A marine battery thermal runaway early warning method based on independent double channels

Abstract

The embodiment of the application provides a kind of marine battery thermal runaway early warning method based on independent double channel, it is applied to ship battery safety technical field, utilize inherent swing frequency in ship navigation as initiative excitation signal, respectively from the electrochemical impedance channel between the outer wall of seawater cooling pipeline and seawater obtains pipe wall following degree, and from the contact resistance channel of battery pole post connection row obtains contact following degree;The difference of pipe wall following degree and contact following degree is obtained following degree deviation, and the historical sliding average of following degree deviation under normal navigation state is obtained as deviation baseline, and then the relative deviation amount is calculated;When relative deviation amount exceeds preset threshold, output thermal runaway early warning signal.The invention converts ship swing interference into excitation source, through double-channel independent physical quantity mutual check, can identify thermal runaway precursor in advance when conventional parameters are not exceeded, has strong anti-interference ability, low false alarm rate, applicable to battery management system under marine environment.

Description

Technical Field

[0001] This application relates to the field of marine battery safety technology, and in particular to a method for early warning of thermal runaway of marine batteries based on independent dual channels. Background Technology

[0002] Current marine battery thermal runaway warning systems mostly rely on temperature or voltage thresholds, triggering an alarm when these parameters exceed set values. However, the temperature rise before thermal runaway is delayed, often only triggering an alarm in the irreversible stage, failing to provide sufficient emergency response time for crew members, resulting in a severely inadequate early warning lead time.

[0003] Shipboard operating environments are subject to interference such as strong salt spray corrosion, wide-band vibration, and drastic temperature fluctuations. Single sensors, such as gas sensors or temperature sensors, are highly prone to generating false alarms. Frequent false alarms can reduce crew trust in the early warning system and even cause genuine thermal runaway to be overlooked.

[0004] Some solutions employ a dual-sensor redundancy design to improve reliability, but typically use similar sensors to measure the same physical quantity. This redundancy cannot withstand common-cause failures; when environmental factors, such as humidity or electromagnetic interference, simultaneously affect both sensors, the redundancy design fails, and true independent calibration cannot be provided. Summary of the Invention

[0005] The embodiments of this application provide a method for early warning of thermal runaway of marine batteries based on independent dual channels. This method transforms ship rolling into an active excitation source and detects the deviation in following the trajectory of the thermal runaway through two independent physical channels: the impedance of the seawater cooling pipe wall and the contact resistance of the electrode post. It can provide early warning of thermal runaway before conventional parameters exceed limits, and is adaptive to sea state changes and resistant to interference. To achieve the above objectives, this application adopts the following technical solution: A method for early warning of thermal runaway in marine batteries based on independent dual channels, the method comprising: Read raw triaxial acceleration data from the ship's inertial measurement unit and generate a raw acceleration array; The original acceleration array is subjected to DC removal processing to generate a zero-mean acceleration array; Perform a Fast Fourier Transform on the zero-mean acceleration array to generate a complex acceleration array; Calculate the amplitude at each frequency point of the complex acceleration array to generate an acceleration amplitude array; Sort the acceleration amplitude array in descending order to generate a descending amplitude array; The frequency corresponding to the first amplitude in the descending amplitude array is taken as the ship's rolling frequency signal; A unit amplitude sine wave is generated using the ship's sway frequency signal as the frequency, and used as the excitation signal; The excitation signal is input into a second-order bandpass filter, and the center frequency of the bandpass filter is set to the frequency of the excitation signal to generate a filtered waveform. Perform a Hilbert transform on the filtered waveform to generate an analytic signal; Calculate the magnitude sequence of the analytic signal to generate the envelope sequence; Divide each sampling point of the filtered waveform by the value at the corresponding position in the envelope sequence to generate a normalized waveform, which serves as the swing reference waveform. The resistance between the outer wall of the seawater cooling pipe and the seawater is obtained, and a pipe wall resistance sequence is generated. The pipe wall resistance sequence is synchronously detected with the swing reference waveform to obtain the pipe wall frequency fluctuation amplitude and generate the pipe wall following degree. Obtain the contact resistance value of the battery terminal connector and generate a contact resistance sequence; The contact resistance sequence is synchronously detected with the swing reference waveform to obtain the contact frequency fluctuation amplitude and generate the contact following degree. Calculate the difference between the pipe wall following degree and the contact following degree to generate the following degree deviation, obtain the historical sliding average value of the following degree deviation under normal ship navigation conditions, and generate the deviation baseline; The difference between the following deviation and the deviation baseline is calculated to generate a relative deviation. When the relative deviation is greater than a preset threshold, a thermal runaway warning signal is output.

[0006] In some possible implementations, obtaining the resistance between the outer wall of the seawater cooling pipe and the seawater, and generating a pipe wall resistance sequence, includes: A constant alternating current of a preset frequency is injected between the outer wall of the seawater cooling pipe and the seawater to generate a test alternating current; the alternating voltage response between the outer wall and the seawater is collected to generate a response alternating voltage. Calculate the ratio of the root mean square value of the response AC voltage to the root mean square value of the test AC current, and generate the complex impedance modulus value as the single tube wall resistance value. The injection, acquisition, and calculation actions are repeated at preset sampling time intervals, and the single pipe wall resistance values ​​obtained each time are arranged in chronological order to generate a pipe wall resistance value sequence.

[0007] In some possible implementations, obtaining the contact resistance values ​​of the battery terminal connectors and generating a contact resistance sequence includes: A constant DC current of a preset amplitude is applied to both ends of the battery terminal connector to generate a DC excitation current. The DC voltage drop across the connector is measured to generate a voltage drop value. Calculate the ratio of the voltage drop value to the amplitude of the DC excitation current to generate the instantaneous value of the contact resistance; The actions of applying, measuring, and calculating are repeated every preset sampling period. The instantaneous values ​​of contact resistance obtained each time are arranged in chronological order to generate a contact resistance sequence.

[0008] In some possible implementations, obtaining the historical sliding average of the follow-through deviation under normal navigation conditions and generating a deviation baseline includes: Read a preset number of historical deviation values ​​from the memory, arrange them from oldest to newest, and generate a historical deviation list; Calculate the sum of all values ​​in the historical deviation list to generate the total deviation; The sum of the deviations is divided by the preset number to obtain the arithmetic mean, which is used as the deviation baseline.

[0009] In some possible implementations, the process of removing DC from the original acceleration array to generate a zero-mean acceleration array includes: Calculate the average value of all values ​​in the original acceleration array to generate the array average value; Iterate through each value in the original acceleration array, subtract the average value of the array from each value, and generate a difference array; The difference array is output in its original order as the zero-mean acceleration array.

[0010] In some possible implementations, performing a fast Fourier transform on the zero-mean acceleration array to generate a complex array of accelerations includes: Obtain the length of the zero-mean acceleration array and generate the number of points N; Create a complex array of length twice N, and generate an empty complex array; The values ​​of the zero-mean acceleration array are sequentially assigned to the first N positions of the empty complex number array to generate the front-fill array; The zero values ​​are assigned to the last N positions of the front-filled array to generate the array to be transformed; Perform a Fast Fourier Transform operation on the array to be transformed to generate a complex array of accelerations.

[0011] In some possible implementations, calculating the amplitude at each frequency point of the complex acceleration array to generate the acceleration amplitude array includes: Extract the real part value of each frequency point from the complex acceleration array to generate a real part list; Extract the imaginary part value of each frequency point from the complex acceleration array to generate an imaginary part list; Square each value in the real part list to generate a real part square list; Square each value in the list of imaginary parts to generate a list of squared imaginary parts. Add the values ​​at corresponding positions in the list of squares of the real part to the list of squares of the imaginary part to generate a list of sums of squares; Take the square root of each value in the sum of squares list to obtain the amplitude list, which serves as the acceleration amplitude array.

[0012] In some possible implementations, the calculation of the sum of all values ​​in the historical deviation list to generate the deviation sum includes: Read the total deviation calculated from the previous sliding window and generate the previous total deviation; extract the value of the head position from the historical deviation list and generate the head deviation value. Calculate the difference between the previous total deviation and the head-of-line deviation value to generate a total without the head-of-line value; obtain the newly calculated following deviation value at the current moment to generate the current deviation value; The sum of the values ​​after removing the first value is added to the sum of the current deviation values ​​to generate the updated total deviation.

[0013] As can be seen from the above technical solution, this application has the following beneficial effects: 1. This application significantly improves the lead time and anti-interference capability of thermal runaway early warning. Unlike traditional threshold methods, this invention detects the early mechanical and electrochemical effects of electrolyte leakage and shell bulging, which appear tens of minutes or even hours before a sharp temperature rise. Utilizing the inherent rolling motion of the ship as an active excitation source, weak signals are extracted through synchronous detection, transforming the ship's rolling motion from an interference source into a signal source, effectively filtering out broadband noise such as random wave impacts and main engine vibrations. Even under severe sea conditions, it can stably extract the same-frequency fluctuation components in pipe wall resistance and contact resistance, achieving early and reliable early warning.

[0014] 2. This application significantly reduces the false alarm rate and simplifies engineering applications through an independent dual-channel design and an adaptive baseline mechanism. The two channels are physically independent: the first channel is based on electrochemical impedance, and the second channel is based on contact resistance. The two measurement circuits have independent power supplies and signal transmissions, avoiding common-cause failures. An alarm is triggered only when the tracking deviations of both channels simultaneously deviate from the historical baseline, greatly improving alarm confidence. Simultaneously, the historical moving average baseline can adaptively track changes in sea state, battery aging, and ambient temperature fluctuations, eliminating the need for manual setting of absolute thresholds or periodic calibration, significantly reducing maintenance costs and making it suitable for long-term unattended ship applications. Attached Figure Description

[0015] The invention will now be further described with reference to the accompanying drawings.

[0016] Figure 1 A flowchart for acquiring ship roll frequency signals and generating roll reference waveforms provided in this application embodiment; Figure 2 A flowchart for the synchronous detection and following accuracy calculation of the first channel tube wall resistance provided in this application embodiment; Figure 3 A flowchart for calculating the synchronous detection and tracking accuracy of the second-channel contact resistance provided in this application embodiment; Figure 4 The flowchart for baseline update of follow-up deviation and decision-making for thermal runaway early warning provided in the embodiments of this application is shown. Detailed Implementation

[0017] The terms "first," "second," and "third," etc., used in this application specification, claims, and drawings are for distinguishing different objects, not for specifying a particular order.

[0018] In the embodiments of this application, the words "exemplary" or "for example" are used to indicate examples, illustrations, or descriptions. Any embodiment or design described as "exemplary" or "for example" in the embodiments of this application should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of the words "exemplary" or "for example" is intended to present the relevant concepts in a specific manner.

[0019] Research has revealed that existing technologies, which rely on temperature or voltage thresholds for judgment, exhibit a lag in temperature rise before thermal runaway, often triggering alarms only in the irreversible stage, resulting in severely insufficient early warning. Furthermore, single sensors are susceptible to interference from salt spray, vibration, and other factors, leading to false alarms, while dual-sensor redundancy cannot withstand common-cause failures, resulting in poor reliability.

[0020] To address the aforementioned issues, this application provides a method for early warning of thermal runaway in marine batteries based on independent dual channels: Example 1: To solve the above problems, as follows... Figures 1-4 As shown, the technical problem to be solved by the present invention is: how to issue an early and reliable warning signal in the early stage of thermal runaway and before the conventional parameters exceed the limit, and avoid false alarms caused by marine environmental interference.

[0021] To solve this technical problem, the present invention adopts a method for early warning of thermal runaway of marine batteries based on independent dual channels. The core idea is to take the rolling motion of the ship under the action of waves as a known and continuous external excitation source, measure the degree of response of the two different physical links of the battery system to this excitation, and construct a joint early warning index that is strongly correlated with the precursors of thermal runaway.

[0022] Specifically, the first channel measures the electrochemical impedance between the outer wall of the seawater cooling pipe and the seawater. When a small amount of electrolyte leaks from the battery, the leaked organic vapor combines with condensation or salt mist outside the cooling pipe wall to form a thin liquid film with altered conductivity. This causes a change in the impedance value at the pipe wall-seawater interface, and this impedance value fluctuates with the seawater flow caused by the ship's rolling motion. The second channel measures the contact resistance of the battery terminal connectors. When gas is generated inside the battery, causing the casing to bulge, the terminals experience additional axial force, changing the tightness of the connectors and causing fluctuations in the contact resistance value. These fluctuations are also modulated by the inertial force caused by the ship's rolling motion.

[0023] Because the two channels have different physical mechanisms and are both subjected to the same swing excitation, under normal conditions, the response tracking degree of the two channels, i.e., the degree of synchronization between the response signal and the excitation signal, has a stable difference. When thermal runaway precursors occur, electrolyte leakage and shell bulging often occur sequentially or simultaneously, causing asynchronous changes in the tracking degree of the two channels, resulting in a significant deviation of the tracking degree from its historical normal baseline. By calculating the relative deviation of the tracking degree deviation from the historical moving average baseline and comparing it with a preset threshold, a warning signal can be output before the parameters significantly exceed the limits.

[0024] In the above method, the two channels are completely independent: the signal of the first channel originates from the electrochemical measurement circuit outside the seawater cooling pipe, and the signal of the second channel originates from the DC resistance measurement circuit of the electrode connector. The two measurement circuits are powered and transmitted independently, avoiding common-cause faults. At the same time, this method transforms ship rolling, which is usually considered a disturbance, into an active excitation source, making full use of the unique physical conditions of the marine environment and possessing technical advantages that land-based battery early warning solutions cannot achieve. Detailed Implementation

[0026] I. Overall Process Overview

[0027] This embodiment uses an electric vessel equipped with a seawater-cooled battery pack as an example. During normal navigation, the vessel generates continuous low-frequency rolling motions, with the dominant rolling frequency typically ranging from 0.1Hz to 0.5Hz, depending on the vessel type, loading status, and sea state. The battery management system has a built-in early warning controller, which is connected to the vessel's inertial measurement unit, pipe wall resistance measurement circuit, and contact resistance measurement circuit, and executes an early warning method according to the following steps.

[0028] The early warning controller acquires the ship's rolling frequency signal and generates a sine wave of unit amplitude as an excitation signal. This excitation signal is not used to drive any actuators, but rather serves as a reference for subsequent coherent detection.

[0029] The dominant yaw frequency component is extracted from the excitation signal to obtain a normalized yaw reference waveform. This waveform has the same frequency and phase as the actual ship yaw acceleration, but its amplitude is normalized to 1.

[0030] Two independent detection channels are activated. Channel 1: Measure the resistance between the outer wall of the seawater cooling pipe and the seawater, forming a pipe wall resistance sequence. This sequence is synchronously detected with a swing reference waveform to extract the wave component with the same frequency as the swing, and the amplitude of this component is calculated, defined as the pipe wall following degree. Channel 2: Measure the contact resistance of the battery terminal connector, forming a contact resistance sequence. This sequence is synchronously detected with a swing reference waveform to extract the amplitude of the wave component with the same frequency, defined as the contact following degree.

[0031] The difference between the pipe wall following degree and the contact following degree is calculated to obtain the following degree deviation. Under normal ship navigation and battery health conditions, this deviation value fluctuates within a small range around a stable mean. The early warning controller stores historical following degree deviation data under normal conditions and calculates its moving average as the deviation baseline.

[0032] Calculate the difference between the current following deviation and the deviation baseline, take the absolute value, and divide it by the deviation baseline (plus a minimum divide-by-zero constant) to obtain the relative deviation. When the relative deviation exceeds a preset threshold, it is determined that the battery has a risk of thermal runaway, and a warning signal is output.

[0033] 2. Obtain the ship's rolling frequency signal.

[0034] 2.1 Hardware configuration and data acquisition.

[0035] The ship's inertial measurement unit (IMU) is typically installed near the ship's center of gravity, outputting raw acceleration data along three orthogonal axes. The early warning controller reads the vertical acceleration data from the IMU via the controller area network (CLAN) bus or Ethernet, as the vertical roll component best reflects the ship's undulation motion. The sampling rate is set to a preset sampling frequency, which is significantly higher than the ship's maximum possible roll frequency; for example, the preset sampling frequency is set to more than 10 times the highest roll frequency to satisfy the Nyquist sampling theorem and avoid spectral aliasing.

[0036] 2.2 Generate the original acceleration array.

[0037] Each time the early warning controller reads an acceleration data point, it stores it in a fixed-length first-in-first-out queue. The queue length needs to cover at least dozens of oscillation cycles to ensure that the frequency resolution of the subsequent Fourier transform is sufficiently fine. For example, when the ship's lowest oscillation frequency is 0.1Hz (10-second period), the queue length can be 50 to 100 seconds of data points. This queue is the original acceleration array, where each element is a real value.

[0038] 2.3 DC removal process.

[0039] Because inertial measurement units may experience zero-point drift, and the ship may be in a tilting state, the raw acceleration array includes a DC component. This DC component interferes with subsequent spectrum analysis and therefore must be eliminated.

[0040] Action sequence: The first step is to calculate the arithmetic mean of all values ​​in the original acceleration array to obtain the array average. This average represents the DC bias of the acceleration signal.

[0041] The second step is to iterate through each value in the original acceleration array, subtract the array average from each value, and obtain the difference.

[0042] The third step is to arrange all the differences in their original order to form a zero-mean acceleration array. The values ​​in this array fluctuate around zero, eliminating the effects of sensor bias and static gravity.

[0043] 2.4 Perform a Fast Fourier Transform.

[0044] The zero-mean acceleration array is a long array of zero-mean accelerations. A sequence of real numbers representing points. To obtain its spectrum, the early warning controller performs a Fast Fourier Transform. This invention employs zero-padding technology to improve frequency resolution, specifically as follows.

[0045] Action sequence: The first step is to obtain the length of the zero-mean acceleration array, denoted as the number of points. .

[0046] The second step is to create a file with a length of... The complex array is used as an empty complex array. This length is chosen to facilitate the use of the radix-2 Fast Fourier Transform algorithm, which requires a length that is a power of 2. Padding with zeros makes the length […]. And can be designed The power of 2, along with zero padding, can improve the apparent resolution of the spectrum without increasing the actual data.

[0047] The third step is to assign the values ​​of the zero-mean acceleration array to the first element of the empty complex number array in sequence. Each position is assigned a value, which represents the real part of the complex number, while the imaginary part is set to zero. After the assignment is complete, the front-filled array is obtained.

[0048] The fourth step is to assign the zero value to the last element of the preceding fill array. The position, i.e., the index arrive The real and imaginary parts are both zero, resulting in the array to be transformed.

[0049] Fifth step: Perform a Fast Fourier Transform operation on the array to be transformed, and output a value of the same length. The complex array of accelerations. Each element in the array represents a frequency component, with the real part being the amplitude of the cosine component and the imaginary part being the amplitude of the sine component.

[0050] Technical explanation: Zero padding does not increase the true frequency resolution, but it makes the spectrum curve smoother, facilitating peak location. Since the frequency corresponding to the peak needs to be extracted later, zero padding is beneficial.

[0051] 2.5 Calculate the amplitude array.

[0052] The complex array of accelerations includes complex values ​​at each frequency point, which need to be converted into an amplitude spectrum, i.e., the energy amplitude at each frequency point.

[0053] Action sequence: The first step is to extract the real part value of each frequency point from the complex acceleration array, arrange them in frequency order, and obtain a list of real parts.

[0054] The second step is to extract the imaginary part value of each frequency point from the complex acceleration array to obtain the imaginary part list.

[0055] The third step is to square each value in the list of real parts to obtain a list of squared real parts.

[0056] The fourth step is to square each value in the list of imaginary parts to obtain a list of squared imaginary parts.

[0057] Fifth, add the values ​​at corresponding positions in the list of squares of the real part to the list of squares of the imaginary part to obtain the list of sums of squares.

[0058] The sixth step is to take the square root of each value in the sum of squares list to obtain the amplitude list, which serves as the acceleration amplitude array. Each element in this array is a non-negative real number, representing the intensity of the vibrational energy at that frequency point.

[0059] Special Note: Because the input is a sequence of real numbers, the complex array of accelerations output by the Fast Fourier Transform (FFT) exhibits conjugate symmetry, meaning the latter half is symmetrical to the first half and does not include additional information. Therefore, in practical processing, the warning controller can only use the first half of the array (index 0 to...). The corresponding frequency points are used to save computation. Index 0 corresponds to the DC component, and index... Corresponding to the Nyquist frequency.

[0060] 2.6 Extract the ship's swaying frequency.

[0061] The energy of ship rolling is mainly concentrated in the low-frequency range, usually showing a significant peak in the acceleration amplitude array. The frequency corresponding to this peak is the dominant rolling frequency of the ship. To reliably extract this frequency, the early warning controller uses a sorting method.

[0062] Action sequence: The first step is to copy the acceleration amplitude array to a temporary array, and then sort the temporary array in descending order to obtain a descending amplitude array. The first element of the descending amplitude array is the global maximum value, denoted as . .

[0063] The second step is to iterate through the original acceleration amplitude array and find the values ​​that match the input values. The element with the smallest absolute difference is recorded at its index, denoted as the peak index. The allowable error threshold is set to... One-thousandth of the value is used to avoid floating-point errors causing lookup failures.

[0064] The third step is to use the peak index. Sampling frequency of the inertial measurement unit and number of transformation points Calculate the actual frequency value: ; This frequency value This refers to the ship's rolling frequency signal.

[0065] The fourth step is to generate a unit amplitude sine wave at this frequency, expressed as follows: ,in Time is the time. This sine wave is denoted as the excitation signal. Unit amplitude refers to a peak value of 1; using unit amplitude allows the subsequent tracking accuracy to have standardized dimensions.

[0066] 3. Extract the swing reference waveform.

[0067] Although the excitation signal is a standard sine wave, the actual ship rolling waveform may not be an ideal sine wave and may contain harmonic components. To obtain a purer fundamental reference, the excitation signal needs to be narrowband filtered and amplitude normalized. In this embodiment, the generated excitation signal is an ideal sine wave, which is used as input to subsequent bandpass filtering, Hilbert transform, and normalization steps. However, for generality, filtering and normalization are still performed in these steps to ensure a stable reference waveform can be obtained even if the excitation signal has amplitude fluctuations.

[0068] Action sequence: The first step is to input the excitation signal into a second-order bandpass filter. The filter's center frequency is dynamically set to the frequency of the excitation signal, with a high quality factor, such as 10, resulting in a very narrow passband that only allows the swing frequency component to pass through, filtering out any possible second harmonics and noise. The output is the filtered waveform.

[0069] The second step is to perform a Hilbert transform on the filtered waveform to obtain the analytic signal. The analytic signal is a complex sequence, with its real part being the original filtered waveform and its imaginary part being the result of its Hilbert transform.

[0070] The third step is to calculate the magnitude of the analytic signal at each sampling point, and then take the square root of the sum of the squares of the real and imaginary parts to obtain the envelope sequence. The envelope sequence reflects the instantaneous amplitude of the filtered waveform at each moment.

[0071] The fourth step is to divide each sample point of the filtered waveform by the value at the corresponding position in the envelope sequence. This division is performed point by point, ensuring that the amplitude of each point in the resulting new sequence falls within the range of... The waveform is between [a certain value] and has the same shape as the filtered waveform, but its amplitude is calibrated to a constant value of 1. This normalized waveform is the swing reference waveform. It only includes the phase and frequency information of the swing's dominant frequency and is unaffected by the wave size.

[0072] Additional explanation: In actual ship environments, even if an ideal sine wave is generated, the resulting sine wave may have slight amplitude fluctuations due to errors in the digital oscillator or frequency quantization. Therefore, the above normalization step is still of practical significance.

[0073] IV. First Channel: Pipe Wall Resistance Sequence and Pipe Wall Follow-up Degree.

[0074] 4.1 Introduction to Measurement Principle

[0075] Seawater cooling pipes are typically made of copper-nickel alloys or titanium alloys, and a natural electrochemical interface exists between the outer wall of the pipe and the seawater. This interface exhibits complex impedance characteristics, and its modulus is affected by factors such as the composition, temperature, and flow rate of the electrolyte near the interface. When electrolyte leakage occurs, the leaked organic solvent mixes with condensate or salt spray on the outer surface of the pipe wall, forming a thin liquid film. The conductivity of this liquid film is much higher than that of pure condensation, resulting in a significant decrease in the impedance of the pipe wall-seawater interface. Simultaneously, the oscillation and flow of seawater outside the pipe caused by the ship's rolling motion causes periodic changes in the thickness of the liquid film, thus modulating the impedance value. Therefore, at the dominant frequency of the rolling motion, the pipe wall resistance exhibits a fluctuation component with the same frequency as the rolling motion. The amplitude of this fluctuation component reflects the degree of leakage.

[0076] 4.2 Inject AC current.

[0077] Action sequence: The first step is to generate a sine wave digital signal with a preset frequency, such as 1kHz, which is then converted into an analog voltage by a digital-to-analog converter. This voltage is then converted into an AC current with a constant amplitude by a voltage-to-current conversion circuit. This current is injected between the outer wall of the pipe and the seawater to generate a test AC current.

[0078] 4.3 Acquire response voltage.

[0079] Action sequence: The first step is to acquire the voltage signal between the outer wall and the seawater through a differential amplifier, and then convert it into a digital signal by an analog-to-digital converter after anti-aliasing filtering, so as to obtain a digital sequence of the AC voltage response.

[0080] 4.4 Calculate the single resistance value.

[0081] Action sequence: The first step is to extract a complete cycle from the digital sequence of the response AC voltage, for example, by pre-setting an integer number of sampling points. Calculate the square of the value at each sampling point, sum the results, divide by the number of sampling points, and then take the square root to obtain the root mean square value of the response voltage. .

[0082] The second step is to read the known root mean square value of the test AC current. (Preset constant).

[0083] The third step is to calculate. The complex impedance modulus is obtained and used as the single tube wall resistance.

[0084] 4.5 Generate the pipe wall resistance sequence.

[0085] Action sequence: The first step is to repeat the above injection, acquisition and calculation actions every preset sampling time interval, such as 0.1 seconds, and generate a new single tube wall resistance value each time.

[0086] The second step involves appending each obtained single tube wall resistance value to the end of the array in chronological order. If the array length exceeds a preset maximum length, the old value at the beginning of the array is removed to maintain a fixed queue length. This ultimately yields the tube wall resistance value sequence. .

[0087] 4.6 Synchronous detection yields the pipe wall following accuracy.

[0088] The principle of synchronous detection (coherent demodulation) is as follows: the signal to be detected is multiplied by a reference signal of the same frequency, and then low-pass filtered. The DC component is proportional to the amplitude of the component in the signal to be detected that has the same frequency as the reference signal. In this embodiment, the reference signal is a swing reference waveform. .

[0089] Mathematical derivation: Let the signal to be detected be The reference signal is Multiplying them together gives: ; After low-pass filtering, high-frequency components Noise is filtered out, leaving only the DC component. When the reference signal and the signal to be detected are in phase At that time, the DC component is multiplied by That is, the amplitude is obtained .

[0090] Action sequence: The first step is to sequence the pipe wall resistance values. Time alignment with the swing reference waveform is performed. If the sampling rates of the two waveforms differ, a linear interpolation method is used. Resample to the same sampling rate as the swing reference waveform.

[0091] The second step is to multiply the two aligned sequences point by point to obtain the product sequence. .

[0092] The third step is to input the product sequence into a first-order low-pass digital filter. The filter recursive formula is: ; in The input is a product sequence. For filtered output, This is the smoothing coefficient. (Take...) When the sampling interval is 0.1 seconds, the time constant is It can effectively filter out 2 times and high-frequency noise within seconds. The filtered output DC component .

[0093] Fourth step: Multiply the DC component by 2 to obtain the pipe wall following degree: ; Definition: Pipe wall following degree It is a dimensionless real number representing the intensity of the fluctuation of the pipe wall resistance in sync with the ship's rolling motion.

[0094] V. Second Channel: Contact Resistance Sequence and Contact Follow-up Degree.

[0095] 5.1 Introduction to Measurement Principle

[0096] The battery terminals are fastened to the connector pins with bolts, and the tightening force determines the contact resistance. Under normal conditions, the contact resistance is very small (in the milliohm range) and relatively stable. When gas is generated inside the battery, causing the casing to bulge, the terminals are pushed upwards, changing the preload of the fastening nuts and thus altering the contact resistance. The inertial force caused by the ship's rolling motion causes a slight displacement of the connector pins, further modulating the contact resistance. Therefore, the contact resistance also includes a fluctuation component with the same frequency as the rolling motion. The amplitude of this fluctuation component reflects the degree of casing bulging.

[0097] 5.2 Apply DC current.

[0098] Action sequence: The first step is to apply a constant DC current of a preset amplitude (e.g., 10A) to both ends of the battery terminal connector through a constant current source circuit to generate a DC excitation current. Simultaneously, the actual value of this current is measured (which can be deduced by inversely calculating the voltage drop across the precision sampling resistor) and stored.

[0099] 5.3 Measure the voltage drop.

[0100] Action sequence: The first step is to acquire the DC voltage drop across the connector using a differential amplifier, and then convert the voltage drop value to an analog-to-digital converter. .

[0101] 5.4 Calculate the instantaneous value of contact resistance.

[0102] Action sequence: The first step is to calculate the instantaneous value of the contact resistance according to Ohm's law: ; 5.5 Generate the contact resistance sequence.

[0103] Action sequence: The first step is to repeat the application, measurement, and calculation process every preset sampling period, such as 0.2 seconds, to generate a new instantaneous value of the contact resistance each time.

[0104] The second step is to append the instantaneous contact resistance values ​​obtained each time to the end of the array in chronological order, keeping the queue length fixed, to obtain the contact resistance sequence. .

[0105] 5.6 Contact tracking accuracy is obtained through synchronous detection.

[0106] Action sequence: The first step is to sequence the contact resistance. Align with the swing reference waveform (using the same interpolation method), multiply point by point to obtain the product sequence. .

[0107] The second step is to input the product sequence into the same first-order low-pass filter as the first channel. The same applies, yielding the DC component. .

[0108] The third step is to multiply the DC component by 2 to obtain the contact tracking accuracy: ; Definition: Contact Follow-through It is a dimensionless real number representing the degree to which the contact resistance fluctuates synchronously with the ship's rolling motion.

[0109] VI. Calculate the following deviation and the deviation baseline.

[0110] 6.1 Calculate the following deviation.

[0111] Action sequence: The first step is to calculate the difference between the pipe wall following degree and the contact following degree: ; 6.2 Obtain the historical moving average.

[0112] The early warning controller needs to store historical data on follow-altitude deviations during normal navigation. Normal navigation is determined by conditions such as speed exceeding a preset threshold, battery charging / discharging current within the normal range, and battery temperature within the normal range. The preset sliding window length is [value missing]. ( It should cover at least 20 swing cycles. For example, with a sampling period of 0.1 seconds and a swing cycle of 5 seconds, it can be taken as... This corresponds to 50 seconds covering 10 cycles; if 20 cycles are needed, then... ).

[0113] Action sequence: The first step is to read the most recently stored data from memory. Each historical deviation value is sorted by time from oldest to newest to generate a historical deviation list. The list is empty when the system first runs; no warning is issued at this time, and a "System learning incomplete" signal is output. The list will be updated once the historical deviation values ​​have accumulated. After obtaining the data, proceed with the subsequent steps.

[0114] The second step is to calculate the sum of all values ​​in the historical deviation list to obtain the total deviation. The first calculation requires traversing the entire list.

[0115] The third step is to divide the total deviation by The arithmetic mean is obtained and used as the baseline for deviation: ; 6.3 Efficient update of sliding window summation.

[0116] To reduce computational load, a recursive approach is used to update the total deviation.

[0117] Action sequence: The first step is to read the total deviation calculated in the previous sliding window and record it as the previous total deviation. .

[0118] The second step is to extract the value of the head position from the historical deviation list, which is the oldest data point, and record it as the head deviation value. .

[0119] The third step is to calculate the sum of previous deviations minus the deviation value of the head of the queue, thus obtaining the sum after removing the value of the head of the queue: .

[0120] The fourth step is to obtain the newly calculated following deviation value at the current moment and record it as the current deviation value. .

[0121] Fifth step: Calculate the sum of values ​​after removing the first value of the queue, add the current deviation value, and obtain the updated total deviation: .

[0122] Step 6: Divide the updated total deviation by A new deviation baseline is obtained. .

[0123] Step 7: Delete the first element of the historical deviation list and append the current deviation value as the new last element to form an updated list for use in the next iteration.

[0124] Definition: The first deviation value refers to the value stored first in the historical deviation list in chronological order.

[0125] 7. Calculate the relative deviation and output an early warning.

[0126] 7.1 Calculate the relative deviation.

[0127] Action sequence: The first step is to calculate the following deviation minus the deviation baseline, obtaining the signed difference: .

[0128] The second step is to take the absolute value of this difference to obtain the absolute deviation: .

[0129] The third step is to divide the absolute deviation by the deviation baseline (plus a minimum divide-by-zero constant). ), to obtain the relative deviation: ; in Take 1% of the typical value of the deviation baseline or a fixed value of 0.001 to avoid When the value approaches zero, the division overflows.

[0130] 7.2 Threshold comparison.

[0131] Action sequence: The first step is to compare the relative deviations. With preset threshold Threshold Statistical analysis based on thermal runaway simulation experiments determined a typical value of 1.5–2.0, indicating a relative deviation exceeding the normal average by 50%–100%. If... If so, it is determined that the warning conditions are met.

[0132] 7.3 Output early warning signals.

[0133] Action sequence: First step, when At that time, the warning sign location will be set to true.

[0134] The second step is to send an early warning message via the digital output port or communication bus, outputting a thermal runaway warning signal. If only one channel malfunctions while the other channel is normal, a lower-level maintenance reminder can be output without triggering a thermal runaway warning.

[0135] Complete application example: An electric vessel is sailing normally at 12 knots in calm waters with good sea conditions, experiencing stable, low-frequency rolling motions. The early warning controller in the battery management system has been initialized and is establishing and operating a thermal runaway early warning system as follows.

[0136] The following key parameters are pre-set in the early warning controller, and their selection is based on the ship's typical rolling characteristics and hardware measurement capabilities: The sampling frequency of the inertial measurement unit is set to 50 Hz, which is 100 times the ship's highest rolling frequency (about 0.5 Hz) and far exceeds the Nyquist requirement, ensuring the accuracy of spectrum analysis.

[0137] The sampling interval for the first channel (pipe wall resistance) is 0.1 seconds, which means 10 pipe wall resistance points are collected per second.

[0138] The sampling interval for the second channel (contact resistance) is 0.2 seconds, which means that 5 contact resistance points are sampled per second.

[0139] The update cycle for the follow-up deviation is the slower sampling interval between the two channels, that is, a new follow-up deviation is calculated every 0.2 seconds to ensure data synchronization.

[0140] The sliding window length is set to 1000 follow-degree deviation points. Since it is updated every 0.2 seconds, the window corresponds to a time length of 200 seconds, which is sufficient to cover about 40 swing cycles, if the swing cycle is 5 seconds.

[0141] The low-pass filter used in synchronous detection is a first-order infinite impulse response filter, and its recursive formula is: ,in When the input sampling interval is 0.1 seconds, the filter time constant is approximately 10 seconds; the sampling interval for the product sequence of the second channel is 0.2 seconds, but to maintain the same filtering characteristics, its smoothing coefficient should be adjusted to... (because (seconds). This example is for simplification; all times are used consistently. It accepts a slightly different time constant for the second channel, which can be calculated separately in actual engineering.

[0142] Zero constant in relative deviation calculation Taking 0.001, which is approximately 3% to 10% of the normal deviation baseline (0.01 to 0.03), can effectively prevent division overflow.

[0143] Warning threshold The value is set to 1.5. This value was obtained through statistical analysis of thermal runaway simulation tests on batteries of the same model. The average relative deviation under normal conditions is 0.105, and the standard deviation is 0.02. The mean is taken and three times the standard deviation is added to get 0.165. In order to fully ensure that there are no false alarms and leave a margin, the preset threshold is set to 1.5.

[0144] Monitoring process during normal navigation.

[0145] First, the early warning controller reads the most recent 500 vertical acceleration data points from the inertial measurement unit, covering a 10-second time window. Due to potential ship tilt and sensor zero-point drift, the raw data contains a DC component. The controller calculates the average of these 500 data points and then subtracts this average from each data point to obtain a zero-mean acceleration sequence. Next, the controller zero-paddings the sequence to 1000 points, which improves the apparent resolution of the spectrum and facilitates subsequent peak localization. A Fast Fourier Transform is performed on the zero-padding sequence to obtain the complex acceleration spectrum, and then the amplitude at each frequency point is calculated (the square root of the sum of the squared real and imaginary parts) to obtain the amplitude spectrum.

[0146] A noticeable peak appears in the low-frequency range of the amplitude spectrum. The controller sorts the amplitude array in descending order, takes the first element as the global peak, and then looks up the frequency index corresponding to the peak in the original array. In this example, the sampling frequency is 50 Hz, the number of transformation points is 1000, and the frequency resolution is 0.05 Hz. The peak appears at index 4, corresponding to the frequency... Hertz. Therefore, the controller generates a unit amplitude sine wave at a frequency of 0.2 Hz as the excitation signal.

[0147] To obtain a clean and amplitude-stable reference waveform, the excitation signal is fed into a second-order bandpass filter. The filter's center frequency is dynamically set to 0.2 Hz, with a quality factor of 10, resulting in a bandwidth of 0.02 Hz, effectively filtering out harmonics and noise. The filtered waveform undergoes a Hilbert transform to obtain an analytic signal; the magnitude of the analytic signal is the waveform's envelope. Dividing the original waveform point-by-point by the envelope sequence yields a normalized waveform with a constant amplitude of 1, i.e., the sway reference waveform. This waveform is in phase and frequency with the actual sway of the ship but is unaffected by changes in wave height.

[0148] Simultaneously, the first detection channel begins operation. The controller injects a constant alternating current with a frequency of 1 kHz and an amplitude of 100 microamps between the outer wall of the seawater cooling pipe and the seawater. The injection frequency is much higher than the oscillation frequency, which avoids the 50 Hz power frequency interference of the ship's electrical system. At the same time, AC injection can avoid polarization effects. The controller collects the response voltage between the outer wall of the pipe and the seawater. By calculating the ratio of the root mean square of the response voltage to the root mean square of the injected current, the complex impedance modulus of the pipe wall-seawater interface is obtained, i.e., the single-cycle pipe wall resistance. The injection, collection, and calculation are repeated every 0.1 seconds, and the pipe wall resistance values ​​of the most recent 200 seconds are continuously recorded to form a pipe wall resistance sequence.

[0149] Subsequently, the controller synchronously detects the tube wall resistance sequence and the swing reference waveform. Since the sampling rates of the two sequences may differ (the swing reference waveform is typically sampled at 50 Hz), the tube wall resistance sequence is first resampled to 50 Hz using linear interpolation to align their times. Then, they are multiplied point-by-point to obtain the product sequence. The product sequence contains a DC component, a second harmonic component, and noise. The product sequence is then input into a first-order low-pass filter, with the recursive formula as follows: ,in It is a product sequence. This is the filtered DC component. Multiplying the filtered DC component by 2 gives the pipe wall following accuracy. In mathematics, ,in This is a sequence of pipe wall resistance values. This is the reference waveform for the oscillation. Under normal navigation conditions, the pipe wall following accuracy remains stable at around 0.05, indicating that there is no electrolyte leakage causing enhanced impedance modulation.

[0150] The second detection channel operates simultaneously. The controller applies a constant DC current of 10 amps to both ends of the battery terminal connector; this current value will not cause a significant temperature rise in the connector (power approximately 0.1 watts). The DC voltage drop across the connector is measured, and the instantaneous contact resistance is calculated using Ohm's law. Data is collected every 0.2 seconds, and the most recent 200 seconds of data are stored to form a contact resistance sequence. This sequence is then subjected to the same synchronous detection operation as the swing reference waveform to obtain the contact tracking accuracy. Under normal conditions, the contact follow-through is stable at around 0.03, indicating that the electrode connection is tight and the contact resistance is minimally affected by the swaying inertial force.

[0151] The controller calculates the following deviation. At this time, To obtain a normal baseline that updates slowly over time, the controller maintains a sliding window of length 1000 to store historical follow-altitude deviation values. The window slides every 0.2 seconds, removing the oldest value and adding the latest one each time. The sum of all values ​​in the window divided by 1000 gives the baseline deviation. When the system starts up for the first time, the controller first collects the first 1000 data points. The data, summed and averaged, were used as the initial baseline, during which no warnings were issued. After a sufficiently long period of smooth sailing, the baseline stabilized around 0.018.

[0152] The controller then calculates the relative deviation. ,in To prevent a baseline that is too small from causing division by zero, a constant is used. Substitute the value: denominator The value is much lower than the warning threshold of 1.5, so the system determines that the battery is safe and does not trigger an alarm.

[0153] Changes after electrolyte microleakage occurs.

[0154] Approximately two hours into the voyage, a minor electrolyte leak began to occur in one of the battery cells due to a manufacturing defect. The leaked organic vapor diffused to the outer wall of the seawater cooling pipes, mixing with salt mist condensation on the pipe surface to form a thin liquid film with high conductivity. This film altered the electrochemical properties of the pipe wall-seawater interface, making the pipe wall resistance extremely sensitive to seawater flow caused by the ship's rolling motion. A distinct fluctuation component, synchronized with the rolling frequency of 0.2 Hz, began to appear in the pipe wall resistance sequence.

[0155] The controller detects the pipe wall following degree in real time. It gradually increases, slowly rising from 0.05 to 0.12. The contact following accuracy of the second channel. It remains around 0.03 because the battery casing has not yet bulged significantly, and the terminal connection is unaffected. Therefore, the following deviation is... .

[0156] Meanwhile, the sliding window is continuously updated. As a new deviation value of 0.09 enters the window, the oldest normal deviation value is removed, and the baseline slowly rises from 0.018 to 0.023. Because the leak occurred relatively recently, most of the historical data within the window remains normal, so the baseline increase is limited. The controller recalculates the relative deviation versus the absolute deviation. denominator , This value has far exceeded the preset threshold of 1.5.

[0157] The controller immediately detected a risk of thermal runaway in the battery and sent a warning message to the bridge via the digital output port, while simultaneously activating the audible and visual alarms. The crew then inspected the relevant battery module and found that the cell casing had slightly discolored, and the surface temperature was only 2 degrees Celsius higher than the ambient temperature, far below the traditional temperature alarm threshold. However, because the warning controller had issued a signal in advance, the crew was able to isolate the battery cluster from the system in time and initiate cabin ventilation and fire prevention procedures. Subsequently, the cell continued to heat up slowly while isolated, but thermal spread did not occur, preventing a serious accident.

[0158] Adaptability under different sea conditions.

[0159] If a ship encounters large waves, the dominant sway frequency may become 0.3 Hz, and the sway amplitude will also increase significantly. At this time, the pipe wall following degree and contact following degree under normal conditions will be slightly higher than in calm sea conditions, for example, becoming 0.07 and 0.05 respectively, but the difference between the two remains around 0.02. The sliding window automatically incorporates new normal fluctuations into the history, and the baseline is gradually updated to around 0.02. Therefore, the relative deviation will not falsely trigger a warning due to changes in sea state. Only when electrolyte leakage or casing bulging causes asynchronous changes in the following degree of the two channels will the deviation significantly deviate from the baseline. As can be seen from the above examples, this invention can reliably identify precursors of thermal runaway in the early stages before the battery's conventional parameters exceed their limits, utilizing the inherent swaying excitation of the ship and the difference in response between the two channels, providing sufficient response time for ship safety.

[0160] The foregoing has shown and described the basic principles, main features, and advantages of this application. Those skilled in the art should understand that this application is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of this application. Various changes and modifications can be made to this application without departing from the spirit and scope thereof, and all such changes and modifications fall within the scope of this application as claimed. The scope of protection of this application is defined by the appended claims and their equivalents.

Claims

1. A method for early warning of thermal runaway in marine batteries based on independent dual channels, characterized in that, The method includes: Read raw triaxial acceleration data from the ship's inertial measurement unit and generate a raw acceleration array; The original acceleration array is subjected to DC removal processing to generate a zero-mean acceleration array; Perform a Fast Fourier Transform on the zero-mean acceleration array to generate a complex acceleration array; Calculate the amplitude at each frequency point of the complex acceleration array to generate an acceleration amplitude array; Sort the acceleration amplitude array in descending order to generate a descending amplitude array; The frequency corresponding to the first amplitude in the descending amplitude array is taken as the ship's rolling frequency signal; A unit amplitude sine wave is generated using the ship's sway frequency signal as the frequency, and used as the excitation signal; The excitation signal is input into a second-order bandpass filter, and the center frequency of the bandpass filter is set to the frequency of the excitation signal to generate a filtered waveform. Perform a Hilbert transform on the filtered waveform to generate an analytic signal; Calculate the magnitude sequence of the analytic signal to generate the envelope sequence; Divide each sampling point of the filtered waveform by the value at the corresponding position in the envelope sequence to generate a normalized waveform, which serves as the swing reference waveform. The resistance between the outer wall of the seawater cooling pipe and the seawater is obtained, and a pipe wall resistance sequence is generated. The pipe wall resistance sequence is synchronously detected with the swing reference waveform to obtain the pipe wall frequency fluctuation amplitude and generate the pipe wall following degree. Obtain the contact resistance value of the battery terminal connector and generate a contact resistance sequence; The contact resistance sequence is synchronously detected with the swing reference waveform to obtain the contact frequency fluctuation amplitude and generate the contact following degree. Calculate the difference between the pipe wall following degree and the contact following degree to generate the following degree deviation, obtain the historical sliding average value of the following degree deviation under normal ship navigation conditions, and generate the deviation baseline; The difference between the following deviation and the deviation baseline is calculated to generate a relative deviation. When the relative deviation is greater than a preset threshold, a thermal runaway warning signal is output.

2. The method according to claim 1, characterized in that, The step of obtaining the resistance between the outer wall of the seawater cooling pipe and the seawater, and generating a pipe wall resistance sequence, includes: A constant alternating current of a preset frequency is injected between the outer wall of the seawater cooling pipe and the seawater to generate a test alternating current; the alternating voltage response between the outer wall and the seawater is collected to generate a response alternating voltage. Calculate the ratio of the root mean square value of the response AC voltage to the root mean square value of the test AC current, and generate the complex impedance modulus value as the single tube wall resistance value. The injection, acquisition, and calculation actions are repeated at preset sampling time intervals, and the single pipe wall resistance values ​​obtained each time are arranged in chronological order to generate a pipe wall resistance value sequence.

3. The method according to claim 1, characterized in that, The step of obtaining the contact resistance value of the battery terminal connector and generating a contact resistance sequence includes: A constant DC current of a preset amplitude is applied to both ends of the battery terminal connector to generate a DC excitation current. The DC voltage drop across the connector is measured to generate a voltage drop value. Calculate the ratio of the voltage drop value to the amplitude of the DC excitation current to generate the instantaneous value of the contact resistance; The actions of applying, measuring, and calculating are repeated every preset sampling period. The instantaneous values ​​of contact resistance obtained each time are arranged in chronological order to generate a contact resistance sequence.

4. The method according to claim 1, characterized in that, The process of obtaining the historical sliding average value of the follow-through deviation under normal navigation conditions and generating a deviation baseline includes: Read a preset number of historical deviation values ​​from the memory, arrange them from oldest to newest, and generate a historical deviation list; Calculate the sum of all values ​​in the historical deviation list to generate the total deviation; The sum of the deviations is divided by the preset number to obtain the arithmetic mean, which is used as the deviation baseline.

5. The method according to claim 1, characterized in that, The process of removing DC from the original acceleration array to generate a zero-mean acceleration array includes: Calculate the average value of all values ​​in the original acceleration array to generate the array average value; Iterate through each value in the original acceleration array, subtract the average value of the array from each value, and generate a difference array; The difference array is output in its original order as the zero-mean acceleration array.

6. The method according to claim 1, characterized in that, The step of performing a Fast Fourier Transform on the zero-mean acceleration array to generate a complex acceleration array includes: Obtain the length of the zero-mean acceleration array and generate the number of points N; Create a complex array of length twice N, and generate an empty complex array; The values ​​of the zero-mean acceleration array are sequentially assigned to the first N positions of the empty complex number array to generate the front-fill array; The zero values ​​are assigned to the last N positions of the front-filled array to generate the array to be transformed; Perform a Fast Fourier Transform operation on the array to be transformed to generate a complex array of accelerations.

7. The method according to claim 1, characterized in that, The calculation of the amplitude at each frequency point of the complex acceleration array to generate the acceleration amplitude array includes: Extract the real part value of each frequency point from the complex acceleration array to generate a real part list; Extract the imaginary part value of each frequency point from the complex acceleration array to generate an imaginary part list; Square each value in the real part list to generate a real part square list; Square each value in the list of imaginary parts to generate a list of squared imaginary parts. Add the values ​​at corresponding positions in the list of squares of the real part to the list of squares of the imaginary part to generate a list of sums of squares; Take the square root of each value in the sum of squares list to obtain the amplitude list, which serves as the acceleration amplitude array.

8. The method according to claim 4, characterized in that, The calculation of the sum of all values ​​in the historical deviation list to generate the deviation sum includes: Read the total deviation calculated from the previous sliding window and generate the previous total deviation; extract the value of the head position from the historical deviation list and generate the head deviation value. Calculate the difference between the previous total deviation and the head-of-line deviation value to generate a total without the head-of-line value; obtain the newly calculated following deviation value at the current moment to generate the current deviation value; The sum of the values ​​after removing the first value is added to the sum of the current deviation values ​​to generate the updated total deviation.

Images