can be turned on in the scanning process to adjust different brightness and light positions, so that the problems of old, wrinkling and the like are solved, and the scanning effect is better.

This object conveyer among scanning equipment can realize non-contact scanning, can realize carrying out automatic scanning to the object of different specifications and irregular shape, and the user only need will wait to scan the object put the transparent board of placing on can. On one hand, the top surface shading cloth and the lifting shading cloth in the shading device can prevent the influence of external light on a scanning area, so that the scanning effect is better, and the shadow caused by uneven light is avoided; on the other hand, dust can be prevented from falling onto the scanning lens, because the dust on the lens is difficult to clean.

The specific design idea of the single chip microcomputer controller is as follows:

design of overall system function

The single chip microcomputer controller comprises an STM32 single chip microcomputer, an L298 motor driving circuit and an intelligent controller in an STM32 single chip microcomputer, wherein the intelligent controller comprises 3 NARX neural network models, an ESN neural network model, a PID controller, an LSTM neural network model, a dynamic recursive wavelet neural network model, a plurality of Elman neural network models and 1 beat-by-beat delay line TDL; an STM32 single chip microcomputer, an L298 motor driving circuit, a lifting frame, a piece of shading cloth and a displacement sensor form a shading cloth folding and unfolding adjusting platform, and an intelligent controller in the STM32 single chip microcomputer realizes intelligent adjustment of folding and unfolding of the shading cloth; the shading cloth folding and unfolding adjusting platform and the intelligent controller are shown in figure 5.

Design process of intelligent controller

1. NARX neural network model design

The 3 NARX neural network models are respectively an NARX neural network model 1, an NARX neural network model 2 and an NARX neural network model 3, and the output of the NARX neural network model 1 is respectively the input corresponding to the PID controller and the LSTM neural network model and the input corresponding to the dynamic recursive wavelet neural network model; the sum of the output value of the PID controller and the output value of the LSTM neural network model is used as the input of the NARX neural network model 2, and the output of the NARX neural network model 2 is the input corresponding to the dynamic recursive wavelet neural network model, the input of the L298 motor driving circuit and the input corresponding to the LSTM neural network model respectively; the displacement sensor detects the displacement values of the lifting frame as the input of the beat delay line TDL and the NARX neural network model 3 respectively, the output of the NARX neural network model 3 is the input corresponding to the dynamic recursive wavelet neural network model and the input corresponding to the LSTM neural network model respectively, and the error change rate of the displacement given value of the lifting frame and the output value of the ESN neural network model are used as the input of the NARX neural network model 1. The NARX neural network model (Nonlinear Auto-Regression with External input neural network) is a dynamic feedforward neural network, the NARX neural network model is a Nonlinear autoregressive network with External input, the NARX neural network model has a dynamic characteristic of multistep time delay and is connected with a plurality of layers of closed networks through feedback, and the recurrent neural network of the NARX neural network model is a dynamic neural network which is widely applied in a Nonlinear dynamic system and has the performance generally superior to that of a full recurrent neural network. Before application, the delay order and the number of hidden layer neurons of the input and output are generally determined in advance, the current output of the NARX neural network model not only depends on the past output y (t-n), but also depends on the current input prediction vector X (t), the delay order of the input prediction vector and the like, wherein an input signal is transmitted to the hidden layer through an epitaxial layer, the hidden layer processes the input signal and then transmits the processed signal to the output layer, the output layer linearly weights the output signal of the hidden layer to obtain a final neural network output signal, and the epitaxial layer delays a signal fed back by the network and a signal output by the input layer and then transmits the final neural network output signal to the hidden layer. The NARX neural network model has the characteristics of nonlinear mapping capability, good robustness, adaptability and the like, and is suitable for predicting flight linear change parameters. x (t) represents the external input of the NARX neural network model, and m represents the delay order of the external input; y (t) is the output of the NARX neural network model, n is the output delay order; s is the number of hidden layer neurons; the output of the jth implicit element can be found as:

in the above formula, w_jiAs a connection weight between the ith input and the jth implicit neuron, b_jIs the bias value of the jth implicit neuron, the value of the output y (t +1) of the networkComprises the following steps:

y(t+1)＝f[y(t),y(t-1),…,y(t-n),x(t),x(t-1),…,x(t-m+1)；W] (2)

2. PID controller design

The 3 outputs of the dynamic recursive wavelet neural network model are respectively used as the proportional, differential and integral coefficients input by the PID controller, the output of the NARX neural network model 1 is used as the error of the PID controller, the PID controller calculates the three inputs of the PID controller according to the output of the NARX neural network model 1 at the previous moment and the output of the NARX neural network model 1 at the previous two moments, and the calculation formula is shown as (3):

the algorithm of the controller of the PID is:

the 3 outputs of the dynamic recursive wavelet neural network model are respectively used as the proportional, differential and integral coefficients of the PID controller input, and the inputs of the dynamic recursive wavelet neural network model are the outputs of the NARX neural network model 1, the NARX neural network model 2 and the NARX neural network model 3. Wavelet Neural network WNN (wavelet Neural networks) theoretical basis is a feedforward network provided by taking a wavelet function as an excitation function of a neuron and combining an artificial Neural network, wherein the expansion and contraction, the translation factor and the connection weight of wavelets in the wavelet Neural network are adaptively adjusted in the optimization process of an error energy function. An input signal with a dynamic recursive wavelet neural network model can be represented as an input one-dimensional vector x_i(i ═ 1,2, …, n), the output signal is denoted y_k(k ═ 1,2, …, m), the calculation formula of the dynamic recursive wavelet neural network model output value is:

in the formula omega_ijInputting the connection weight between the i node of the layer and the j node of the hidden layer,as wavelet basis functions, b_jIs a shift factor of the wavelet basis function, a_jScale factor, omega, of wavelet basis functions_jkThe connection weight between the node of the hidden layer j and the node of the output layer k. The difference between the dynamic recursive wavelet neural network model and the common static wavelet neural network is that the dynamic recursive wavelet neural network model has two associated layer nodes which play a role in storing the internal state of the network, and a self-feedback loop with fixed gain is added on the two associated layer nodes to enhance the memory performance of time sequence characteristic information, so that the tracking precision of the control of the displacement evolution track of the lifting frame is enhanced to ensure better control precision; the first associated layer node is used for storing the state of the phase point of the hidden layer node at the previous moment and transmitting the state to the hidden layer node at the next moment; the second correlation layer node is used for storing the state of the phase point of the output layer node at the previous moment and transmitting the state to the hidden layer node at the next moment; feedback information of neurons of the hidden layer and the output layer can affect the dynamic processing capacity of proportional, integral and differential coefficients of a PID controller output by the dynamic recursive wavelet neural network model, two associated layers belong to state feedback inside the dynamic recursive wavelet neural network model, the dynamic memory performance specific to the recursion of the dynamic recursive wavelet neural network model is formed, and the accuracy and the dynamic performance of the dynamic recursive wavelet neural network model and the PID controller participating in controlling the displacement of the lifting frame are improved; a group of connection weights are added between the first association layer node and the output layer node of the dynamic recursive wavelet neural network model to enhance the dynamic approximation capability of the dynamic recursive wavelet neural network model for controlling the displacement of the lifting frame and improve the control precision of the displacement of the lifting frame. The weight and threshold correction algorithm of the dynamic recursive wavelet neural network model in the patent adopts a gradient correction method to update the network weight and the wavelet basis function parameters, so that the output of the dynamic recursive wavelet neural network is continuously close to the expected output.

3. LSTM neural network model design

The NARX neural network model 1, the NARX neural network model 2 and the NARX neural network model 3 are respectively used as corresponding inputs of the LSTM neural network model, and the sum of the output value of the LSTM neural network model and the output value of the PID controller is used as the input of the NARX neural network model 2;

the temporal Recurrent Neural Network (RNN) model, which consists of Long Short Term Memory (LSTM) elements, is called the LSTM temporal recurrent neural network, also commonly referred to as the LSTM network. The LSTM neural network model introduces mechanisms of Memory cells (Memory cells) and hidden layer states (Cell states) to control the transfer of information between hidden layers. The memory unit of an LSTM neural network has 3 Gates (Gates) as Input Gate, forgetting Gate and Output Gate. Wherein, the input gate can control the adding or filtering of new information; the forgetting door can forget the information to be lost and keep the useful information in the past; the output gate enables the memory unit to output only information related to the current time step. The 3 gate structures carry out operations such as matrix multiplication, nonlinear summation and the like in the memory unit, so that the memory still cannot be attenuated in continuous iteration. The long-short term memory unit (LSTM) structure unit is composed of a unit (Cell), an Input Gate (Input Gate), an Output Gate (Output Gate) and a forgetting Gate (Forget Gate). The unit is responsible for remembering values at arbitrary time intervals, and all three gates can be considered as conventional artificial neurons for computing a weighted sum of activation functions. The LSTM neural network model is a model which can last for a long time and has short-term memory, is suitable for work such as prediction of time sequences and the like, effectively prevents gradient disappearance during RNN training by the LSTM, and is a special RNN by a long-short-term memory (LSTM) network. The model can learn long-term dependency information while avoiding the gradient vanishing problem. LSTM adds a structure called a Memory Cell (Memory Cell) to a neural node of a hidden layer of an internal structure RNN of a neuron to memorize past information, and adds three kinds of gate structures (Input, form, Output) to control use of history information. Let the number sequence of the input LSTM neural network model be (x)₁,x₂,…x_T) The hidden layer state is (h)₁,h₂,…h_T) Then, time t has:

i_t＝sigmoid(W_hih_t-1+W_xiX_t) (6)

f_t＝sigmoid(W_hfh_t-1+W_hfX_t) (7)

c_t＝f_t⊙c_t-1+i_t⊙tanh(W_hch_t-1+W_xcX_t) (8)

o_t＝sigmoid(W_hoh_t-1+W_hcX_t+W_coc_t) (9)

h_t＝o_t⊙tanh(c_t) (10)

wherein i_t、ft、o_tRepresenting input, forget and output doors, C_tRepresenting a cell, W_hRepresents the weight of the recursive connection, Wx represents the weight of the input layer to the hidden layer, and sigmoid and tanh are two activation functions. The method comprises the steps of firstly establishing an LSTM time recurrent neural network model, establishing a training set by utilizing preprocessed data output by an NARX neural network model 1, an NARX neural network model 2 and an NARX neural network model 3, and training the model, wherein the LSTM neural network model takes the time sequence and nonlinearity of the data output by the NARX neural network model 1, the NARX neural network model 2 and the NARX neural network model 3 into consideration, so that the control accuracy of the displacement of the lifting frame is improved.

4. Design of multiple Elman neural network models

Historical data of the displacement of the lifting frame is used as the input of a corresponding beat delay line TDL, the historical data of the displacement of the lifting frame output by the beat delay line TDL for a period of time is used as the input of a plurality of Elman neural network models, the output of the Elman neural network models is used as the corresponding input of an ESN neural network model, the output of the ESN neural network model is used as a feedback value of the displacement of the lifting frame, and the error change rate of a set value of the displacement of the lifting frame and the output value of the ESN neural network model are used as the output of a NARX neural network model 1Entering; the ESN neural network model realizes the fusion of a plurality of output values of the Elman neural network model and the accurate prediction of the displacement of the lifting frame; the Elman neural network model can be regarded as a forward neural network with a local memory unit and a local feedback connection, and a special association layer is arranged besides a hidden layer; the correlation layer receives the feedback signal from the hidden layer, and each hidden layer node is connected with the corresponding correlation layer node. The association layer takes the hidden layer state at the previous moment and the network input at the current moment as the input of the hidden layer, which is equivalent to state feedback. The transfer function of the hidden layer is generally a Sigmoid function, the output layer is a linear function, and the associated layer is also a linear function. In order to effectively solve the problem of approximation precision in the displacement prediction of the lifting frame, the function of the associated layer is enhanced. Setting the number of an input layer, an output layer and a hidden layer of the Elman neural network model as m, n and r respectively; w is a₁，w₂，w₃And w₄Respectively representing the connection weight matrixes from the structural layer unit to the hidden layer, from the input layer to the hidden layer, from the hidden layer to the output layer and from the structural layer to the output layer, wherein the expressions of the hidden layer, the associated layer and the output layer of the lifting frame displacement predictor of the Elman neural network model are respectively as follows:

c_p(k)＝x_p(k-1) (12)

5. ESN neural network model design

The output of the Elman neural network models is used as the corresponding input of the ESN neural network model, the output of the ESN neural network model is used as the displacement feedback value of the lifting frame, and the error change rate of the displacement given value of the lifting frame and the output value of the ESN neural network model are used as the input of the NARX neural network model 1; the ESN neural network model realizes the fusion of a plurality of output values of the Elman neural network model and the accurate prediction of the displacement of the lifting frame; an ESN (Echo state network, ESN) is a novel dynamic neural network, has all the advantages of the dynamic neural network, and can better adapt to nonlinear system identification compared with a common dynamic neural network because the Echo state network introduces a reserve pool concept. The reserve pool is a randomly connected reserve pool which is formed by converting a part connected among traditional dynamic neural networks, and the whole learning process is a process of learning how to connect the reserve pool. The "pool" is actually a randomly generated large-scale recursive structure in which the interconnection of neurons is sparse, usually denoted SD as the percentage of interconnected neurons in the total number of neurons N. The state equation of the ESN neural network model is as follows:

wherein W is the state variable of the neural network, W_inInput variables of the ESN neural network model; w_backConnecting a weight matrix for an output state variable of the ESN neural network model; x (n) represents the internal state of the ESN neural network model; w_outA connection weight matrix among a nuclear reserve pool of the ESN neural network model, the input of the neural network and the output of the neural network;is the output deviation of the ESN neural network model or may represent noise; f ═ f [ f₁,f₂,…,f_n]N activation functions for neurons within the "pool of stores"; f. of_iIs a hyperbolic tangent function; f. of_outIs the epsilon output functions of the ESN neural network model. The above embodiments are merely illustrative of the technical concepts and features of the present invention, and the purpose of the embodiments is to enable those skilled in the art to understand the contents of the present invention and implement the present invention, and not to limit the protection scope of the present invention. All equivalent changes and modifications made according to the spirit of the present invention should be covered by the protection of the present inventionWithin the range.