Open Access
Add to BibSonomy
Abstract
A multi-channel optical sensing system for heavy metal concentration detection is presented in this paper. The system utilizes a multi-channel optical path combined with a unique circuit design and BP neural network (BP-ANN) to perform the online analysis of multi-wavelength signals, achieving accurate concentration detection of a heavy metal solution. An array photodiode is used to detect the transmission light intensities at multiple wavelengths through the optical path of the system, which enables the collection of useful spectral information of the solution. The system uses a unique signal acquisition method to effectively improve the efficiency of both signal acquisition and operation. BP-ANN is applied to the online analysis of multi-channel information, which overcomes the influential issue of nonlinear effect on data detection, optimizes the anti-interference ability, and lowers the detection limit of the system. This system eliminates the necessary employment of the expensive and large spectrometers and therefore greatly reduces the instrument cost and occupying space. Additionally, the detection limit of the system is extended lower than that of the conventional spectrophotometer. Compared with the detection limits of heavy metal solution obtained by using a single characteristic light wavelength, the detection limits of Cd2+, Cu2+ and Cr6+ achieved through using multi-channel detection system can be reduced by 42.64%, 38.12%, and 20.62%, respectively, and these detection limits are found as 0.0041mg/L, 0.0091mg/L, and 0.0112mg/L, respectively.
© 2022 Optica Publishing Group under the terms of the Optica Open Access Publishing Agreement
1. Introduction
Cadmium (Cd2+), chromium (Cr6+), mercury (Hg2+), copper (Cu2+) and lead (PB2+) in water are harmful to human body [1]. The amount of each of these heavy metals in water is strictly controlled and monitored in many countries [2,3]. At present, how to overcome noise interference is still a big challenge for in-field detection of these heavy metal elements. Therefore, it is especially important to develop a method of fast detection of heavy metal content in a target water body in the condition of external noise interference [4].
Currently, heavy metal detection instrument for water environment can be categorized into optical sensing and non-optical sensing. Anodic stripping voltammetry (electrochemistry) is a common non-optical sensing method, which has the characteristics of high sensitivity and easy operation [5]. However, this method has the problems of being interfered by organic substances in the water and easily causing secondary pollution to the water environment.
Instruments based on optical sensing methods can be protected from contamination due to the characteristic of non-contact detection [6]. Optical sensing methods can be categorized into atomic absorption spectrometry and spectrophotometry [7]. Atomic absorption spectrometer has the advantages of fast detection, high sensitivity, and strong anti-interference ability. But the instrument can only be used to detect one element, and the instrument equipment is expensive [8]. The existing spectrophotometers detect the concentration of the solution by the characteristic peak signals corresponding to different heavy metal ions. In essence, the single-wavelength signal is used to calibrate the concentration of heavy metal ions, and other optical signal responses near the characteristic peak wavelength are ignored [9]. Spectrometer can analyze heavy metals by detecting complete spectral signals, but due to its internal structure design, there exist its shortcomings such as being bulky and expensive. To solve these problems, it is very important to design an intelligent and miniaturized optical sensing instrument which is portable, low-cost, and fast-detection [10].
In recent years, portable spectrophotometers have been widely used in field detection, such as ZZW- II tester and PORS-15V spectrometer [11]. However, such instruments can only detect the heavy metal content in water roughly, and the detection limit does not meet the international standard of heavy metal content in drinking water.
In summary, the article proposes a multi-channel optical sensing system that can quickly take measurements in real-time, based on the Beer-Lambert Law [12]. According to the characteristic peaks of the absorption spectrum of heavy metals ions, the laser light source corresponding to each characteristic peak and its adjacent wavelengths is selected as the detection light source. The multi-wavelength signal obtained by the detection was used to analyze and detect the heavy metal solution, and the detection sensitivity was higher than that of the ordinary spectrophotometer [13,14]. Compared with the single-channel detection instrument, it improves the reliability of the system, overcomes the problem of detection inaccuracy in the nonlinear detection area, and reduces the detection error.
2. Principle and design
In this paper, a design of a multi-channel optical detection system for heavy metals in water environment based on spectrophotometry is proposed. This system was used to detect cadmium (Cd2+), chromium (Cr6+) and copper (Cu2+) in water environment. Multi-wavelength signals are collected by an array photodiode, which accomplishes the accurate concentration detection of a solution [15]. It effectively reduces the system’s spatial volume, lowers the system cost, shortens the detection time, and improves the sensing sensitivity. The system uses unique signal acquisition methods to improve the efficiency of both signal acquisition and operation, and effectively increases the response rate of the system. The designed photoelectric conversion circuit has the ability to amplify the signal and resist high-frequency interference. A closed-loop acquisition circuit composed of analog-to-digital conversion (A/D) and digital-to-analog conversion (D/A) is designed to enhance the signal acquisition accuracy of the system. The system uses BP neural network (BP-ANN) to predict multi-channel data acquisition, which overcomes the influential issue of nonlinear effects on data detection, improves anti-interference ability, and lowers the detection limit of the system [16–18].
2.1. Design of the sensing optical path
The transmission path of the system is composed of an array photodiode that is used as a light detector and multiple different single-wavelength light sources that are used as detection light sources [19]. Compared with the spectrometer using a photomultiplier tube, our system uses a photodiode instead, which has the advantages of low noise, low cost, small size, light weight, long life, high quantum efficiency, and good frequency effect. It is suitable for rapid change of optical signal detection. Photodiode also has better current linearity, which is conducive to the design of a low-cost, anti-interference portable optical sensing system.
Due to the poor selectivity of a single photodiode, the continuous signal within the detection range will be collected without selection. The system uses a fixed wavelength laser as a detection light source to improve monochromaticity and stability. Multiple photodiode arrays are used to realize the simultaneous detection of multiple wavelength signals, which not only improves the detection rate but also avoids the problems caused by the instability of monochromator performance related to the sensitivity, selectivity, and linear relationship of the calibration curve of the system. Moreover, the laser beam is less affected by the liquid flow and has higher sensitivity when transmitting the solution. It can effectively offset the influence of adaptive noise, downsize the system volume, and reduce the cost of the whole system.
In this paper, the use of four different wavelength semiconductor lasers with the same number of photodiodes as the transmission path is proposed for the first time. It achieves the collection of spectral information of heavy metals. CPS (Thorlabs) series 405nm, 450nm, 520nm, 532nm semiconductor lasers were used as light sources [20]. FDS100 (Thorlabs) photodiodes were used for light signal acquisition. A cuvette holding the test solution was placed directly between the light source and the detector to avoid the loss of the light source in the air. The optical transmission and sensing path is shown in Fig. 1. The system can effectively reduce the system volume, lower the system cost and shorten the detection time under the premise of improving the sensitivity as much as possible.
2.2. Sensing principle and circuit design
The system innovates the detection mode of absorbance and demarcates absorbance according to the value of detected voltage difference, which transforms the sensing signal from logarithmic relationship to linear relationship, making the sensing signal more intuitive. Therefore, the calculation amount of the system is reduced, and the response rate is improved. A signal calibration and acquisition circuit with signal amplification and anti-interference performance is designed to monitor the sensor signals collected from each channel in real-time. A closed-loop feedback control circuit formed by a D / A module and an A / D module is continuously controlled by the MCU (Microcontroller Unit). The system can more accurately obtain the light intensity signal even out of small concentration changes, thereby improving the resolution of the system to detect the concentration of the solution.
The main objectives of our system design include integration and miniaturization. To achieve this goal, the system uses MSP430 as the core controller, with a visual interface and key control module [21]. The model of the whole system is designed using SolidWorks, as shown in Fig. 1, and the schematic diagram of the control panel is shown in Fig. 2.
The optical circuit of the system designed in this paper is based on Beer-Lambert Law, which is the theoretical basis for colorimetric analysis and spectrophotometry. It is given in (1).
Photodiode and AD8304 logarithmic amplifier are used to build photoelectric conversion circuit. The system detects the absorbance of the solution to be tested more quickly and accurately. The logarithmic relationship between the logarithmic amplifier base to emitter voltage (VBE) and collector current (IC) is shown in (2) [22]. The linear relationship between the output voltage of the system and the absolute loss of the detected light intensity is shown by (3). ${P_d}$ is expressed as: The voltage signals Vin and Vt corresponding to the optical power of the incident light and the transmitted light, respectively, which are detected in the system. The voltage difference detected is: The absorbance detected by the system is expressed as In order to filter out the high-frequency noise signal and collect more stable signal values, the external circuit of the light intensity detection circuit and the chip constitute a Sallen-Key filter, which can filter out the high-frequency noise signal with a frequency greater than 1 kHz [23]. The non-inverting operational amplifier is designed for the circuit output to obtain the gain, so that the output voltage gain reaches three times, thereby improving the sensitivity of signal acquisition. The system uses the multi-channel selection chip 74HC4051 to control the turn-on of multiple photodiodes to achieve the purpose of collecting information from multiple optical sensing channels at the same time.This module allows to reduce the number of the logarithmic amplifier, avoid the problem that too much power consumption of operational amplifier device affects the insufficient supply of other modules, lower the power consumption of the system, and make the system design more integrated and the circuit structure much simpler. The circuit schematic diagram is shown in Fig. 3 (a).
Fig. 3. Schematic diagram of multi-channel light intensity acquisition circuit including multi-channel photoelectric conversion circuit and ad-da calibration circuit.
The signal closed-loop control acquisition circuit module is designed to read the digital signals more accurately, which is realized by the MCU controlling the A / D circuit and D / A circuit. The A / D circuit module and the D / A circuit module use the ADS1115 analog-to-digital chip and DAC8552 digital-to-analog converter chip of TI (Texas Instrument). The system detects the light intensity signal of each channel in real-time through the MCU, continuously corrects the reference voltage, and controls ADS1115 chip for appropriate PGA (Programmable Gain Amplifier Configuration) configuration, which is to realize the closed-loop control detection of light intensity. The system can obtain a more accurate digital signal. The circuit schematic diagram is shown in Fig. 3 (b).
2.3. Feature recognition and BP-ANN establishment
In this paper, BP neural network is applied to the multi-channel signal acquisition system to achieve accurate prediction of detection data, overcome the influence of nonlinear effect on data detection and reduce the detection limit of the system [24,25].
The detection limit of the detection instrument is inversely correlated with the sensitivity, and is positively correlated with the standard deviation of the sample detection. Sb is the Standard Deviation of the instrument test sample data, and its expression is shown in (8). The relationship between Sb and the detection limit is shown in (7).
Fig. 4. BP-ANN structure diagram. The neural network has three hidden layers (the number of hidden layers is 5, 4, and 4 respectively).
According to the Beer-Lambert Law, the absorbance is affected by the optical path of the sensing path, the properties of the absorbing material, and the wavelength λ of the incident light. Information collected by single-channel sensing is only determined by a single channel, which is vulnerable to external interference such as the input voltage fluctuation of the light source, the external temperature change, and contamination dusts in the optical path. Single-wavelength sensing detection is extremely unstable. Multi-channel sensing combined with BP-ANN was proposed for concentration detection in this paper. BP-ANN was used to overcome the influence of the nonlinear effect on concentration detection and improve the detection range of the system for the solution to be tested with a nonlinear effect [27,28]. Multiwavelength sensing detection mode has a higher anti-interference ability and fault-tolerant rate compared with Single wavelength sensing detection mode, which effectively avoids the noise influence, improves the accuracy of low concentration solution detection, and lowers the detection limit of the system.
3. Experiment and analysis
3.1. Stability of the sensing system
The circuit layout is extremely important to the stability of the sensing system. The design of the PCB (Printed Circuit Board) is shown in Fig. 5. The total supply voltage of the system is designed with a DC-DC step-down circuit (The DC-DC chip is TI's TPS5430), which converts the external input 12V to 5V. Considering that the MCU control voltage of the system is different from the supply voltage of the entire system, the 5V output is reduced to 3.3V through a linear LDO (Low Dropout Regulator) as the MCU control voltage. To avoid the mutual influence of the two voltages, the MCU control module is placed separately from other modules. The light intensity signal collected by this system will be converted into a micro-current signal, and the temperature of the source module will increase due to the increase in the time used. Therefore, the photoelectric signal conversion module on the PCB of the system design is kept at a certain distance from the source module.
3.2. Calibration of sensor circuit acquisition
With the increase of the working time of the laser, the working temperature of the laser will increase, the wavelength shift will occur in the system detection, and the detected signal will be affected. The system is equipped with a temperature acquisition module to detect the temperature change of the laser. The system needs to be preheated each time the system is used. When the temperature of the laser reaches about 35°C, the light source of the laser gradually enters a stable state. Prior to each use of the system, it is necessary to check the data of the blank group to set it as the baseline in advance and perform the zero-adjustment operation on the system.
3.3. Experiment process
Data collected by the system is the digital signal that is read by the A/D acquisition module using different gain amplifier configurations. The data value is proportional to the absolute power detected by the photodiode according to the above formula (1). The digital signal corresponding to the emitted light and the incident light collected by the system is proportional to the absorbance of the solution according to (7). The system is turned on to perform the pre-heat treatment until the system is stable. Then the button of data collection is pressed to gather data on the incident light. After the collection is complete, the system will display a reminder for the solution to be placed for testing. After the solution is in place, the data collection button is pressed again. Next, the solution is removed and the button of data recording is pressed. Once the data is recorded, a data completion reminder will be displayed. Finally, the collected data is saved into the PC, and the plots as shown in Fig. 6 are obtained after the drawing processing is completed.
Fig. 6. (a), (b), and (c) are the graphs of the voltage difference collected from the detection of Cd2+, Cr6+, and Cu2+ solutions at different concentrations over time. (d) is the curve diagram of the voltage difference collected by the system detecting the Cd2+, Cr6+, and Cu2+ solutions as a function of concentration.
The detection of the experiment is based on spectrophotometry. The reagents used for the detection of three heavy metal ions are: Cu2+ solution using DDTC (Diethyldithiocarbamate, (C2H5)2NCSSNa)reagent, Cd2+ using sodium hydroxide ethanol solution with Cadion(C18H14N6O2) added, Cr6+ using Diphenyl carbamide (OC (HNNHCsHs)2) in acetone solution [29,30].
Different kinds of solutions show different colors after their corresponding chromogenic reagents are added, and each color depth is also different. The solution with Cd2+ turns purple-red after Cadion is added; when the concentration of Cd2+ is increased, the solution turns brown-red. The solution with Cu2+ and the solution with Cr6+ change from transparent liquids to yellow-brown and purple, respectively, after their corresponding chromogenic reagents are added in. As the concentration increases, the color gradually turns darker.
The resulting colors of the solution do have a negative effect on the laser beam. The loss effect of the laser beam is greater when the solution grows darker in color. If the color of the solution is relatively light, the laser beam’s loss effect will be less significant. Therefore, different types of ions require different PGA configurations. The detection of solution with Cd2+ uses PGA with a differential voltage of ±1.096V, and the detection of solution with Cu2+ and the one with Cr6+ both use a PGA with a differential voltage of ±0.256V. According to the characteristic light wavelength corresponding to the solution with Cd2+, the 450nm laser sensing channel among the multi-channels is selected since it has the closest matching to the characteristic light wavelength, and the detection result is shown in Fig. 6(a). The solution with Cu2+ and the one with Cr6+ respectively select the data read by the 450nm and 532nm laser sensing channels which are similar to their corresponding characteristic wavelengths, as shown in Figs. 6(b) and 6(c). From Figs. 6 (a), (b), and (c), it can be seen that under the same condition of ADC16 (16-bit analog-to-digital conversion digits), the reading range is smaller and the accuracy is higher when detecting solution with Cu2+ and with Cr6+. However, the influence of external noise is greater when the acquisition accuracy is higher, which results in greater fluctuations in the acquisition curve.
As an example, the data obtained by testing the 0.2mg/L, 0.4mg/L, 0.6mg/L, and 0.8mg/L solutions in the experiment, the voltage amplitude obtained becomes larger and larger when the concentration increases. As shown in Fig. 6(d), it can be concluded that the collected data difference (voltage difference) is proportional to the absorbance of the testing solution based on Beer-Lambert's law.
3.4. Experiment analysis
The system detects different concentrations of heavy metal solutions of Cd2+, Cu2+, and Cr6+, respectively and collects their sensing data of different wavelengths. The concentration and voltage amplitude relationship curves are established with the data collected by the detection channels of the corresponding characteristic wavelengths, and the results obtained by data fitting are shown in Fig. 7.
Fig. 7. (a), (b), and (c) are the fitting curve diagrams of the data obtained by the system collecting Cd2+, Cu2+, and Cr6+, respectively.
It can be seen from Fig. 7 that the detection curves of Cu2+ and Cr6+ are linear. When the detection range is exceeded, the system will not be able to detect. The Cd2+ detection plot shows a linear relationship only in the range of 0∼1mg/L, which is shown in Fig. 7(a). Outside the detection range, there is no linear relationship between the absorbance and the concentration of the tested solution The detected voltage amplitude reaches a threshold at 2.34ppm.
3.4.1 Reduce detection error
The system uses single-channel mode and multi-channel mode to detect different concentrations of heavy metal solutions, and the system operation process is shown in Fig. 8.
Taking Cr6+ as an example, the same volume of chromogenic reagent is added dropwise to different concentrations of Cr6+ solution, and the system performs 6 detections on the solution with the same concentration. The test results are available in Table 1.
Table 1. The test result table of the system on the Cr6+ solution
According to the table, the percentage error can be greatly reduced when the system uses multi-channel detection mode to detect different concentrations of Cr6+. The experimental results show that the percentage error reduction effect is more significant in the system using multi-channel detection mode than using single-channel detection mode to detect the solution.
As shown in Fig. 9 (a), the results obtained by the system using different detection modes and the line of its standard deviation can clearly show that the detection percentage error and stability are smaller when the system uses the multi-channel detection mode to detect low-concentration Cr6+ solutions. According to the standard line of the actual concentration, the concentration detected by a single-channel sensor is quite different from the actual concentration. As shown in Fig. 9 (b), the accuracy of single-channel sensing is lower than that of multi-channel sensing, the measurement accuracy can be maintained above 92% when the system uses the multi-channel mode to detect low-concentration Cr6+ solutions.
Fig. 9. (a)is a data analysis chart of the system detecting low-concentration Cr6+ (Including the measured mean value and sample standard line of the two detection modes, and the actual concentration.). (b) is a line graph of the measurement accuracy of different concentrations of Cr6+ in different detection modes.
3.4.2. Overcome the influence of non-linear effects
BP-ANN can overcome the interference of the nonlinear function relation and predict the target value more accurately. The method of combining the multi-channel sensing and BP-ANN to detect the concentration of a solution is proposed in this paper. The solution of Cd2+ with a non-linear region of sensing relationship can be detected by the system which also conducts network training on the collected data. The trained network can achieve an accuracy rate of more than 96% in predicting the concentration of the solution. To verify the reliability of the network in the nonlinear detection area, the system detects different concentrations of solution with Cd2+, and the detection results are shown in Table 2.
Table 2. The test result table of the system for different concentration levels of Cd2+ solution
Fig. 10. (a) is a data analysis graph of Cd2+ solution at different concentration levels (Including the measured mean value and sample standard line of the two detection modes, and the actual concentration.) (b) is a comparison of detection error of Cd2+ in different concentrations between multi-channel detection mode and single-channel detection mode.
When the system detects close to the concentration detection limit, the detection concentration data has a large error compared with the actual concentration. The system can reduce the detection error by 12% with multi-channel detection mode compared with single-channel detection mode. The system detection concentration accuracy rate is as high as 97% in the linear detection range.
It can be concluded from the standard deviation data in Fig. 10(a) and Table 2 that the stability of the multi-channel detection mode is higher than that of the single-channel detection mode. Therefore, the reliability of the detection data is higher with the multi-channel detection mode. As shown in Fig. 10(b), as the concentration of solution increases and enters the non-linear detection range, the system can still maintain high accuracy by the multi-channel detection mode, but the curve of single-channel sensing detection will gradually shift away from the actual concentration curve and the error gradually increases. Therefore, the single-channel sensing detection is only suitable in the linear detection range. The system overcomes the impact issue of non-linear effects on detection by the use of multiple channels combined with BP-ANN, and as a result, expands the detection of solution concentration with Cd2+ from the existing range of 0.005 mg/L ∼ 1 mg/L to the new range of 0.005mg/L∼2 mg/L.
3.4.3. Lower the detection limit of the system
Since the system detects heavy metal ion solutions based on colorimetry, the system has the highest usage response in the same concentration range detected among the methods using the characteristic light sensor channel corresponding to the solution. Whichever detection mode the system uses to detect heavy metal solutions, the sensitivity of the system is basically the same. In order to verify whether the trained network can lower the detection limit, the trained BP-ANN is implanted on the single-chip microcomputer. The system detects the three ionic solutions of Cd2+, Cu2+, and Cr6+, respectively. Compared with the blank group, the system can detect the concentrations of the three solutions as low as 0.005mg/L, 0.005mg/L, 0.006mg/L, respectively. The results obtained after testing multiple sample data are shown in Table 3 and Fig. 11.
Table 3. The Precision Test Result Table of the System
Fig. 11. (a) is the result analysis of the detection of low-concentration heavy metal ion solutions by the system using the multi-channel detection mode and the single-channel detection mode respectively (Including the measured mean value and sample standard line of the two detection modes, and the actual concentration.). (b) is the LOD of the system using different detection modes.
It can be seen from Table 3 that the system uses the multi-channel mode instead of the single-channel detection mode to detect Cu2+ solution, and the detection limit reaches 0.0091mg/L which is a 42.64% reduction and the detection error at low concentrations is smaller. Similarly, the detection limits of the system for detecting Cd2+ and Cr6+ have reached 0.0041 mg/L and 0.0112 mg/L, respectively, corresponding to their 38.12% and 20.62% by deduction. The detection range of the system for Cu2+ solution is 0.009∼3.5 ppm, the detection range for Cd2+ solution is 0.004∼2.34 ppm, and the detection range for Cr6+ solution is 0.011ppm∼8 ppm. Highlighted in blue font in 3.4.3 of the revised manuscript.
The above experimental results show that the performance of the system has been improved in all aspects for the system using multi-channel combined with BP-ANN compared to the system using the single-channel detection mode to detect the concentration of heavy metal solutions. For different concentration levels of the solution, the multi-channel detection mode can greatly reduce the detection error, overcome the impact problem of nonlinear effects on the system detection, improve the stability of the detection of low-concentration solutions, and lower the detection limit of the system.
4. Conclusion
Aiming at the detection of heavy metal ion concentration in a water environment, this paper presents the design of a portable multi-channel optical sensor detection system. The system can detect a solution’s concentrations of Cd2+, Cu2+ and Cr6+. The experimental results demonstrate that the voltage difference between incident light and the transmitted light is proportional to the solution concentration. The linearity between the voltage difference and the solution concentration directly verifies the theory of the sensing circuit in Section 2.2 based on Lambert-Beer's law. The proposal of this sensing method not only simplifies the complexity of the optical path design but also reduces the computational complexity of the MCU. The computing volume of the system will be reduced to 21.38%, and the computing time will be reduced by 28.5us. Highlighted in green font in section 4 of the revised manuscript. The system combines BP-ANN with a single-chip microcomputer and applies it to the design of a multi-channel optical sensing system, which overcomes the detection problems in the non-linear region, improves the detecting accuracy of low-concentration solutions, enhances the anti-interference performance, and lowers the detection limit of the system. The system uses multi-channel mode to detect Cu2+ solution, where the detection limit is reduced by 42.64% with the detection limit as 0.0091mg/L, and the detection limits for Cd2+ and Cr6+ have reached 0.0041mg/L and 0.01124mg/L, respectively. Presently, the conventional spectrophotometer has a detection limit of 0.028mg/L for Cu2+. The detection limit of the portable spectrophotometer PORS-15V for Cu2+ can only reach 0.2mg/L. Therefore, the performance of the system in detecting heavy metals in the water environment is much better than that of the portable spectrophotometer. Compared with the heavy metal detection instrument MODEL 9830, which is also based on spectrophotometry, the detection time of the system is less than 2 minutes (the detection time of MODEL 9830 is less than 20 minutes), the volume is only 0.34%, and the cost price is 10%. Therefore, under the condition that the water quality testing standards can be met, the advantages of the system in terms of price, volume, and response rate are much better than high-precision optical testing instruments.
Funding
National Natural Science Foundation of China (62175021, 11904038, 51902033); Sichuan Province Science and Technology Support Program (2021ZYD0033, 2020YFQ0040, 2020YJ0429, 2020YJ0431, 2021YFG0020); Chengdu Technology Innovation and Research and Development Project (2021-YF05-02420-GX, 2021-YF05-02422-GX, 2021-YF08-00159-GX); Finance Science and Technology Project of Hainan Province (ZDKJ2020009); Open Project Program of State Key Laboratory of Vanadi-um and Titanium Resources Comprehensive Utilization (2021P4FZG08A); Open Project Program of State Key Laboratory of Marine Resource Utilization in South China Sea (MRUKF2021036).
Disclosures
The authors declare no conflicts of interest.
Data availability
The data that support the findings of this study are available from the corresponding author upon reasonable request.
References
1. S. Haron and A. K. Ray, “Optical biodetection of cadmium and lead ions in water,” Medical engineering & physics 28(10), 978–981 (2006). [CrossRef]
2. M. Tian, L. Fang, X. Yan, W. Xiao, and K. H. Row, “Determination of heavy metal ions and organic pollutants in water samples using ionic liquids and ionic liquid-modified sorbents,” Journal of Analytical Methods in Chemistry 2019(14-15), 1–19 (2019). [CrossRef]
3. G. Shtenberg, N. Massad-Ivanir, and E. Segal, “Detection of trace heavy metal ions in water by nanostructured porous Si biosensors,” Analyst (Cambridge, U. K.) 140(13), 4507–4514 (2015). [CrossRef]
4. M. Faraji, Y. Yamini, A. Saleh, M. Rezaee, M. Ghambarian, and R. Hassani, “A nanoparticle-based solid-phase extraction procedure followed by flow injection inductively coupled plasma-optical emission spectrometry to determine some heavy metal ions in water samples,” Anal. Chim. Acta 659(1-2), 172–177 (2010). [CrossRef]
5. AJ. Borrill, N. E. Reily, and J. V. Macpherson, “Addressing the practicalities of anodic stripping voltammetry for heavy metal detection: a tutorial review,” Analyst 144(23), 6834–6849 (2019). [CrossRef]
6. P. Yeh, N. Yeh, C. H. Lee, and T. J. Ding, “Applications of LEDs in optical sensors and chemical sensing device for detection of biochemicals, heavy metals, and environmental nutrients,” Renewable and Sustainable Energy Reviews 75, 461–468 (2017). [CrossRef]
7. E. Pehlivan and S. Cetin, “Sorption of Cr(VI) ions on two Lewatit-anion exchange resins and their quantitative determination using UV–visible spectrophotometer,” J. Hazard. Mater. 163(1), 448–453 (2009). [CrossRef]
8. L. Jiang, J. Wu, Q. Li, G. Deng, X. Zhang, Z. H. Li, K. Chen, and K.S. Chiang, “Photochromic dye doped polymeric Mach-Zehnder interferometer for UV light detection,” Journal of Materials Chemistry C 7(21), 6257–6265 (2019). [CrossRef]
9. B. Wang, Y. Huang, Z. Liang, and Z. Zheng, “Parameter estimation of multi-wavelength interdigital sensors based on optimized neural network,” 2014 IEEE International Conference on Mechatronics and Automation (ICMA). IEEE (2014).
10. H. Endo, Y. Yonemori, K. Hibi, H. Ren, T. Hayashi, W. Tsugawa, and K. Sode, “Wireless enzyme sensor system for real-time monitoring of blood glucose levels in fish,” Biosens. Bioelectron. 24(5), 1417–1423 (2009). [CrossRef]
11. C. Chai, G. Liu, L. Feng, X. Liu, B. Yao, and L. Wang, “Towards the development of a portable sensor based on a molecularly imprinted membrane for the rapid determination of salbutamol in pig urine,” Anal. Chim. Acta 675(2), 185–190 (2010). [CrossRef]
12. K. Fuwa and B. L. Valle, “The physical basis of analytical atomic absorption spectrometry. the pertinence of the beer-lambert law,” Anal. Chem. 35(8), 942–946 (1963). [CrossRef]
13. K. Venkataraayan, S. Askra Ba, K. E. Alameh, and C. L. Smith, “Photonic-based multi-wavelength sensor for object identification,” Opt. Express 18(4), 3774–3783 (2010). [CrossRef]
14. K. Venkataraayan, S. Askra Ba, K. E. Alameh, and C. L. Smith, “Optical-cavity-based multiwavelength sensor for spectral discrimination and object position detection,” J. Lightwave Technol. 29(16), 2365–2371 (2011). [CrossRef]
15. D. N. Buchanan and J. G. Thoene, “Photodiode array detection for liquid chromatographic profiling of carboxylic acids in physiological fluids: 3-hydroxy-3-methylglutaric aciduria,” Clin. Chem. (Washington, DC, U. S.) 32(1), 169–171 (1986). [CrossRef]
16. J. Zhang, Y. Xue, Q. Sun, T. Zhang, and P. Wang, “A miniaturized electronic nose with artificial neural network for anti-interference detection of mixed indoor hazardous gases,” Sensors and Actuators B Chemical 326(1), 128822 (2021). [CrossRef]
17. Y. Yi, D. Kun, R. Li, K. Ni, and W. Ren, “Accurate temperature prediction with small absorption spectral data enabled by transfer machine learning,” Opt. Express 29(25), 40699–40709 (2021). [CrossRef]
18. L. Shi, Y. Xu, Z. Qiao, X. Xing, Z. Li, C. Zhou, and Y. He, “High-throughput Chiral Molecule Determination Based on Multi-channel Weak Measurement,” IEEE Photonics J. 13(3), 1–12 (2021). [CrossRef]
19. T. Alfredson and T. Sheehan, “Recent developments in multichannel, photodiode-array, optical LC detection,” J. Chromatogr. Sci. 24(11), 473–482 (1986). [CrossRef]
20. W. Zhang, Y. Zhao, H. Da, W. Cai, and P. Wang, “Light-addressable potentiometric sensor based on precise light intensity modulation for eliminating measurement error caused by light source,” Sensors & Actuators A Physical 185, 139–144 (2012). [CrossRef]
21. E. Dallago, “An autonomous power supply system supporting low-power wireless sensors,” IEEE Trans. Power Electron. 27(10), 4272–4280 (2012). [CrossRef]
22. J. Lee, H. G. Rhew, D. R. Kipke, and M. P. Flynn, “A 64 channel programmable closed-loop neurostimulator with 8 channel neural amplifier and logarithmic ADC,” IEEE J. Solid-State Circuits 45(9), 1935–1945 (2010). [CrossRef]
23. A. M. Durham, W. Redman-White, and J. B. Hughes, “High-linearity continuous-time filter in 5-V VLSI CMOS,” IEEE J. Solid-State Circuits 27(9), 1270–1276 (1992). [CrossRef]
24. E. Giacoumidis, A. Matin, J. Wei, N. J. Doran, L. P. Barry, and X. Wang, “Blind Nonlinearity Equalization by Machine-Learning-Based Clustering for Single- and Multichannel Coherent Optical OFDM,” J. Lightwave Technol. 36(3), 721–727 (2018). [CrossRef]
25. A. Caggiano, “Tool Wear Prediction in Ti-6Al-4V Machining through Multiple Sensor Monitoring and PCA Features Pattern Recognition,” Sensors 18(3), 823 (2018). [CrossRef]
26. G. G. Agarwal and W. J. Studden, “Asymptotic integrated mean square error using least squares and bias minimizing splines,” Ann. Statist. 8(6), 1307–1325 (1980). [CrossRef]
27. D. Liu, L. Guo, Y. Wu, H. Lv, and G. Xu, “Anti-interference function of scale-free spiking neural network under ac magnetic field stimulation,” IEEE Trans. Magn. 57(2), 3400205 (2021). [CrossRef]
28. C. Feng, N. Zhao, G. Yin, T. Gan, and J. Duan, “Artificial neural networks combined multi-wavelength transmission spectrum feature extraction for sensitive identification of waterborne bacteria,” Spectrochimica Acta Part A Molecular and Biomolecular Spectroscopy 251, 119423 (2021). [CrossRef]
29. M. N. Uddin, M. A. Salam, and M. A. Hossain, “Spectrophotometric measurement of Cu(DDTC)2 for the simultaneous determination of zinc and copper,” Chemosphere 90(2), 366–373 (2013). [CrossRef]
30. H. Watanabe and H. Ohmori, “Dual-wavelength spectrophotometric determination of cadmium with cadion,” Talanta 26(10), 959–961 (1979). [CrossRef]
