CN110580943B - An intelligent infusion system based on LoRa communication - Google Patents

Abstract

The invention discloses an intelligent infusion system based on LoRa communication, comprising an infusion monitoring unit, a pager, a repeater, a Lora communication module and a medical care station host computer. The upper computer of the medical care station is connected with the repeater through a network cable. The invention improves the robustness and security of information transmission through LoRa wireless communication technology, and improves the accuracy of data measurement by setting a weighing calibration module.

Description

An intelligent infusion system based on LoRa communication

technical field

The invention relates to the field of medical information, in particular to an intelligent infusion system based on Lora communication.

Background technique

The intelligent infusion system can automatically determine the remaining dose of the infusion, estimate the remaining time, and determine the situation of blocked needles, leaking needles, empty bottles, and dripping stops. It is an important research direction of smart medical care that nurses can see the infusion process of each bed in the ward and change the liquid in time in the medical station. The existing intelligent infusion systems mostly use Wifi and ZigBee for wireless communication, the transmission distance is small, and they are easily affected by interference.

SUMMARY OF THE INVENTION

The purpose of the present invention is to provide an intelligent infusion system based on Lora communication.

The technical solution to achieve the purpose of the present invention is: an intelligent infusion system based on Lora communication, including an infusion monitoring unit, a repeater, a pager, a wireless communication module and a medical care station host computer, the infusion monitoring unit and the pager pass through. The Lora communication module is connected with the repeater, and the upper computer of the medical station is connected with the repeater through a network cable. The infusion monitoring unit is responsible for monitoring the weight of the liquid medicine, infusion progress and other information, and transmits it to the repeater through LoRa wireless communication technology; the pager is used to send an emergency call signal and transmits it to the repeater through LoRa wireless communication technology; the repeater Obtain the infusion-related information from the infusion monitoring unit, obtain the call signal from the pager, and transmit it to the upper computer of the medical care station through the network cable; while the upper computer of the medical care station provides power to the repeater, it sends out a reminder of the end of infusion according to the information obtained from the repeater or call reminder.

Compared with the prior art, the advantages of the present invention are: 1) the present invention improves the robustness and security of information transmission through LoRa wireless communication technology; 2) the present invention improves data measurement by setting a weighing calibration module accuracy.

Description of drawings

Figure 1 is a structural diagram of an intelligent infusion system based on LoRa communication.

Figure 2 is a flowchart of polling the Lora gateway.

Figure 3 is a schematic diagram of a secure access Lora gateway.

FIG. 4 is a schematic diagram of a site according to an embodiment of the present invention.

Fig. 5 is the concrete flow chart of wavelet neural network and artificial fish swarm algorithm.

Detailed ways

The solution of the present invention will be further described below with reference to the accompanying drawings and specific embodiments.

As shown in Figure 1, the intelligent infusion system based on Lora communication includes an infusion monitoring unit, a repeater, a pager, a wireless communication module and a medical care station host computer. The infusion monitoring unit and the pager communicate with the relay through the Lora communication module. The upper computer of the medical care station is connected with the repeater through a network cable. The infusion monitoring unit is responsible for monitoring the weight of the liquid medicine, infusion progress and other information, and transmits it to the repeater through LoRa wireless communication technology; the pager is used to send an emergency call signal and transmits it to the repeater through LoRa wireless communication technology; the repeater Obtain the infusion-related information from the infusion monitoring unit, obtain the call signal from the pager, and transmit it to the upper computer of the medical care station through the network cable; while the upper computer of the medical care station provides power to the repeater, it sends out a reminder of the end of infusion according to the information obtained from the repeater or call reminder.

In some embodiments, the Lora communication module adopts a star network structure. Since the Lora gateway can only exchange information with one terminal node at a certain time, in the process of wireless transmission, the Lora communication module adopts the polling method to allow the node data to access the Lora gateway in turn according to a certain period. The polling process is shown in the figure. 2 shown.

Usually, the infusion monitoring unit measures the weight of the liquid medicine through a load cell to determine the infusion progress. However, the load cell is easily affected by temperature and produces a zero-point drift phenomenon, which leads to inaccurate weighing information, resulting in the advance or lag of the end of the infusion alarm. In order to deal with the zero drift phenomenon of the load cell affected by temperature, the present invention sets a weighing calibration module in the infusion monitoring unit, and uses wavelet neural network and artificial fish swarm algorithm to compensate the zero point temperature drift of the load cell.

Before temperature compensation, the voltage data output by the weighing sensor under different conditions is collected by fixing the weight of the liquid medicine, changing the temperature, and fixing the temperature and changing the weight of the liquid medicine; then use the wavelet neural network to learn the data samples, Through its own training, the output of the neural network is gradually approached to the expected output value; because the wavelet neural network is easy to fall into the local optimum, the artificial fish swarm algorithm is introduced to test it in order to obtain the global optimal solution.

The specific process of wavelet neural network and artificial fish swarm algorithm is described as follows:

及误差函数E；(1) Initialization parameters: for the number of nodes in the input layer m, the number of nodes in the hidden layer n, the number of nodes in the output layer l, the scaling factor a _j of the wavelet, the translation factor b _j , the difference between the input layer and the hidden layer The connection weight v _ji and the connection weight w _kj from the hidden layer to the output layer are given to the initial value, and the neuron function of the hidden layer is selected.

and the error function E;

(2) Start the wavelet neural network search: input the learning sample (gravity weighed by the load cell under different temperature conditions) and the corresponding expected output (the voltage output by the load cell), and use formula (1) to calculate the output of the output layer , using formula (2) to calculate the error;

和

分别为第i(i＝1,2,…,M)个模式的第j(j＝1,2,…,N)个期望输出和实际输出；Among them, M is the number of patterns of input samples,

and

are the jth (j=1, 2,...,N) expected output and actual output of the ith (i=1, 2,...,M) mode respectively;

(3) Judging whether to end the neural network algorithm: when the error is less than a predetermined threshold, stop the learning of the neural network, and transfer to the artificial fish swarm algorithm, otherwise input the next sample, go back to step 2, and continue the neural network learning;

(4) Fish school initialization: use the output result of the output layer of the wavelet neural network to initialize N artificial fish, the output result of the wavelet neural network is the state of the ith artificial fish, and the error function in step 2 is the artificial fish The fitness value function, at this time, the artificial fish groups gather in the same place, and the optimal solution on the bulletin board is the output result of the wavelet neural network;

(5) Verification of the optimal solution: the fish group performs foraging behavior, and searches for places with higher food concentration within the visual field (Visual) according to a certain step size (that is, the fitness value is smaller and the error is smaller). place), if it cannot be searched, it will move forward randomly. At this time, the fitness value of the artificial fish changes, but the optimal solution on the bulletin board is not updated. The output result of the wavelet neural network is the global best result; if the fish finds a place with higher food concentration during the foraging process, the optimal solution on the bulletin board changes, and the result is fed back to the wavelet neural network to continue. Perform training, and so on until the global optimal solution is found.

In some embodiments, the Lora communication module is provided with a security verification access module, and the physical unclonable function property of the chip is used to realize the security verification access of the Lora communication module, the infusion monitoring unit and the pager. The process of the security verification access is as follows. As shown in Figure 3, the details are as follows:

Due to the differences in the chip manufacturing process in each infusion monitoring module and pager, the Lora gateway sends excitation signals to different infusion monitoring units and pagers respectively, and will obtain different response signals. These responses are unique and unreproducible, and can be As the ID of each infusion monitoring unit or pager, that is, the identity "verification code"; the Lora gateway puts these identity "verification codes" into the storage module for ID access verification when the infusion monitor and pager are polled .

As shown in FIG. 4 , in some embodiments, the infusion monitoring unit and the pager are arranged near the hospital bed, and the repeater and the Lora communication module are arranged in the hospital corridor. Since the hospital bed is relatively close to the infusion monitoring unit and the pager, radio frequency identification technology is used to bind each hospital bed to an infusion monitoring unit and a pager respectively. RFID tags are set on the beds in the infusion area, and each infusion monitor and pager is bound to the label on the bed. After binding, the host of the medical station can distinguish the data sent by the infusion monitoring unit and pager from different beds. information.

In order to verify the effectiveness of the scheme of the present invention, the following simulation experiments are carried out.

1. Lora communication test

In this embodiment, the Lora communication module adopts the product E32-433T20DC of Chengdu Ebyte Company, which is a wireless serial port module (UART) based on the SX1278 radio frequency chip of SEMTECH Company, and the module adopts TTL level. Keep the default settings of the E32-433T20DC product unchanged, that is, the air transmission rate is 2.4kbps, and the transmit power is 20dBm. In the environment where there are obstacles such as walls, electronic equipment, optical equipment, etc., including 4G, WiFi, microwave, millimeter wave and other interference , conduct plane through-wall test and cross-floor test.

(1) Plane through-wall test

First, the data packets are sent and received in different rooms on the same floor. In the case of sending 150,000 packets, the test results are shown in Table 1. It can be seen that when the transmission path is as long as 60 meters (six walls are separated), the packet loss rate is only 1.187%, which is higher than the communication quality of Wifi, ZigBee and other methods.

Table 1 Plane through-wall test results table

Number of partition walls number of packets sent number of lost packets Packet loss rate 1 (10m) 150000 416 0.277% 2(20m) 150000 425 0.283% 3(30m) 150000 478 0.319% 4(40m) 150000 579 0.386% 5(50m) 150000 1216 0.811% 6(60m) 150000 1780 1.187%

(2) Cross-floor test

Taking the third floor as the benchmark, data packets are sent to other floors respectively. The test results are shown in Table 2. It can be seen that when the signal is transmitted to the sixth floor, the packet loss rate is only 1.271%, which is higher than the communication quality of Wifi, ZigBee and other methods.

Table 2 Cross-floor test result table

interval floor number of packets sent number of lost packets Packet loss rate 1 (to the fourth floor) 120000 355 0.296% 2 (to the fifth floor) 120000 390 0.325% 3 (to the sixth floor) 120000 1525 1.271%

2. Weighing calibration test

In the weighing calibration module, there are many optional weighing sensors. Different weighing sensors have different zero temperature drift laws due to differences in the manufacturing process. Therefore, the temperature compensation should be carried out according to the own laws of the weighing sensor. Taking the PT14 pressure sensor as an example, the output voltage data under different conditions are measured, as shown in Table 3.

Table 3 Measurement data table of PT14 pressure sensor

t(℃) P(KPa) 100 101 102 103 104 105 106 107 108 23.4 U_P(mV) 21.6 43.0 61.8 81.0 99.1 118.9 139.2 153.4 183.5 30 U_P(mV) 20.0 39.6 58.4 78.9 98.7 120.2 137.0 150.2 179.9 37 U_P(mV) 18.8 39.4 58.9 77.9 95.4 113.9 133.3 152.8 173.9 44 U_P(mV) 5.5 27.9 45.6 63.1 83.0 101.4 120.0 139.5 157.7 54 U_P(mV) 7.9 27.2 46.4 63.5 84.5 102.4 120.0 139.4 158.3 60 U_P(mV) 4.6 21.2 37.6 54.1 73.1 95.4 103.8 126.9 148.7 70 U_P(mV) 4.9 18.5 31.3 49.6 67.4 86.9 96.8 118.9 137.3

In Table 3, t is the ambient temperature where the pressure sensor is located, P is the pressure carried by the pressure sensor, and U _P is the output voltage of the pressure sensor.

After the data collection is completed, the wavelet neural network and artificial fish swarm algorithm are executed according to Figure 5. Parameter initialization includes: setting the input layer nodes of the wavelet neural network to 9 according to the 9 values of the pressure P (100-108), determining the number of modes of the input sample to 7 according to the 7 values of the temperature t, and setting the output node to 1 is used to output the results; according to the experience of existing research, set the number of nodes in the hidden layer to 16, the maximum training times of the network to 500, the learning rate to 0.01, the momentum factor to 0.95, and the training target error sum of squares to 0.0001; The hidden layer neuron function uses the Morlet wavelet function, and its function expression is:

In addition, set the size of the fish school N=20, the field of view Visual=0.7, the step size Step=0.5, and the maximum number of iterations is 400. The above parameters are not intended to limit the present invention, and improvements and optimizations can be made in the embodiments according to specific conditions.

Next, the training of the wavelet neural network is carried out, the gravity and the output voltage of the load cell under different temperature conditions are used as the learning samples, the output result is calculated by formula (1), and the error is calculated by formula (2); If the error is less than the set threshold, it is transferred to the artificial fish swarm algorithm, and the output result of the wavelet neural network is used as the initial state of the fish swarm.

Whether the output is globally optimal is verified by letting the artificial fish perform foraging behavior. Let the current state of the i-th artificial fish be X _i , and the corresponding fitness value is Y _i (calculated by formula (2)), and randomly select a state X _v within its Visual range, such as formula (4), if the corresponding If the fitness value Y _v <Y _i , the optimal solution on the bulletin board is updated to X _v , and feedback is given to the wavelet neural network to continue training to improve the result; otherwise, the artificial fish randomly advances one step and becomes state X _inext , as in Equation (5), and the bulletin board does not need to be updated.

X _v =X _i +Rand()*Visual (4)

X _inext =X _i +Rand()*Step (5)

Repeat the above steps until the wavelet neural network outputs the best result, that is, the temperature compensation reaches the best effect under the specified error.

Claims (5)

1. a kind of intelligent infusion system based on LoRa communication, it is characterized in that: comprise infusion monitoring unit, pager, repeater, Lora communication module and medical station host computer, described infusion monitoring unit, pager pass through Lora communication module and The repeater is connected, and the upper computer of the medical station is connected with the repeater through a network cable; Among them, the infusion monitoring unit is equipped with a weighing calibration module, and the wavelet neural network and artificial fish swarm algorithm are used to compensate the zero-point temperature drift of the weighing sensor. The specific process is described as follows: (1) Initialization parameters: for the number of nodes in the input layer m, the number of nodes in the hidden layer n, the number of nodes in the output layer l, the scaling factor a _j of the wavelet, the translation factor b _j , the difference between the input layer and the hidden layer The connection weight v _ji and the connection weight w _kj from the hidden layer to the output layer are given to the initial value, and the neuron function of the hidden layer is selected.

and the error function E; (2) Start the wavelet neural network search: input the learning sample and the corresponding expected output, use the formula (1) to calculate the output of the output layer, and use the formula (2) to calculate the error;

Among them, M is the number of patterns of input samples,

and

are the jth, j=1, 2,..., N expected outputs and actual outputs of the i, i=1, 2, ..., M modes, respectively; (3) Judging whether to end the neural network algorithm: when the error is less than a predetermined threshold, stop the learning of the neural network, and transfer to the artificial fish swarm algorithm, otherwise input the next sample, go back to step 2, and continue the neural network learning; (4) Fish school initialization: use the output result of the output layer of the wavelet neural network to initialize N artificial fish, the output result of the wavelet neural network is the state of the ith artificial fish, and the error function in step 2 is the artificial fish The fitness value function, at this time, the artificial fish groups gather in the same place, and the optimal solution on the bulletin board is the output result of the wavelet neural network; (5) Verify the optimal solution: the fish group performs foraging behavior and searches for places with higher food concentration within the field of view according to a certain step size. If it cannot be found, it will randomly move forward one step. Change but the optimal solution on the bulletin board is not updated. After a certain number of iterations, the optimal solution on the bulletin board remains unchanged, and the output result of the wavelet neural network is determined to be the global optimal result; if the fish is in the foraging process If a place with a higher concentration of food is found, the optimal solution on the bulletin board will change, and the result will be fed back to the wavelet neural network to continue training, and so on until the global optimal solution is found. 2. The intelligent infusion system based on LoRa communication according to claim 1, is characterized in that: Lora communication module adopts star network structure, and in the information transmission process, adopts polling mode to allow node data to visit Lora gateway in turn. 3. the intelligent infusion system based on LoRa communication according to claim 1, is characterized in that: described Lora communication module is provided with security verification access module, utilizes the physical unclonable function property of chip, realizes Lora communication module and infusion monitoring unit and secure access to the pager. 4 . The intelligent infusion system based on LoRa communication according to claim 1 , wherein the infusion monitoring unit and the pager are arranged near the hospital bed, and the repeater and the LoRa communication module are arranged in a hospital corridor. 5 . 5. The intelligent infusion system based on LoRa communication according to claim 1, is characterized in that: each sickbed corresponds to an infusion monitor and a pager, and the RFID tag is respectively set on the sickbed of the infusion area, and the radio frequency identification technology Bind each infusion monitor, pager to the label on the bed.

Images