Electrostatic dust concentration self-calibration method and system

Abstract

The application discloses a static dust concentration self-calibration zero method and system, and belongs to the technical field of environmental monitoring. A built-in relay is used to cut off the collection end to establish a zero reference in a low dust window period. A median and a median absolute deviation are used for robust rough zero estimation. A multivariate zero point compensation model is established with working time, dust concentration cumulative amount and polarization duration as inputs. In subsequent operation, a least square type adaptive algorithm is used to update model parameters only in a low dust period. Parameters are frozen in a non-low dust period to avoid learning effective signals. Finally, the model parameters are solidified into non-volatile storage and form a full life cycle management. The application does not need external standard devices and manual inspection, can long-term inhibit zero point drift under non-stop conditions, and improves long-term stability and reliability of measurement data.

Description

Technical Field

[0001] This invention relates to the field of environmental monitoring and industrial process control technology. Specifically, it relates to a method and system for self-calibrating electrostatic dust concentration, which is particularly suitable for zero-point drift suppression and long-term stability maintenance of electrostatic dust concentration measuring instruments in stationary pollution source emission monitoring, workplace occupational health protection, and process control. Background Technology

[0002] Electrostatic induction measurement of dust concentration is widely used in industrial dust emission monitoring and workshop dust concentration early warning due to its advantages such as high sensitivity, convenient installation and maintenance, and non-contact measurement. Its basic principle is to utilize the friction or induction of electric charge between flowing dust particles and measuring electrodes to generate charge, and to calculate the dust concentration by detecting the amount of charge.

[0003] Electrostatic dust concentration meters are widely used in industrial emission monitoring, mine ventilation, and cleanroom and environmental online monitoring. They characterize concentration by measuring the electrical signal generated by particle charging or induction processes, and have advantages such as simple structure, fast response, and long-term online operation. However, in conditions containing water mist, humidity, strong adhesion, or complex particle size distribution, the electrodes and sampling chamber are easily contaminated, resulting in baseline rise, decreased sensitivity, increased drift, and more false alarms, seriously affecting the reliability of the measurement.

[0004] Existing improvement solutions mostly rely on manual disassembly and cleaning, additional pre-filtration, fixed-band filtering, or periodic shutdown calibration. These solutions either require frequent maintenance and incur high downtime costs, or are only effective against certain types of interference. They lack sufficient support for early identification of contamination accumulation, online self-recovery, and measurement compensation, making it difficult to maintain consistency and traceability during long-term operation.

[0005] Therefore, there is an urgent need for a highly pollution-resistant measurement method that is designed for complex pollution scenarios and features online pollution identification, graded treatment, self-calibration, robust signal extraction, and joint compensation under operating conditions, so as to continuously obtain stable, accurate, and traceable dust concentration results without shutting down the system. Summary of the Invention

[0006] The purpose of this invention is to overcome the shortcomings of the prior art and provide a self-calibration method and system for electrostatic dust concentration. Through intelligent low dust window identification, robust coarse zero estimation, multivariate zero-point compensation model, adaptive tracking and update constraints, and full life cycle management, it achieves the technical effect of suppressing zero-point drift for a long time without external standards or manual disassembly and inspection, and without stopping the machine.

[0007] To achieve the above objectives, the present invention adopts the following technical solution:

[0008] In a first aspect, the present invention provides an online self-calibration method for electrostatic dust concentration meters, comprising the following steps:

[0009] S1: Triggering and Low Dust Window Detection

[0010] The system continuously collects measurements from the electrostatic sensing channel; calculates the DC mean and standard deviation within a fixed-length sliding window, and monitors the rate of change and fluctuation of the standard deviation; when the standard deviation is less than a preset threshold, the rate of change of the mean is less than a preset threshold, the fluctuation of the amplitude is within the allowable range, and the auxiliary concentration criterion shows that the particle concentration is less than a preset threshold, it is determined to enter the low dust window and allows the self-calibration zeroing process to be started.

[0011] The core of this step is to accurately identify the truly suitable low-dust period for zeroing through multi-dimensional statistical criteria, so as to avoid accidentally triggering zeroing when the dust concentration is high or the operating conditions fluctuate, which would cause the zeroing result to be contaminated by dust signals.

[0012] S2: Zero-reference setup and electrostatic sensing channel contamination assessment

[0013] An internal relay is used to disconnect the electrostatic acquisition terminal, and the current DC average and AC values ​​are measured to establish a new zero reference point. The contamination level of the electrostatic sensing channel is evaluated by the DC change, and the contamination level of the electrostatic sensing channel is positively correlated with the DC change.

[0014] This step measures the pure circuit background and the offset caused by electrode contamination after the relay cuts off the electrode signal. By comparing the initial values, the degree of contamination can be quantitatively assessed. At the same time, the AC noise background is measured to assess the health of the circuit.

[0015] S3: Coarse Zero Estimation and Robust Screening

[0016] The channel output is collected for a period of time during the zero reference hold period; the median of the sample is used as the coarse zero estimate, and the stability is evaluated by the median absolute deviation or equivalent robustness index; when the robustness index is less than the threshold and the reference duration is not less than the minimum requirement, the coarse zero estimate is deemed valid; otherwise, the sampling is automatically extended or the current self-calibration zero is terminated.

[0017] This step uses the median instead of the mean for estimation, which can effectively resist the effects of transient disturbances and abnormal spikes; the median absolute deviation, as a robustness indicator, can objectively reflect the stability of the zero reference value.

[0018] S4: Construction of Multivariable Zero-Point Compensation Model

[0019] A zero-point compensation model is established with working time, cumulative dust concentration, and polarization duration since the last zeroing as inputs. The model includes a first-order term, a second-order term, and a time-decaying component to characterize the influence of the environment and working conditions on the zero point. The initial estimation and validity verification of the model parameters are completed using the zero reference data obtained from the coarse zero estimation step.

[0020] The model constructed in this step comprehensively considers three main drift mechanisms: long-term aging, dust accumulation effect, and charge accumulation. It can more accurately describe the evolution of the zero point and lay the foundation for subsequent adaptive tracking.

[0021] S5: Adaptive Zero Tracking and Constraint Update

[0022] In subsequent operation, the model parameters are updated with small steps only during periods when the low dust window or active zero reference is met again, achieving online tracking of zero-point estimation. Parameters are frozen during non-low dust periods to avoid passively learning effective signals in real dusty scenarios. The step size and update frequency are constrained by stability rules. The model is only updated when there is confirmed no dust signal interference; parameters are frozen during other periods, fundamentally solving the problem of traditional methods easily learning dust signals.

[0023] S6: Solidification and Lifecycle Management

[0024] Write the solidified model parameters, timestamps, reference types, and confidence levels into non-volatile storage; maintain the zero-point history curve, drift rate, trigger frequency, and failure rate health indicators; when the electrostatic sensing channel contamination index is abnormal, automatically generate maintenance prompts and, if necessary, increase the self-calibration zero trigger frequency or switch to an active zero reference strategy.

[0025] This step incorporates the zero-calibration results into the entire lifecycle management, forming a traceable historical record, and predicts maintenance needs through health indicators, thereby realizing the transformation from passive maintenance to proactive prevention.

[0026] Secondly, the present invention provides an online self-calibration zeroing system for electrostatic dust concentration meters, comprising:

[0027] The electrostatic sensor module includes electrostatic induction electrodes and a charge amplifier, used to generate an electrical signal related to dust concentration.

[0028] Relay switching module: Connected between the electrostatic sensor electrode and the charge amplifier, it is used to cut off the electrode signal and establish a zero reference state when triggered;

[0029] Signal acquisition and processing module: Connected to the output of the charge amplifier, used to acquire signals and perform analog-to-digital conversion and filtering;

[0030] Environmental parameter monitoring module: includes at least one of temperature sensor, humidity sensor, and pressure sensor, used to synchronously collect environmental parameters;

[0031] Storage module: Used to store model parameters, historical zeroing data, health indicators, and lifecycle archives;

[0032] Central control module: connected to each of the above modules respectively, and configured to execute the steps of the method described in the first aspect of the present invention;

[0033] Output and Interaction Module: Used to output zero-point status, drift trend, maintenance prompts, and zeroing results;

[0034] Communication module: Used for data interaction with host computer or cloud platform, receiving remote zeroing commands, and uploading zeroing records.

[0035] Beneficial effects

[0036] Compared with the prior art, the present invention has the following significant advantages:

[0037] (1) No manual intervention required, achieving fully automatic online zeroing.

[0038] This invention, through its built-in relay and intelligent control algorithm, eliminates the need for external standards and manual disassembly and inspection, and can automatically complete zero-point calibration without shutting down the system, significantly reducing the workload of on-site maintenance.

[0039] (2) Intelligent identification of low dust windows to avoid false zero calibration

[0040] This invention accurately identifies the low-dust periods that are truly suitable for zeroing by using multi-dimensional statistical criteria (standard deviation, mean change rate, amplitude fluctuation) and auxiliary concentration criteria (light scattering comparison, historical pattern matching, process parameters).

[0041] (3) Robust estimation method with strong anti-interference ability

[0042] This invention uses the sample median instead of the arithmetic mean for coarse zero estimation, significantly enhancing its resistance to transient disturbances and abnormal spikes. The median absolute deviation, as a stability indicator, can objectively quantify the reliability of the zero reference value.

[0043] (4) Multivariate compensation model, accurately characterizing drift pattern

[0044] The zero-point compensation model established in this invention comprehensively considers three main drift mechanisms: working time (aging), dust accumulation (dust buildup), and polarization duration (charge accumulation). The model structure includes a first-order term, a second-order term, and a decay term, which can more accurately describe the nonlinear evolution law of the zero point.

[0045] (5) Update constraints for working condition perception to prevent learning error signals

[0046] The core innovation of this invention lies in "updating the model only during low-dust windows and freezing parameters during non-low-dust periods." This mechanism fundamentally solves the problem of traditional adaptive methods passively learning effective signals in dusty scenarios, ensuring that the model always learns the true zero-point drift, rather than the measured signal.

[0047] (6) Online assessment of pollution levels to enable predictive maintenance

[0048] This invention quantifies electrode contamination by measuring DC variation and monitors circuit health by analyzing AC noise background, enabling online diagnosis of sensor health. When contamination exceeds a threshold, a maintenance prompt is automatically generated, shifting from "routine maintenance" to "predictive maintenance."

[0049] (7) Full lifecycle management, data traceability

[0050] This invention permanently stores the model parameters, timestamps, confidence levels, and other information from each zeroing calibration, forming a complete lifecycle archive. Health indicators such as zero-point history curves, drift rates, and trigger frequencies provide reliable data for equipment management and metrological traceability.

[0051] (8) Multi-channel support, strong scalability

[0052] This invention supports two modes: multi-channel synchronous zero calibration and polling zero calibration, which can meet the needs of multi-point monitoring.

[0053] (9) Environmental compensation to eliminate the impact of environmental change

[0054] This invention introduces environmental parameters such as temperature, humidity, and pressure for compensation, normalizes the zero reference value under different environmental conditions to a unified benchmark, eliminates the interference of environmental changes on zero-point estimation, and improves the sample consistency of model training.

[0055] (10) Verification mechanism to ensure the validity of zero calibration.

[0056] This invention incorporates a zero-calibration validity verification step, which verifies the correctness of the zero-calibration result by comparing it with multiple reference values. If verification fails, it automatically rolls back to the last valid parameter, preventing erroneous zero-calibration from affecting measurement data. This mechanism provides a final guarantee for the system's robustness. Attached Figure Description

[0057] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below.

[0058] Figure 1 This is an overall flowchart of the online self-calibration zeroing method provided in the embodiments of the present invention;

[0059] Figure 2 This is a detailed flowchart of the triggering and low dust window determination in an embodiment of the present invention;

[0060] Figure 3 This is a schematic diagram illustrating zero-reference establishment and contamination assessment in an embodiment of the present invention;

[0061] Figure 4 This is a flowchart of the coarse zero estimation and robustness screening in an embodiment of the present invention;

[0062] Figure 5 This is a schematic diagram of the structure of the multivariable zero-point compensation model in an embodiment of the present invention;

[0063] Figure 6 This is a flowchart of the adaptive zero-point tracking and constraint update in an embodiment of the present invention;

[0064] Figure 7 This is a schematic diagram illustrating the solidification and lifecycle management in an embodiment of the present invention;

[0065] Figure 8 This is a schematic diagram of environmental parameter compensation in an embodiment of the present invention;

[0066] Figure 9 This is a schematic diagram of multi-channel synchronous zeroing in an embodiment of the present invention;

[0067] Figure 10 This is a structural block diagram of the online self-calibration zeroing system provided in an embodiment of the present invention. Detailed Implementation

[0068] The technical solution of the present invention will now be described in detail with reference to the accompanying drawings and specific embodiments. It should be understood that these embodiments are only for explaining the present invention and are not intended to limit the scope of protection of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0069] Example 1:

[0070] This embodiment provides an online self-calibration method for electrostatic dust concentration meters, such as... Figure 1 As shown, the method includes the following steps:

[0071] S1: Triggering and Low Dust Window Detection

[0072] An electrostatic dust concentration meter was installed at the kiln tail of a cement plant. Its measuring range is 0-50 mg / m³, and its background noise is 0.1 mg / m³. The system continuously collects measurements from the electrostatic sensing channel at a sampling frequency of 100 Hz.

[0073] Set a fixed-length sliding window of 10 seconds (i.e., 1000 sample points). Calculate within each window:

[0074] DC mean: The arithmetic mean of all sample points within the window;

[0075] Standard deviation: σ is the standard deviation of the sample points within the calculation window;

[0076] Rate of change of standard deviation: the percentage change in standard deviation between adjacent windows;

[0077] Amplitude fluctuation: The difference between the maximum and minimum values ​​within the window, divided by the mean.

[0078] Set the following thresholds:

[0079] Standard deviation threshold: σ_th = 0.15 mg / m³ (1.5 times the background noise);

[0080] Threshold for mean change rate: ≤ 0.05 mg / m³ / min (0.1% / min of full scale).

[0081] Permissible range of amplitude fluctuation: ≤ 0.3 mg / m³ (±30% of baseline mean).

[0082] The auxiliary concentration criteria are combined in the following ways:

[0083] Compared with a light scattering dust meter installed in the same location: when the light scattering reading is < 2 mg / m³, it is judged as low dust;

[0084] Based on process parameters: when the kiln tail dust collector is operating normally and the production load is < 30%, it is judged as low dust.

[0085] When all the above conditions are met simultaneously and the duration reaches 30 seconds, it is determined that the effective low dust window has been entered, and a self-calibration zero trigger signal is generated.

[0086] S2: Zero-reference setup and electrostatic sensing channel contamination assessment

[0087] Upon receiving a trigger signal, the built-in relay (double-pole double-throw type) is activated to disconnect the electrostatic sensor electrode from the charge amplifier input terminal and switch the charge amplifier input terminal to ground potential.

[0088] After the relay switches, the output signal of the charge amplifier is acquired:

[0089] DC average measurement: Sampling time 2 seconds, sampling rate 100Hz, 200 sample points are obtained, and the arithmetic mean is calculated as the DC zero reference value V_dc_current = 2.3 mV;

[0090] AC value measurement: Bandpass filtering (1Hz-1kHz) is applied to the same signal segment, and the effective value is calculated as the AC noise floor V_ac_current = 0.8 mV.

[0091] According to the factory records, the initial DC zero reference value V_dc_initial = 1.2 mV, and the DC output range V_dc_range = 5000 mV. The contamination degree of the electrostatic sensing channel is calculated as: P_contamination = 100 × |2.3 - 1.2| / 5000 = 0.022 = 2.2%.

[0092] Pollution thresholds are set as follows: 10% for light pollution, 20% for moderate pollution, and 30% for heavy pollution. The current pollution level of 2.2% is within the normal range.

[0093] The factory default value for AC noise floor is 0.5 mV, and the current value of 0.8 mV is 1.6 times that value, which is less than twice the threshold, indicating that the circuit is in normal health condition.

[0094] S3: Coarse Zero Estimation and Robust Screening

[0095] While keeping the relay in the off state, continue to collect channel output data. Set the sampling rate to 200Hz and the collection time to 5 seconds to obtain 1000 sample points.

[0096] Calculate the sample median as a rough zero estimate: Z0 = median{x1, x2, ..., x1000} = 2.31mV

[0097] Calculate the median absolute deviation: MAD = median{|xi - Z0|} = 0.12 mV

[0098] Set the normal zero-value fluctuation range Z0_range = 1.0 mV (based on historical data statistics), and calculate the relative stability index: S = MAD / Z0_range = 0.12 / 1.0 = 0.12

[0099] Set the stability threshold S_th = 0.10. The current S = 0.12 > 0.10, indicating insufficient stability.

[0100] The sampling time was automatically extended to 10 seconds (twice the original duration), and 2000 sample points were re-acquired. The results were then recalculated: Z0 = 2.32 mV, MAD = 0.09 mV, S = 0.09.

[0101] At this point, S < 0.10, and the reference duration of 10 seconds is greater than the minimum required duration of 5 seconds, so the coarse zero estimate is deemed valid.

[0102] S4: Construction of Multivariable Zero-Point Compensation Model

[0103] The zero-point compensation model is established as follows: Z(t) = a1×t_op + a2×C_acc + a3×T_pol + b1×t_op² + b2×C_acc² + b3×T_pol² + c×exp(-t_op / τ) + Z0

[0104] in:

[0105] t_op: Cumulative working time of the sensor, in hours;

[0106] C_acc: Cumulative dust concentration, obtained by integrating the real-time concentration value over time, in mg·h / m³;

[0107] T_pol: Duration of polarization since the last zeroing, in hours.

[0108] Using 23 successful zero-calibration data points from the past 6 months, the least squares method was used for parameter estimation, yielding: a1 = 0.0021, a2 = 0.0015, a3 = 0.0032; b1 = 0.00003, b2 = 0.00002, b3 = 0.00004; c = 0.85; τ = 720 (hours).

[0109] Validity verification:

[0110] The goodness of fit R² = 0.89 > 0.8, which meets the requirements;

[0111] The maximum relative standard error of the parameters is 22% < 30%, which meets the requirements;

[0112] Residual analysis showed that the residual series had no significant autocorrelation (Durbin-Watson statistic = 2.1).

[0113] S5: Adaptive Zero Tracking and Constraint Update

[0114] In subsequent runs, after each new effective coarse zero estimate Z_measured is obtained, the model parameters are updated using a least mean square adaptive algorithm:

[0115] Z_new = Z_old + μ × (Z_measured - Z_old)

[0116] The coarse zero estimate Z_measured = 2.32 mV, the current model prediction Z_old = 2.28 mV, and the difference Δ = 0.04 mV.

[0117] Based on the robustness index S = 0.09, determine the step size μ by referring to the table:

[0118] 0.05 ≤ S < 0.1 → μ = 0.03

[0119] The updated zero-point value is: Z_new = 2.28 + 0.03 × 0.04 = 2.2812 mV

[0120] Update frequency is constrained by stability rules:

[0121] Number of updates within the same hour: This is the first time (≤3 times, allowed);

[0122] Changes in two consecutive updates: last update change was 0.03 mV, this update change was 0.0012 mV, total change was 0.0312 mV. This is less than 5% of full scale (2.5 mV), which is normal.

[0123] During periods of low dust levels (such as peak production periods or periods of significant fluctuations), model parameters are frozen, and monitoring data is recorded but not used for updates.

[0124] S6: Solidification and Lifecycle Management

[0125] After this self-calibration is completed, the following data will be written to the ferroelectric memory:

[0126] Model parameters: a1=0.0021, a2=0.0015, ... current values;

[0127] Timestamp: 2024-03-15 03:27:45 UTC;

[0128] Reference type: Low dust window passively triggered;

[0129] Confidence score: 92% based on S=0.09 and model fit.

[0130] Update health metrics:

[0131] Zero-point history curve: Recorded zero-point value as 2.2812 mV;

[0132] Drift rate: The average drift rate over the past 30 days was calculated to be 0.12 mV / day;

[0133] Trigger frequency: Successfully triggered 8 times this month;

[0134] Failure rate: 8 out of 9 attempts this month were successful, resulting in a failure rate of 11.1%.

[0135] The current pollution level of 2.2% is far below the moderate pollution threshold, so no maintenance prompt will be generated.

[0136] Example 2:

[0137] This embodiment provides a detailed description of step S1, such as... Figure 2 As shown.

[0138] S11: Sliding window calculation

[0139] Set a fixed-length sliding window L, and adaptively adjust it according to the fluctuation characteristics of the operating conditions:

[0140] Normal operating condition: L = 10 seconds;

[0141] When fluctuations are large: L = 20 seconds (to increase statistical stability);

[0142] Rapidly changing operating conditions: L = 5 seconds (improves response speed).

[0143] Calculate within each window:

[0144] Mean μ = (1 / N)Σxi

[0145] Standard deviation σ = sqrt((1 / N)Σ(xi-μ)²)

[0146] The rate of change of standard deviation Δσ = |σ_current - σ_previous| / σ_previous

[0147] The amplitude fluctuation range R = max(xi) - min(xi)

[0148] S12: Threshold determination

[0149] Set threshold:

[0150] The standard deviation threshold σ_th = k × σ_noise, where σ_noise is the instrument's background noise, and k is taken as 1.5-3.

[0151] Use 1.5 for clean environments and 3 for harsh environments.

[0152] Mean change rate threshold: Change per minute not exceeding 0.5%-2% of full scale.

[0153] For high-precision requirements, use 0.5%; for general monitoring, use 2%.

[0154] Allowable amplitude fluctuation range: R ≤ p × |μ|, where p is between 0.1 and 0.3.

[0155] For stable operating conditions, use 0.1; for fluctuating operating conditions, use 0.3.

[0156] S13: Auxiliary Concentration Criteria

[0157] Use one or more of the following methods in combination:

[0158] Method A: Compare the light scattering method with a real-time light scattering dust meter installed at the same location. If the light scattering method reading is < C_th (e.g., 2 mg / m³) and lasts for T seconds, it is judged as low dust.

[0159] Method B: Historical pattern matching is used to establish a feature library of historical low concentration periods (including time periods, operating parameters, spectral features, etc.), and the similarity between the current period features and the feature library is calculated. When the similarity is > 85%, it is judged as low dust.

[0160] Method C: The process parameter judgment is connected to the DCS system signal. Low dust is judged when the following conditions are met:

[0161] Production equipment shut down;

[0162] Dust collector cleaning cycle;

[0163] Operating at low load (load <30%).

[0164] S14: Comprehensive Judgment

[0165] When all statistical criteria (σ < σ_th, Δμ < Δμ_th, R < R_th) and auxiliary concentration criteria are satisfied simultaneously, and the duration reaches the minimum stable duration T_min (usually 10-30 seconds), it is determined that the effective low dust window has been entered, and a self-calibration zero trigger signal is generated.

[0166] Example 3:

[0167] This embodiment provides a detailed description of step S2, such as... Figure 3 As shown.

[0168] S21: Relay switching

[0169] The relay switching module uses a double-pole double-throw magnetic latching relay, which has the following characteristics:

[0170] Contact resistance: <0.05Ω (measurement condition)

[0171] Isolation resistance: >10 10 Ω (Off state)

[0172] Switching time: <5ms

[0173] Lifespan: >10 7 Second-rate

[0174] Switching timing:

[0175] t0: Trigger signal received

[0176] t1: Relay action (5ms)

[0177] t1-t2: Stable wait (50ms, avoiding switching transients)

[0178] Start at t2: Acquire zero reference signal

[0179] S22: DC Zero Reference Measurement

[0180] The output signal of the charge amplifier is acquired at a sampling rate of 200Hz and a sampling duration of T_dc = 2 seconds, resulting in 400 sample points.

[0181] DC average value calculation: V_dc = (1 / N)Σv_i

[0182] To further improve accuracy, multiple measurements can be performed and the average taken: V_dc_final = (V_dc1 + V_dc2 + V_dc3) / 3

[0183] S23: AC noise background measurement

[0184] Perform digital bandpass filtering (1Hz-1kHz) on the same signal segment to remove DC components and ultra-low frequency interference, and calculate the effective value:

[0185] V_ac_rms = sqrt((1 / N)Σ(v_i - μ)²)

[0186] Alternatively, calculate the peak-to-peak value: V_ac_pp = max(v_i) - min(v_i)

[0187] S24: Pollution Level Assessment

[0188] Contamination level calculation formula: P_contamination = k × |V_dc_current - V_dc_initial| / V_dc_range × 100%

[0189] Where k is the normalization coefficient, usually taken as 100, so that P is expressed as a percentage.

[0190] Tiered early warning:

[0191] Mild pollution: P ≥ 10% and < 20%, log the information, no further warnings will be issued.

[0192] Moderate contamination: P ≥ 20% and < 30%, maintenance reminder generated, planned cleaning recommended.

[0193] Severe pollution: P ≥ 30%, alarm generated, prompting immediate cleaning.

[0194] S25: Circuit Health Assessment

[0195] AC noise floor ratio: R_ac = V_ac_current / V_ac_initial

[0196] Health status assessment:

[0197] R_ac < 2: Normal

[0198] 2 ≤ R_ac < 3: Pay attention to monitoring; it may be due to moisture or component aging.

[0199] 3 ≤ R_ac < 5: Warning, circuit check recommended.

[0200] R_ac ≥ 5: Fault, prompts immediate repair.

[0201] Example 4:

[0202] This embodiment provides a detailed description of step S3, such as... Figure 4 As shown.

[0203] Step S31: Parameter Initialization

[0204] set up:

[0205] Base sampling rate: fs = 200 Hz

[0206] Base data collection duration: T_base = 5 seconds

[0207] Minimum number of valid samples: N_min = 500

[0208] Stability threshold: S_th = 0.10

[0209] Maximum extension factor: K_max = 5

[0210] S32: Data Acquisition

[0211] With the relay off, the fs acquisition channel output is used to obtain N = fs × T_base sample points.

[0212] S33: Median Estimation

[0213] Sort the sample points and take the median: If N is odd, Z0 = x_{(N+1) / 2}; if N is even, Z0 = (x_{N / 2} + x_{N / 2+1}) / 2

[0214] S34: Calculation of robustness indicators

[0215] Calculate the absolute deviation of each sample point from the median: d_i = |x_i - Z0|

[0216] Take the median of these absolute deviations: MAD = median{d_i}

[0217] Calculate the relative stability index: S = MAD / Z0_range

[0218] Z0_range represents the normal zero-value fluctuation range, which can be obtained based on historical data statistics (such as taking the 95th percentile of historical MAD).

[0219] S35: Validity Judgment

[0220] If S < S_th and T ≥ T_min, the condition is valid, and proceed to step 407.

[0221] If S ≥ S_th and the current extension factor < K_max, proceed to step 406.

[0222] If S ≥ S_th and the current extension factor ≥ K_max, the determination fails, and proceed to step 408.

[0223] S36: Adaptive Extended Sampling

[0224] Double the extension factor: K = K × 2 New collection duration: T_new = T_base × K Re-collect data and return to step 403.

[0225] S37: Effective Output

[0226] Record the rough zero estimate Z0, robustness index S, and actual data collection time T, and proceed to the next stage.

[0227] S38: Failure Handling

[0228] Each failure event is recorded, and the failure counter is incremented by 1. If the failure counter exceeds the failure count threshold (e.g., 5 times), a hardware health check is triggered or the frequency of active zeroing is increased.

[0229] Example 5:

[0230] This embodiment provides a detailed description of step S4, such as... Figure 5 As shown.

[0231] S41: Variable Definition and Data Acquisition

[0232] Define three input variables:

[0233] t_op: The sensor's cumulative working time (hours) is accumulated from the first power-on and recorded in non-volatile memory, which is not lost when power is off.

[0234] C_acc: Cumulative dust concentration (mg·h / m³) C_acc(t) = ∫C(τ)dτ, the concentration integral from 0 to t reflects the total amount of dust on the electrode surface, and is strongly correlated with the degree of pollution.

[0235] T_pol: Polarization duration (in hours) since the last zeroing. T_pol = t_current - t_last_calib reflects the time of charge accumulation on the electrode surface.

[0236] S42: Model Structure Design

[0237] The zero-point compensation model takes the following form:

[0238] Z(t) = a1·t_op + a2·C_acc + a3·T_pol + b1·t_op² + b2·C_acc² +b3·T_pol² + c·exp(-t_op / τ) + Z0

[0239] The model consists of three parts:

[0240] The linear component, a1·t_op + a2·C_acc + a3·T_pol, represents uniform drift and is the principal component.

[0241] The quadratic component: b1·t_op² + b2·C_acc² + b3·T_pol² characterizes accelerated drift, reflecting the accelerated aging effect.

[0242] The exponential decay component: c·exp(-t_op / τ) characterizes the initial rapid aging stage, where τ is a time constant.

[0243] S43: Parameter Estimation Methods

[0244] Method A: Least Squares Method (Batch Processing)

[0245] Collect at least 20 sets of historical zero-calibration data {(t_op_i, C_acc_i, T_pol_i, Z_i)}, construct the design matrix X and the observation vector Y, and solve for: β = (X^TX)^{-1} X^TY

[0246] Method B: Recursive Least Squares Method (Online Update) After obtaining new data each time, the parameters are updated recursively: K_k = P_{k-1} x_k / (λ + x_k^T P_{k-1} x_k) β_k = β_{k-1} + K_k (y_k - x_k^T β_{k-1}) P_k = (I - K_k x_k^T) P_{k-1} / λ

[0247] Where λ is the forgetting factor, which is usually taken as 0.95-0.99.

[0248] Method C: Bayesian estimation combines the prior distribution p(β) and the likelihood function p(y|β) to update the posterior distribution: p(β|y) ∝ p(y|β). p(β) takes the posterior expectation as the parameter estimate.

[0249] S44: Validity Verification

[0250] The verification indicators include:

[0251] Goodness-of-fit R²: R² = 1 - SS_res / SS_tot Requirement: R² ≥ 0.8

[0252] Relative standard error of parameters: RSE(β_j) = σ(β_j) / |β_j| × 100% Requirement: RSE ≤ 30%

[0253] Residual autocorrelation test: Calculate the Durbin-Watson statistic: d = Σ(e_t - e_{t-1})² / Σe_t². d close to 2 indicates no autocorrelation, and d should be between 1.5 and 2.5.

[0254] Once all validations pass, the model is valid and ready for application.

[0255] Example 6:

[0256] This embodiment provides a detailed description of step S5, such as... Figure 6 As shown.

[0257] S51: Status Monitoring

[0258] Real-time monitoring of the following statuses:

[0259] Operating conditions: Steady state / Unsteady state

[0260] Low dust window indicator: True / False

[0261] Alarm status: Normal / Alarm

[0262] Environmental parameter changes: rate of change of temperature and humidity

[0263] S52: Update condition judgment

[0264] Updates are only allowed if all of the following conditions are met:

[0265] The low dust window indicator is True.

[0266] Non-alarm state

[0267] The current time is within the preset zero-time period (e.g., 02:00-04:00).

[0268] If the update conditions are not met, proceed to step 605 (parameter freeze).

[0269] S53: Step size adaptive adjustment

[0270] After obtaining the coarse zero estimate Z_measured, the update step size μ is determined based on the robustness index S:

[0271] S range μ range meaning S < 0.01 0.1-0.3 Fast tracking, high confidence 0.01 ≤ S < 0.05 0.05-0.1 Smooth tracking 0.05 ≤ S < 0.1 0.01-0.05 Update with caution S ≥ 0.1 0 No updates

[0272] S54: Parameter Update

[0273] Employing a minimum mean square adaptive algorithm:

[0274] Z_new = Z_old + μ × (Z_measured - Z_old)

[0275] When updating model parameters, keep the model structure unchanged and only adjust the intercept term Z0, or fine-tune all parameters.

[0276] S55: Parameter Freeze

[0277] During periods of low dust levels, the model parameters are completely frozen:

[0278] No updates will be made.

[0279] Only record monitoring data

[0280] It can calculate the theoretical zero point but is not used to correct measured values.

[0281] S56: Update Frequency Constraint

[0282] Apply the following constraints:

[0283] Updates ≤ 3 times within the same hour

[0284] The change in zero point between two consecutive updates is ≤ 5% of the full scale.

[0285] Daily updates ≤ 10 times

[0286] Minimum interval between two updates ≥ 10 minutes

[0287] If the constraint is violated, the update is rejected and the exception is logged.

[0288] Example 7:

[0289] This embodiment provides a detailed description of step S6, such as... Figure 7 As shown.

[0290] S61: Data Solidification

[0291] After each self-calibration to zero, the following data is written to non-volatile storage:

[0292] Data Items Format illustrate Model parameters float[8] a1,a2,a3,b1,b2,b3,c,τ Timestamp uint32 Unix timestamp Reference type uint8 0: Passively triggered, 1: Actively activated Confidence uint8 0-100% Pollution level float P_contamination value robustness indicators float S value Environmental parameters float[3] T,RH,P

[0293] S62: Health Indicator Maintenance

[0294] Maintain time series of the following health indicators:

[0295] The zero-point history curve records the zero-point value Z0 after each zeroing, forming a time series {Z0_t1, Z0_t2, ...,Z0_tn}, which is used to visualize the drift trend.

[0296] Drift rate calculation: Short-term drift rate: ΔZ / Δt (last 24 hours); Long-term drift rate: Slope obtained by linearly fitting data from the last 30 days.

[0297] Trigger frequency statistics: daily trigger count, weekly trigger count, monthly trigger count, and success / failure count.

[0298] Failure rate calculation: Failure rate = Number of failures / Total number of attempts × 100%. An alert is triggered when the failure rate > 20%.

[0299] S63: Anomaly Detection and Maintenance Prompts

[0300] Monitoring the pollution level indicator P_contamination:

[0301] When P ≥ 20%, a maintenance prompt is generated: "Electrode moderately contaminated, cleaning recommended soon."

[0302] When P ≥ 30%, a maintenance prompt is generated: "Electrode heavily contaminated, please clean immediately."

[0303] Monitoring zero-point drift rate:

[0304] When the drift rate > 0.5 mg / m³ / month, the message "Zero-point drift acceleration, inspection recommended" will be displayed.

[0305] Monitoring AC noise background:

[0306] When R_ac > 3, the message "Abnormal circuit noise, it is recommended to check the circuit board" is displayed.

[0307] S64: Adaptive Triggering Strategy

[0308] The self-calibration zero-trigger strategy will be automatically adjusted under the following conditions:

[0309] Scenario A: Multiple consecutive failures, three consecutive self-calibration failures → Relax the low dust window judgment threshold:

[0310] The standard deviation threshold was increased by 20%.

[0311] The allowable range for amplitude fluctuations has been increased by 20%.

[0312] Increase the frequency of proactive zero-testing

[0313] Scenario B: Drift rate too high (drift rate > 1.0 mg / m³ / month) → Increase zero-calibration frequency:

[0314] Increase from once a day to three times a day

[0315] Shorten the active zeroing interval

[0316] Scenario C: Rapid increase in pollution level, weekly pollution rate > 5% → Switch to proactive zero-reference strategy:

[0317] Forced zeroing at a fixed time every day (e.g., 02:00)

[0318] Not limited by low dust conditions

[0319] Example 8:

[0320] This embodiment provides a detailed implementation of S7 environmental parameter compensation, such as... Figure 8 As shown.

[0321] S71: Synchronous acquisition of environmental parameters

[0322] During the self-calibration process, the following environmental parameters are collected simultaneously:

[0323] Temperature T: Accuracy ±0.5℃, range -40-85℃

[0324] Relative humidity (RH): Accuracy ±3%, range 0-100%RH

[0325] Atmospheric pressure P: accuracy ±1 hPa, range 600-1100 hPa

[0326] S72: Establishment of Environmental Compensation Model

[0327] The environmental compensation function f(T,RH,P) was obtained through experimental calibration.

[0328] In a constant temperature chamber, the temperature was changed (from -20℃ to 60℃, in 10℃ increments), and the zero-point change was recorded to obtain the temperature compensation curve f_T(T); in a constant humidity chamber, the humidity was changed (from 20%RH to 90%RH, in 10% increments) to obtain the humidity compensation curve f_RH(RH); in a pressure tank, the pressure was changed (from 800hPa to 1100hPa, in 50hPa increments) to obtain the pressure compensation curve f_P(P).

[0329] Comprehensive compensation function: f(T,RH,P) = f_T(T) × f_RH(RH) × f_P(P)

[0330] S73: Zero Reference Normalization

[0331] Record the environmental parameters (T0, RH0, P0) during this zeroing process, and reduce the measured zero reference value Z_measured to its equivalent value under standard environmental conditions (T_std, RH_std, P_std):

[0332] Z_normalized = Z_measured / f(T0,RH0,P0) × f(T_std,RH_std,P_std)

[0333] Typically, T_std = 25℃, RH_std = 50%, and P_std = 1013hPa are chosen.

[0334] S74: Normalization of Model Training Samples

[0335] When updating model parameters, all historical zero-calibration data are normalized to a unified environmental benchmark using the method described above, eliminating the impact of environmental changes on sample consistency and improving the accuracy of model training.

[0336] Example 9:

[0337] This embodiment provides a detailed implementation of multi-S8 channel synchronous zeroing, such as... Figure 9 As shown.

[0338] S81: Channel Configuration

[0339] Assume the instrument has four electrostatic measurement channels, each corresponding to one of the four measurement points. Each channel is independently configured with a relay switching module.

[0340] S82: Zeroing Mode Selection

[0341] Based on the system configuration, select one of the following modes:

[0342] Mode A: Simultaneous zeroing. All channels are simultaneously disconnected from the acquisition end, and a zero reference point is established at the same time.

[0343] Advantages: Fast zeroing speed, all channels complete simultaneously.

[0344] Disadvantage: No measurement data is available for any channel during the zeroing period.

[0345] Mode B: Polling zeroing performs zeroing on each channel sequentially, while other channels maintain normal measurement.

[0346] Advantages: No monitoring data is lost.

[0347] Disadvantages: The total time for zeroing is relatively long.

[0348] S83: Synchronous Zeroing Implementation

[0349] All channels simultaneously perform the following operations:

[0350] t0: Simultaneously triggers relays on all channels

[0351] t0+5ms: Relay action completed

[0352] t0+55ms: Stable wait ends

[0353] From t0+55ms to t0+5.055s: Simultaneously acquire the zero reference signal of each channel.

[0354] t0+5.055s: Relay reset, measurement resumes.

[0355] S84: Inter-channel consistency check

[0356] Calculate the zero reference values ​​{Z1, Z2, Z3, Z4} for each channel, and calculate the mean μ_z and standard deviation σ_z.

[0357] Inter-channel consistency metric: C_consistency = σ_z / μ_z × 100%

[0358] When C_consistency > 10%, the following issues may be present:

[0359] Inter-channel crosstalk

[0360] Common reference source failure

[0361] Some channel electrodes are severely contaminated.

[0362] S85: Polling Zero-Correction Timing

[0363] Polling zeroing sequence example:

[0364] t0-t5: Channel 1 zeroing, Channels 2 / 3 / 4 measurement.

[0365] t5-t10: Zeroing channel 2, measuring channels 1 / 3 / 4.

[0366] t10-t15: Zeroing channel 3, measuring channels 1 / 2 / 4.

[0367] t15-t20: Zeroing channel 4, measuring channels 1 / 2 / 3.

[0368] Example 10:

[0369] This embodiment provides a self-calibrating zero system for electrostatic dust concentration, such as... Figure 10As shown, it includes the following modules:

[0370] Electrostatic sensor module 1010

[0371] It includes an electrostatic induction electrode 1011 and a charge amplifier 1012. The electrostatic induction electrode adopts a ring electrode structure and is made of stainless steel, while the insulator is made of polytetrafluoroethylene. The charge amplifier adopts a high input impedance (>10Ω) design. 14 An operational amplifier with low bias current (<10fA) and adjustable gain (Ω).

[0372] Relay switching module 1020

[0373] A double-pole double-throw magnetic latching relay is used, connected between the electrostatic sensor electrode 1011 and the charge amplifier 1012. Main parameters:

[0374] Contact resistance: <0.05Ω

[0375] Isolation resistance: >10 10 Ω

[0376] Switching time: <5ms

[0377] Coil power consumption: <200mW

[0378] Lifespan: >10 7 Second-rate

[0379] The signal acquisition and processing module 1030 includes:

[0380] Programmable gain amplifier 1031: Gain 1-128 times, automatic adjustment

[0381] 24-bit Δ-Σ analog-to-digital converter 1032: adjustable sampling rate from 1kHz to 10kHz, signal-to-noise ratio 110dB.

[0382] Digital Filter 1033: Implements functions such as power frequency notch filtering and low-pass filtering.

[0383] The environmental parameter monitoring module 1040 includes:

[0384] Temperature sensor 1041: DS18B20 or PT100, accuracy ±0.5℃

[0385] Humidity sensor 1042: SHT30 or HIH-4000, accuracy ±3%RH

[0386] Pressure sensor 1043: BMP280 or MPX series, accuracy ±1 hPa

[0387] Storage module 1050 includes:

[0388] Ferroelectric memory 1051: MB85RS series, key parameters for frequent writes

[0389] Flash 1052: W25Q series, for historical data and log storage

[0390] The central control module 1060 uses an ARM Cortex-M4 microcontroller with a main frequency of 168MHz and has the following built-in features:

[0391] Real-time Clock 1061: With backup battery to ensure accurate timekeeping.

[0392] Watchdog timer 1062: Prevents program from crashing

[0393] DMA controller: Enables high-speed data transfer

[0394] The central control module is configured to execute the steps described in this invention, including low dust window determination, relay control, data acquisition, robust estimation, model calculation, adaptive tracking, and lifecycle management.

[0395] The output and interaction module 1070 includes:

[0396] OLED Display 1071: Displays real-time zero point, drift trend, and maintenance prompts.

[0397] Indicator light 1072: Power indicator, operation indicator, alarm indicator

[0398] Button 1073: Used for manual triggering of zeroing and parameter setting.

[0399] Communication module 1080 includes:

[0400] RS485 interface 1081: Supports Modbus RTU protocol

[0401] Ethernet interface 1082: Supports Modbus TCP and MQTT

[0402] 4G Module 1083: Used for remote data transmission and alarm push notifications.

[0403] Power module 1090 input 24V DC, output:

[0404] 5V / 3A: Used in digital circuits

[0405] ±12V / 0.5A: For analog circuits

[0406] 3.3V / 1A: For MCU and peripherals

[0407] 5V / 0.2A isolation: for relay driving.

[0408] Although embodiments of the invention have been shown and described, those skilled in the art will understand that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the claims and their equivalents.

Claims

1. An online self-calibration method for electrostatic dust concentration meters, characterized in that, Includes the following steps: S1: Trigger and Low Dust Window Judgment: Continuously collect the measurement values ​​of the electrostatic sensing channel; calculate the DC mean and standard deviation of the output within a fixed-length sliding window, and monitor the rate of change and fluctuation of the standard deviation; when the standard deviation is less than the preset threshold, the rate of change of the mean is lower than the preset threshold, the fluctuation of the amplitude is within the allowable range, and the auxiliary concentration criterion shows that the particle concentration is lower than the preset threshold, the low dust window is determined to be entered, and the self-calibration zeroing process is allowed to start; S2: Zero reference establishment and electrostatic sensing channel contamination assessment: The built-in relay is used to disconnect the electrostatic acquisition terminal, and the current DC average and AC values ​​are measured to establish a new zero reference point; the electrostatic sensing channel contamination is assessed by the DC change, and the electrostatic sensing channel contamination is positively correlated with the DC change. S3: Coarse zero estimation and robustness screening: Acquire channel output for a period of time during the zero reference hold period; use the sample median as the coarse zero estimate, and evaluate stability using the median absolute deviation or an equivalent robustness metric. When the robustness indicator is less than the threshold and the reference duration is not less than the minimum requirement, the coarse zero estimate is deemed valid. Otherwise, the sampling process will be automatically extended or the current self-calibration to zero will be terminated. S4: Construction of a multivariate zero-point compensation model: A zero-point compensation model is established with working time, cumulative dust concentration, and polarization duration since the last zeroing as inputs; the model includes a first-order term, a second-order term, and a time-decaying component, used to characterize the influence of environment and working conditions on the zero point; The initial estimation and validity verification of the model parameters are completed using the zero-reference data obtained in the coarse zero estimation step; S5: Adaptive Zero-Point Tracking and Update Constraints: In subsequent runs, the model parameters are updated with small steps only when the low dust window or active zero reference is met again, so as to realize online tracking of zero-point estimation; the parameters are frozen during non-low dust periods to avoid passively learning effective signals in real dust scenarios; the step size and update frequency are constrained by stability rules. S6: Solidification and Lifecycle Management: Write solidified model parameters, timestamps, reference types, and confidence levels to non-volatile storage; Maintain health indicators such as zero-point history curve, drift rate, trigger frequency, and failure rate; when the contamination level of the electrostatic sensing channel is abnormal, automatically generate maintenance prompts and, if necessary, increase the self-calibration zero trigger frequency or switch to an active zero reference strategy.

2. The online self-calibration method for electrostatic dust concentration meters according to claim 1, characterized in that, The triggering and low-dust window determination step further includes: The fixed-length sliding window is set to 5-30 seconds, and the window length is adaptively adjusted according to the fluctuation characteristics of the operating conditions. The standard deviation threshold is set to 1.5-3 times the instrument's background noise to ensure signal stability within the window. The threshold for the rate of change of the mean is set to a change of no more than 0.5%-2% of the full scale per minute, in order to exclude trend changes; The allowable range for amplitude fluctuation is set to ensure that the peak signal value does not exceed ±10% to ±30% of the baseline mean, in order to eliminate sudden interference. The auxiliary concentration criteria include the following methods: The dust concentration is compared with the value measured by a light scattering dust meter installed at the same location. If the comparison value is lower than a preset concentration threshold, it is determined to be low dust; and / or Matching with historical low-concentration periods, when the similarity between the current period's characteristics and historical low-concentration period characteristics exceeds a similarity threshold, it is determined to be low dust; and / or Based on process parameters, a low dust level is determined when the production equipment is shut down or under low load; and / or When all the above conditions are met simultaneously and the duration reaches the minimum stable duration, it is determined that the effective low dust window has been entered, and a self-calibration zero trigger signal is generated.

3. The online self-calibration method for electrostatic dust concentration meters according to claim 1, characterized in that, The zero-reference setup and electrostatic sensing channel contamination assessment steps further include: The built-in relay is a double-pole double-throw or single-pole double-throw relay, used to disconnect the electrostatic sensor electrode from the charge amplifier input terminal when triggered, and switch the charge amplifier input terminal to ground or a known reference potential; Measuring the current DC mean includes: after the relay switches, acquiring the DC component of the charge amplifier output signal, with a sampling time of 0.5-5 seconds, and calculating the arithmetic mean as the DC zero reference value; Measuring the current AC value includes: after the relay switches, acquiring the AC component of the charge amplifier output signal, calculating the effective value or peak-to-peak value, as the AC noise background; The electrostatic sensing channel contamination assessment formula is: P_contamination = k × |V_dc_current -V_dc_initial| / V_dc_range, where V_dc_current is the current DC zero reference value, V_dc_initial is the initial DC zero reference value after leaving the factory or after the last cleaning, V_dc_range is the DC output range, and k is the normalization coefficient; When the contamination level exceeds the first threshold, it is marked as lightly contaminated; when it exceeds the second threshold, it is marked as moderately contaminated; when it exceeds the third threshold, it is marked as heavily contaminated and a cleaning and maintenance reminder is given. The AC noise floor is used to assess the health of the circuit. When the AC noise floor exceeds 2-5 times the factory value, it indicates that the circuit board is damp or the electronic components are aging; and / or It also includes multi-channel synchronous zeroing: When the instrument has multiple electrostatic measurement channels, each channel is self-calibrated to zero sequentially using either polling or synchronization methods. For synchronous zeroing, all channels simultaneously disconnect the acquisition end, establish their own zero reference points, and calculate the zero-point consistency index between channels. When the zero-point deviation between channels exceeds the consistency threshold, it indicates that there may be crosstalk between channels or a common reference source failure. For polling zeroing, non-zeroing channels continue normal measurement to ensure that no monitoring data is lost during continuous zeroing.

4. The online self-calibration method for electrostatic dust concentration meters according to claim 1, characterized in that, The coarse zero estimation and robustness screening steps further include: During the zero-reference hold period, the channel output is continuously acquired at a sampling rate of 50-500Hz for a duration of no less than 2 seconds, and no less than 100 sample points are obtained. Using the sample median as the rough zero estimate Z0 = median{x1, x2, ..., xn}, the median is not sensitive to outliers, thus improving the robustness of the estimate; The median absolute deviation is used as a robustness indicator: MAD = median{|xi - Z0|}. After normalization, the relative stability indicator S = MAD / Z0_range is obtained, where Z0_range is the normal zero value fluctuation range. When S is less than the stability threshold and the reference duration is not less than the minimum required duration, the coarse zero estimate is deemed valid. If S exceeds the threshold, the sampling time will be automatically extended to 2-5 times the original duration, and the robustness index will be recalculated. If the requirements are still not met after the extension, the current self-calibration zeroing process will be terminated, and a failure event will be recorded. When the cumulative number of failure events reaches the failure count threshold, a hardware health check is triggered or the frequency of active zero-reference triggering is increased.

5. The online self-calibration method for electrostatic dust concentration meters according to claim 1, characterized in that, The steps for constructing the multivariable zero-point compensation model further include: The zero-point compensation model is in the form of a multivariate function: Z(t) = f(t_op, C_acc, T_pol) + Z0, where: t_op represents the cumulative operating time of the sensor, reflecting its long-term aging trend; C_acc is the cumulative dust concentration, which is obtained by integrating the real-time concentration value over time and reflects the electrode dust accumulation effect. T_pol is the duration of polarization since the last zeroing, reflecting the charge accumulation effect on the electrode surface; The model comprises the following components: The linear term: a1 × t_op + a2 × C_acc + a3 × T_pol, represents linear drift; The quadratic term, b1 × t_op² + b2 × C_acc² + b3 × T_pol², characterizes nonlinear acceleration drift. The decay term, c × exp(-t_op / τ), characterizes the exponential decay during the initial rapid aging phase. Model parameter estimation employs at least one of the following methods: Least squares method: Fitting model parameters based on historical zero-correction data; Recursive least squares method: The parameters are updated recursively after each new zero reference data is obtained; Bayesian estimation: updates posterior parameters by combining prior distribution and current observations; The validity verification includes: calculating the model fit R², which should be no less than 0.8; calculating the confidence intervals of the parameters, which should be no more than 30% relative standard error of the parameter estimates; and performing residual analysis, which should be no significant autocorrelation in the residual sequence.

6. The online self-calibration method for electrostatic dust concentration meters according to claim 1, characterized in that, The adaptive zero-point tracking and constraint update step further includes: The least mean square adaptive algorithm takes the following form: Z_new = Z_old + μ × (Z_measured - Z_old) Where Z_old is the zero-point value predicted by the current model, Z_measured is the coarse zero estimate for this time, and μ is the update step size; The step size μ adopts an adaptive adjustment strategy: When S < 0.01, a larger step size μ = 0.1-0.3 is used for fast tracking; When 0.01≤S<0.05, take a medium step size μ=0.05-0.1 for smooth tracking; When 0.05 ≤ S < 0.1, take a small step size μ = 0.01 - 0.05 and update cautiously; When S≥0.1, the step size is set to 0, and no update is performed; The update frequency is constrained by stability rules: the number of updates in the same hour shall not exceed 3; when the zero-point change of two consecutive updates exceeds 5% of the full scale, a verification mechanism is triggered; when the change of environmental parameters exceeds the preset range, a self-calibration zeroing is forcibly started. The non-low dust periods include: periods when signal fluctuations exceed the steady-state threshold, auxiliary concentration criteria show that the concentration is higher than the low dust threshold, or when there is an alarm. During these periods, the model parameters are frozen, and only monitoring data is recorded but not used for updates.

7. The online self-calibration method for electrostatic dust concentration meters according to claim 1, characterized in that, The solidification and lifecycle management steps further include: The non-volatile memory uses ferroelectric memory or flash memory, and the following data is written after each self-zeroing process: Type parameters: the current values ​​of a1, a2, a3, b1, b2, b3, c, τ; Timestamp: UTC time of this calibration; Reference type: Low dust window passively triggered or actively enabled with zero reference; Confidence level: A confidence score of 0-100% calculated based on the robustness index S and model fit. The health indicators to be maintained include: Zero-point history curve: Record the zero-point value after each zeroing and plot the curve of its change over time; Drift rate: Calculates the change in zero point per unit time, with units of mV / day or mg / m³ / month; Trigger frequency: The number of times the self-calibration zeroing is successfully triggered per unit of time; Failure rate: The ratio of zero attempts to successful attempts during self-calibration; When the contamination index P_contamination of the electrostatic sensing channel exceeds the moderate contamination threshold, a maintenance prompt is automatically generated. The prompt includes suggestions to clean the electrodes, check the insulators, and calibrate the circuit. When self-calibration fails multiple times or the zero drift rate exceeds the preset threshold, the self-calibration trigger frequency is automatically increased, and the low dust window judgment threshold is relaxed by 10%-30% to increase the triggering opportunity. If effective zero calibration still cannot be obtained, the active zero reference strategy is switched to, that is, the acquisition channel is forcibly cut off for zero calibration at a preset time, which is not limited by low dust conditions.

8. The online self-calibration method for electrostatic dust concentration meters according to claim 1, characterized in that, It also includes step S7: Environmental parameter compensation. Simultaneously collect temperature, relative humidity, and atmospheric pressure parameters, and establish a compensation model for the influence of environmental parameters on the zero point: Z_compensated = Z_model × f(T, RH, P) + Z0Moses model Where f(T, RH, P) is the environmental compensation function, which is obtained through experimental calibration; During the self-calibration process, the temperature T0, humidity RH0, and pressure P0 at that time are recorded, and the zero reference value is reduced to the equivalent value under standard environmental conditions to eliminate the influence of environmental changes on zero point estimation. When updating model parameters, the zero reference values ​​obtained under different environmental conditions are normalized to a unified benchmark to improve the consistency of model training samples.

9. An online self-calibration zeroing system for electrostatic dust concentration meters, characterized in that, include: The electrostatic sensor module includes electrostatic induction electrodes and a charge amplifier, used to generate an electrical signal related to dust concentration. Relay switching module: Connected between the electrostatic sensor electrode and the charge amplifier, it is used to cut off the electrode signal and establish a zero reference state when triggered; Signal acquisition and processing module: Connected to the output of the charge amplifier, used to acquire signals and perform analog-to-digital conversion and filtering; Environmental parameter monitoring module: includes at least one of temperature sensor, humidity sensor, and pressure sensor, used to synchronously collect environmental parameters; Storage module: Used to store model parameters, historical zeroing data, health indicators, and lifecycle archives; Central control module: connected to the relay switching module, signal acquisition and processing module, environmental parameter monitoring module and storage module respectively, the central control module is configured to execute the method steps of any one of claims 1-10; Output and Interaction Module: Connected to the central control module, used to output zero-point status, drift trend, maintenance prompts, and zeroing results; Communication module: Used for data interaction with host computer or cloud platform, receiving remote zeroing commands, and uploading zeroing records.

10. The online self-calibration zeroing system for electrostatic dust concentration meters according to claim 9, characterized in that: The relay switching module uses a magnetic latching relay or a solid-state relay with low and high isolation to achieve complete isolation in the zero-calibration state. The signal acquisition and processing module includes: Programmable gain amplifier, used to automatically adjust the amplification factor according to the signal strength; 24-bit high-precision analog-to-digital converter with a sampling rate of no less than 1kHz; Digital filters are used to filter out power frequency interference and high-frequency noise; The central control module uses an ARM Cortex-M series microcontroller or digital signal processor, and has a built-in real-time clock and watchdog timer.

Images