MXPA99011957A - Battery testing and classification - Google Patents

Abstract

Method and apparatus for battery evaluation and classification applies transient microcharge and/or microload pulses to an automotive battery. Classification is made on the basis of analysis of the resultant voltage profile or portions or dimensions thereof. In one embodiment the analysis utilises a neural network or algorithm to assess a microcycle sequence of microload/microcharge tests utilising one of a series of battery parameters including impedance as well as voltage characteristics to effect classification. Another embodiment adopts an optimised (not maximum) level of prior test-based data-training for a self-organising neural network. A third embodiment utilises prior test data correlation to enalbe alorithm-based classification without use of a neural network.

Description

TESTING AND CLASSIFICATION OF BATTERIES This invention relates to a method and apparatus for testing and classifying car batteries and others. In the automotive industry as in other technical areas, there is a need for improvements in systems for battery tests. We have provided a breakthrough in the art in terms of the test system described in EP 0 762 135 A2 (case 12, reference P52759EP). Describing a method and apparatus for testing automotive electronic control units and batteries using neural networks to perform waveform analysis on a digitized signal. The battery test is by waveform analysis of the battery current during transient connection of a load. A network learning stage that employs along with a routine of recognition tests for characteristic waveforms. This approach is based on program simulation or software software of the waveform analysis, which can be carried out visually by a person skilled in the art to distinguish between current or voltage profiles of various battery categories. . References identified in searches made with respect to the subject of the present application, consist of the following: US 4204153 US 5469528 US 5537327 WO 96/35522 WO 96/05508 GB 2285317A GB 2278452A None of these is more relevant than W096 / 35522 which describes the use of a capacitor discharge pulse as a battery test stage, to allow identification of the battery type by reference to a voltage gradient using a voltmeter or oscilloscope or digital signal processing device, thereby Battery can be recognized before sorting discarded batteries for recycling purposes. FIRST ASPECT In a first aspect of the present invention we are seeking to provide a major technical advance in battery testing systems, with which the approach of simulating the visual analysis carried out by a technician, is replaced by an alternative approach in the that the technical attributes of computer systems are used, in a way that exploits its inherent advantages instead of seeking and restricting them to simulate a human analytical approach.

According to this, we have sought to use the capacity of software systems to process and analyze data concerning multiple parameters, in a way that the human brain finds difficult. In accordance with this approach, we realize that the multi-layered PERCEPTRON neural networks of our previously identified EP '135 A2 specification have limitations in terms of their requirement to feed an approximation to the analytical response required in any case. In accordance with this new approach to battery analysis, we seek to provide a system in which multiple battery parameters, not all of them necessarily, are electrical parameters such as thermal parameters, would be fed into the system and the system would be able to recognize particular characteristics of complex data fed and thus perform an efficient classification stage, probably on the basis of relatively limited data compared to those required for waveform analysis. An objective of this aspect of the present invention is to provide a method and apparatus for automotive and other battery classification, according to one or more parameters detected, offering improvements in relation to one or more of the matters discussed above or in general. According to the invention, there is provided a method for producing a battery testing apparatus and also an apparatus for classifying automotive and other batteries, as defined in the accompanying claims. In one embodiment of the invention, the system is provided with a neural network in the form of a self-organized network called a Kohonen network. In the modality, this software network has been submitted (or is based on software that has been submitted) to a controlled degree of training. This training is based on a data feed that includes a representative sample of battery data. This representative sample of battery data comprises a plurality of battery parameters selected from the group including voltage, current, internal resistance, surface load / capacitance and thermal parameters. This controlled degree of self-training provides a basis for the classification stage required for the self-organized network. In the embodiment as well, the method and apparatus provide means of generating test data as part of the system and which are adapted to subject a given battery to test to test routines to generate test data for the classification means. This test data is related to the representative sample battery data in which the software network (software) was trained. For example, the software network can be trained in data from a sequence of transient battery charges or transient battery charging routines, these being equally separated by relatively transient intervals. When this routine has been included in the training data, then the means of generating test data included in the battery test apparatus are arranged to generate corresponding data which leads to relatively efficient battery classification. In connection with the use of multiple transient battery charges in the battery test routine, this approach has practical significance in relation to the capacitance aspects of battery construction, ie the sequence of multiple transient loads is capable of discharging the battery. Surface area of the battery that is present due to the construction of parallel plate of the latter and which can otherwise provide false data qua, arise (for example) the potential difference of 15 volts arising from ripple of the alternator.

The embodiment of the invention is more intense in software (software) than intense in physical equipment when compared with previously known battery systems. SECOND ASPECT According to a second aspect of the invention, there is a need in the automotive and other industries for improved methods and apparatus for the testing and classification of batteries, notably the supply in systems that allow on-site testing of batteries that use a minimum of physical equipment which make low energy requirements, while offering portable operational characteristics and with relatively rapid classification / test results, and an object of this aspect of the present invention is to provide a method and apparatus that offers improvements in relation to one or more of these requirements, or no doubt in general. According to the invention, there is provided a method and apparatus for testing and / or classification of automotive and / or other batteries, as defined in the accompanying claims. In an embodiment of the invention described below, there is provided a method and apparatus in which a battery is subjected to a microcycle test procedure, the result of which is subjected to analysis by a neural network or by an adaptive algorithm and Through this new approach to the test of automotive batteries and others, we have established that it is possible to test and classify batteries according to their useful life and other criteria, in a way that produces advantages according to several of the factors discussed above. In the modality, the data analysis stage is carried out by means of an adaptive algorithm or by a neural network and the battery test procedure includes the application of transient stimuli as part of a microcycle. The physical equipment provided for this purpose is minimized by this design approach and the test connection to the battery allows on-site testing, without violating automotive radio code systems or software (software) of the electronic control unit and similar. The battery test routine, including the micro cycle that is provided for the application of a series of transient stimuli to the battery in the manner shown in the accompanying drawings for example: Voltage test; Micro-charge or energy to be consumed (microload); Surface load removal; Micro-charge (microload); Recovery; Micro-charge (microload); Microcharge or stored energy (microcharge). Within a microcycle of the above type, there may be a series of positive and negative pulses applied to the battery, these are implemented by a gate of Field Effect Transistor (FET = Field Effect Transistor). By the use of a microcycle of charges (charges) and / or transient loads applied to a battery in sequence according to this aspect of the invention, the advantage is provided (with respect to our specification EP '135A previously mentioned ( that the practical and effective cost of the system, particularly for use in a portable application, is unexpectedly advanced in such a proportion as a relatively sophisticated analysis of a battery in a relevant one of the multiplicity of categories by use of an easily operated equipment item , becomes a realistic option in a proportion that has not been the cto date.The number of pulses that are provided in a micro cycle is greater than one slightly at any point in the range of two pulses to 100 or more. voltages in the sequential pulses can be incre or decre or be equal.The number or frequency and voltage levels of the pulses can be determined by u An initial voltage test applied to the battery and / or ambient temperature and / or battery size, and then the pulse characteristics can be determined by data values obtained from subsequent pulses. In other words, the algorithm that determines pulse numbers, frequency and levels, is adaptive. Alternatively, prior knowledge regarding the battery can be used, at least in part to establish certain parameters of the pulse characteristics, these depend on data such as battery history, known behavior, etc. In the embodiments of the invention wherein the method and apparatus provide test and / or battery classification by use of a micro cycle comprising two or more charges and / or transient loads applied to a battery in sequence, The use of this test routine is novel to the best of the knowledge of the applicants in relation to test procedures for automobiles and the like or other battery systems. Conventional testing procedures to date have involved procedures that may be termed micro cycles, which necessarily (and often by intention) involve the use of substantial charge flows and energy release associated with corresponding associated technical disadvantages, including the provision of a thermal collector or similar means and service limitations that restrict the test to battery evaluation not in situ. The described modalities employ an analysis stage for battery data generated from the micro cycle and the modalities employ a neural network or an adaptive algorithm to allow classification or numerical evaluation of a battery. The use of a neural network leads to the data processing advantages established in our previous application of July 19, 1997. In the embodiments, Figures 11 and 12 of the drawings show battery responses to microcharging and microcharging cycles. (microload) for both "good" and "bad" batteries. Data used in the analysis stage of the invention are used directly from the microcycle routine. The data can be in waveform format or in discrete data point format, and can represent relative or absolute values of the characteristics of sampled batteries. The latter may include impedance, rebound, peaks, areas, charge reception, start and end voltages, etc. THIRD ASPECT According to a third aspect of the present invention, there is provided a system wherein the design approach of the first and second aspects of the invention are modified in order to achieve a substantial simplification in terms of analytical procedures and equipment supply. physical / software (software), while still based on the same analytical principle and thus can provide a comparable result in terms of battery classification in the main classes. In simple terms, this aspect of the invention provides a method and apparatus that are capable of achieving accurate battery classification for example in the bad / good classes, although in the downloaded / good classes but with the structural and procedural simplification above mentioned with the corresponding accompanying cost advantages. In accordance with this aspect of the invention, we have discovered that in the method and apparatus of the preceding aspect of the invention, the above-described simplifications can be achieved by eliminating the use of neural networks for analysis of battery parameter profiles for of classification. This simplification of corresponding progress was a relatively unexpected development in the course of our research in this project, despite the background experience with our previous work that involves providing the use of neural networks. However, it was precisely the knowledge of the results of the test procedures involving the neural networks that provided the data with which simplified methods and devices could be adopted. More specifically, knowledge of the results of test procedures with neural networks allows us to identify the fact that a satisfactory co-relationship could be achieved between battery ratings and test procedure profiles, using a relatively simple algorithmic procedure. This procedure uses data previously established with respect to the microload voltage profiles (microcharge) and / or microload (microload), in order to define time intervals (of a defined time such as the microcharge start time or microcharge) (microload) (in which the characteristic sample voltages are taken which in this way become substantially as informative as the analysis of the integral profile itself.) Accordingly, and in accordance with this third aspect of the present invention, they provide a method and apparatus as defined in the ones identified in the accompanying claims.

In the embodiment described below, the reaction of the battery to an applied transient test parameter provides a measure of the condition of the battery. By analyzing the profile or structure of the waveform (by reference to the corresponding voltage at a characteristic point) representative of the reaction of the battery to the test parameter, the corresponding classification step can be carried out. This analysis stage is carried out in the modality without using a neural network by means of an algorithm adapted to identify a corresponding one of two or more battery condition classifications, and the corresponding one of these classifications exhibits a co-relation with (voltage aspects of) the waveform profile of the electrical value. In the modality, voltage aspects of the waveform profile used for this purpose include a voltage value giving an impedance measurement, voltage bounce, voltage peaks, voltage and voltage channels, voltage profile or shape, load recovery and start and end voltages. Also in the modality, the transient period of the micro-charge stored (microload) or micro-charge to be consumed (microcharge) applied to the battery, is in the range of 1 to 1000 milliseconds, and the repetition interval between microloans to be consumed (microload) and / or successive microcharge (microcharge) varies from 2 to 100 times the duration of the application of the micro-load to be consumed (microload) or micro-charge stored (microcharge). In this aspect of the invention, the ability to eliminate the neural network with its accompanying cost despite the retention of an analysis basis (with respect to the test parameter profile) corresponding to that carried out by the neural network, is Significant value and benefit.

The achievement of this stage by a routine based on relatively simple algorithmic physical equipment, allows to achieve a simplification of valuable product. Modes directly relating to a second aspect of the invention will now be described by way of example with reference to the accompanying drawings in which: Figure 1 shows a series of test procedures forming a complete test cycle; Figure 2 shows the voltage profiles plotted against time for a microload test sequence to be consumed (microload); Figure 3 shows the corresponding profile for a stored microload test sequence (microcharge); Figure 4 shows a micro-load circuit diagram to be consumed for application of transient load to a battery and test; Figure 5 shows the corresponding stored micro-charge circuit diagram; Figure 6 shows the microloan circuit to be consumed together with the associated data acquisition and control systems; Figure 7 shows the stored micro-charge circuit corresponding to Figure 6; Figure 8 shows a block circuit diagram for a detailed battery tester installation that includes microcharging functions to be consumed (microload) and microcharging (microcharge), and a corresponding micro control system and an output signal or an output to an exhibitor; Figure 9 shows a trace for the results of a complete battery test cycle, illustrating the voltage / time profile with identification of the recovery period and surface load removal stages; Figure 10 shows a voltage / battery time trace for a typical response to a complete test cycle for an "average" battery; Figure 11 shows a battery terminal voltage / time trace in a Micro-charge cycle with indication of the individual micro-charges applied by the FET and the corresponding voltage drops for both a good battery and a bad battery; Figure 12 shows the corresponding lines for a micro-charge cycle stored with corresponding lines for good and bad batteries; Figure 13 shows a voltage / time trace for overvoltage in recovery of a microload stage to be consumed (microload) that provides one of the classical analysable parameters for the purpose of the present invention; and Figures 14, 15, 16A and 16B show simple algorithms in text and flow diagram format for use in the t aspect of the invention, to identify the relevant 3 and 4 battery categories without the use of a neural network. FIRST APPEARANCE The embodiment of the invention will now be described for the purposes of technical description of the present invention and reference is made to our specification published prior to EP 0 762 135 A2 and herein we incorporate the entire description of that prior specification in the present application. This method is based on the description of the previous specification EP '135 A2 with the modification indicated below. Similarly, the embodiment may be based on the following description (in relation to the second aspect of the invention) which is based on Figures 1 to 8 of the accompanying drawings, and likewise using the Kohonen network described below. The embodiment of this aspect of the present invention uses a Kohonen self-organization map instead of the multi-layer Perceptron network described in our previous specification. Instead of the waveform analysis performed in our prior art, the mode uses readings during transient battery charging or discharging at certain predetermined time intervals. In test routines, instead of the period of 50 to 500 and preferably 150 to 300 milliseconds for transient discharge (in EP'135A2), this modality has used periods in the range of up to 500 milliseconds and preferably 10 to 20 milliseconds In order to avoid unacceptable heat dissipation. The apparatus includes means for generating test data to subject a battery to a test routine based on these periods of load to be consumed transiently or transiently stored load. Kohonen's self-organization map is adapted to discover natural groupings within data presented to its network and is able to learn to respond to different parts of the data feed signal. The network consists of neurons, which are initially randomly assigned "weighting" vectors in preparation for the training routine. Neurons can be visualized as arranged in a two-dimensional grid. The weighting vectors are provided in the same format as the feed signal. A neuron compares its current weight vector with the content of the feed signal and the number of correspondences made between these two vectors is the output of the neuron. The neuron that responds most strongly to a feeding signal is identified and the neighborhood neurons around this neuron are identified and the weights of these neurons are adjusted in order to give a stronger response to the relevant feeding signal. This process of weight adjustment is known as the training process. We have discovered that an excessive training of the neurons leads to the network of a positive response only to signals identical to those in which it was trained and this effect is clearly undesirable. A neuron can train on more than one signal at a time and a network can train itself to respond in other areas of its grid to power signals that are significantly different with a certain one.

In the modalities, data obtained during the operation of a conventional battery tester and which results from sampling at several predetermined intervals, then presented to a self-organizing Kohonen network trained for analysis. By reference to the established data patterns, the network was able to classify the battery. In the modality, the total training data package consists of approximately 550 battery tests, each comprising approximately 20 parameters. These parameters, including a three-phase test procedure, include first surface load removal, second load test and third, heavy load test, followed by a recovery period. Other data provided to the network include Cold Cranking Amps (CCA) regimes and calculated internal battery resistance. This data block represents the basis for providing the Kohonen network with its modest training data feed requirements, for example a data sample of 60 randomly selected battery tests from the previous population and each test comprises 20 data samples per each battery status. It is expected that it will be possible to cause the system to be able to distinguish between different modes of battery failure within the "bad-unserviceable" battery classification mentioned above. In this way, for example, a battery can be considered as exhausted after a long service phase, due to parasitic and irreversible gas evolution reactions, or that the battery may have been damaged in some way. The different electrical characteristics of this battery may well be capable of classification by the Kohonen network. In the mode, the system rating output provides a signal that is used to control a battery charging system, which responds to the battery status. The classification signal is fed to a neural network that controls a charging system speed of the battery in accordance. SECOND ASPECT The second aspect of the invention provides a method and apparatus for testing and / or classification of batteries, wherein a micro cycle of one or more transient stored loads and / or transient loads to be applied to a battery in sequence and data Regarding the battery reaction to the microcycle are analyzed by a neural network or an adaptive algorithm to allow the classification of the battery.

An additional characteristic of the modality is that the analysis stage is performed by the neural network or adaptive algorithm in relation to one or more of the following characteristics of the battery, i.e. impedance, voltage bounce, voltage peaks, voltage channels , profile or shape of voltage, load absorption and start and end voltages. Figure 1 shows a sequence of transient test cycles (cycle 1, cycle 2) with intercalated periods of battery recovery and chasurface removal of the battery. In Figure 2 and Figure 3, the voltages applied to the field effect transistors (FETs) in Figures 4 and 5 are illustrated. As shown in Figure 2, the microloan sequence to be consumed, the time (T2) between successive micro tests (Vx, V2, V3, etc.) is greater than (or equal to) the duration (Tx) of each test. The voltage v is less than V2 is less than V3. Total test time (To) is greater than or equal to 40 seconds. For the microchacycle, the voltage (V4) in the field effect transistor is constant. Figures 4 and 5 show the FETs in the microload circuits to be consumed (microload) and microload stored (microcha that act as switches for the application of chato consume and transient stored chaunder the control of a dynamic adjustment circuit seen in Figure 6. In circuits 10 and 12 of Figures 4 and 5, the battery under test is illustrated at 14 on FETs at 16 and 18. The microcharging voltage source is illustrated at 20. Turning now to system 22 shown at Figure 6, the micro-charging circuit 10 is generally constructed as illustrated in Figure 4 with battery terminals 24, 26 connected to the power MOSFET field effect transistor switch 16 which is controlled by a dynamic adjustment circuit 28 which responds to a micro controller 30. The data acquisition system 32 serves to provide an interface for data acquisition and transmission to the analysis function that is part of the microcontroller 30 and is illustrated in more detail in Figure 8. In general, the MOSFET of power or energy applies a transient load to the battery terminals 24, 26 for a selected period in the range of for example 20 to 100 micro seconds. The reaction of the battery to this transient charge and its recovery are examined and analyzed.

Typical results are shown for example in Figure 11. The power MOSFET typically has a maximum size of 30 millimeters by 20 millimeters and its average current draw is small and thus causes little damage to the vehicle's battery. The current discharge represented by the MOSFET can be altered by changing the gate voltage regulated by the micro controller 30 through adjustment circuits 28. The installation for varying the extracted current provides the basis for determining the internal impedance of the battery. In a similar way, other characteristics of the battery are easily derivable from the voltage profiles obtained in the method of the invention as described in the embodiments, including various aspects of the voltage itself and including voltage bounce, peaks and channels. voltage, load absorption and start and end voltages. It is a feature of this aspect of the present invention that classification can be effected by profile analysis in relation to one or more of these battery characteristics. In several cases, a measure of the magnitude of the relevant quantity is given by examining a portion or dimension of the profile of the voltage trace produced during the test. The use of microload to consume (microload) or microload stored (microcharge) transient in this mode, eliminates the requirement of certain prior techniques that use large carbon stacks, where the mode test apparatus can be in the format of a manual or otherwise mobile unit that can be battery powered or energized from a simple household electricity supply. One advantage of the microcharging technique that we have discovered is that it is effective in removing surface charges from the battery, which would otherwise adversely affect the accuracy of the test. Turning now to the micro-charging circuit 34 of Figure 7, this differs somewhat from the simplified system shown in Figure 5. Battery terminals 24, 26 are coupled to a data acquisition system and a microcontroller 38. A pair of power MOSFETs 40, 42 act as counters to allow the charging and discharging of a capacitor 44, which introduces a very small charge packet into the test battery through terminals 24, 26 very quickly, to provide transient battery charge pulses. The voltage source for the micro-loads is identified (as in Figure 5) at 20 and can be provided by battery charge theft or a backup of an inspection and maintenance station. Figure 12 shows examples of battery response to the microcharging cycle. The ease with which the battery accepts a charge pulse provides a measure of its condition and also the reaction of the battery after the pulse. Internal impedance of the battery and battery conditions in general in this way are accessible by use of this technique. However, the method can be combined with the microcharging method as a more effective total means of battery testing, and a combined system is illustrated in Figure 8. In Figure 8, systems and components corresponding to those described above in relation to the Prior figures are listed in accordance and will not be re-described except as necessary. Additional systems and components shown in Figure 8 include signal conditioning and multi-fold and analog-to-digital circuits 46, 48 and 50 respectively, of which the multiplexer 48 is coupled to a temperature thermistor 52. A unit of power supply 54, based for example on rechargeable batteries, provides a compact power source for the portable system 56 as a whole of Figure 8. The microcontroller identified at 30, 38 (for correlation with the microcontrollers of the previously described circuits (it incorporates a digital interface unit 58, a data storage function 60 and an algorithmic function 62 that provide the stepped analytical function for the electrical parameter feeds received by the interface 58 at 64, 66, 68 and 70. An additional connection 72 to interface 58 provides a user power and control function from a keyboard or other control power device 74. A display device 76 provides direct indication of the recognized battery category, providing a direct indication of the relevant selected based on the waveform resulting from the analysis of the microcharge sequence (microcharge) / microloan to be consumed (microload) applied to the battery. in 62, which is provided by an adaptive or advanced algorithm or a neural network in this embodiment, by the use of a microcycle of charges to be consumed and / or transient stored charges applied to a battery in sequence in accordance with this aspect of the invention, the advantage is provided (with respect to our specification EP '135A previously mentioned ) that the convenience and cost-effectiveness of the system, particularly for use in a portable or manual application, is unexpectedly advanced in such a proportion as a relative sophisticated analysis of a battery in a relevant one of a multiplicity of categories by the use of an item Manual equipment operated easily, becomes a realistic option in a proportion that to date had not been the case. The data storage function 16 cooperates with the neural / algorithmic network function 52, to allow a correlation of battery condition categories in wave form within the analysis sub-system 30, 38. The implementation can be carried out in a system Compatible PC or custom physical equipment, which provides the required digital signal processing functions, for example as that available from Texas Instruments, as noted in the EP'135 A2 specification mentioned above. The information mentioned in algorithms for the purpose is available in the publication "Neural Networks in C ++" (Neural networks in C ++) by Adam Blum published by John Wiley & amp;; Sons. Simplified algorithms are employed in the third aspect of the invention and are illustrated in Figures 14, 15, 16A and 16B in format of text and flow diagrams. Figure 9 illustrates results obtained with the system of Figure 8 for a complete battery test cycle. Figure 9 is a voltage / time trace illustrating a cycle of three transient micro-loads (microloads) 78, 80, 82 separated by two recovery periods 84, 85 followed by a transient microcharge (microcharge) 88. The voltage slope between Transient test stages indicate changes including removal of battery surface charge at 84 between the first two micro-loads. The time interval between the test stages is indicated on a linear scale by the location of the relevant traces on that scale, the micro-load stage 88 closely followed by the third micro-load. Figure 10 shows in a voltage / time trace, the current results obtained for a battery in "average" condition. The four micro-charging stages are indicated by the same reference numbers as in Figure 9. The profiles are characteristic of the battery classification, and taken in combination with the current values of the pre- and post-test voltages detected provide a means characteristic to identify the specific classification that is applicable. Turning now to the graphical data shown in Figures 11 and 12, these show the voltage / time plots not only for the FET gate or control voltages but also the battery responses for good and bad batteries. Figure 11 shows the result of micro-charge (microload) and Figure 12 shows the result of micro-charge (microcharge). Figure 11 shows the effect of 5 microloader cycles, which are achieved within something less than 0.45 seconds (for all five) with intermediate recovery periods in approximately 3/4 of .1 second ie .075 seconds. Each transient individual charge has a duration of approximately 1/4 of .1 second ie .025 second. The gate voltages for the tests show an increase of approximately 4 volts to approximately 6 volts from tests A to C and then remain constant. The five corresponding voltage drops for the trace 90 of a bad battery show characteristic significant drops in A and B with relatively corresponding stable state profiles in C, D and E of which the amplitude and sawtooth profiles are characteristic. In the trace 92 of a corresponding good battery, the substantially reduced voltage sags and the correspondingly modified profiles of the subsidence and intermediate voltages provide the clear basis for a corresponding classification of compliance.

In Figure 12, which corresponds to Figure 11, but shows a sequence of three micro-charge cycles (microcharge) or stages labeled F, G and H, the indications of the micro-load parameters in terms of FET gate voltages and transient time intervals correspond very much to those of Figure 11 and therefore do not require further comment. The battery responses in Figure 12 for a bad battery 90 and a good battery 92 are easily distinguished in terms of the resulting battery voltage staggering markedly improved in the case of the bad battery 90 against the modest or almost negligible voltage shift corresponding to the good battery 92. Figure 13 shows on a much shorter time scale the voltage / time trace during the short recovery period after a micro-charging stage. It can be seen that the surge profile 94 has a sharp profile with well-defined sawtooth at its leading edge 96. This is followed by a relatively uncharacterized and steeper rearward edge 98 from the peak 100, followed by an oscillation profile at 102 before the voltage rises to the level 104 corresponding to the recovery of the microcharging level 106. It will be understood that this profile provides a basis for a clear battery classification separation caused by profile changes in the various identified areas resulting from differences in battery status of the type that is apparent from Figures 11 and 12. THIRD ASPECT This mode represents a simplification of the modality described above (second aspect) wherein the advanced / neural network algorithm 62 of the microcontroller for data analysis 30, 38 is replaced by a relatively simple algorithm that is adapted to identify a particular of two or more condition classifications of batteries that exhibit a correlation with the waveform profile that resulted a of the stages of microcharge stored (microcharge) or microload to consume (microload) discussed above in relation to the preceding modalities. For this purpose, this modality examines voltage levels at characteristic points (previously determined) in the microloan voltage profile to be consumed (microload) or microcharged charge (microcharge) by reference to the time interval from the start of the test. Illustrative algorithms for the purpose of identifying the relevant three or four categories of batteries, are established in Figures 14, 15, 16A and 16B and represent relatively undeveloped stages in the technical development of the algorithms in relation to voltage parameters detected with battery condition categories. Specifically, based on the data storage function 60 comprising the results of battery tests in the relevant limited number of categories or classifications applicable to this simplified apparatus, we have found that a reliable level of accurate classification can be achieved acceptably, without the sophistication and the attenuated cost inherent in the use of neural networks, it is not even necessary to use the advance / adaptive algorithms designed for the modality of Figure 8. The operation of this simple modality generally proceeds according to the matters described above, but with the simplification represented by the use of a reduced number of categories of batteries such as simply "good / good discharged / bad".

Claims (35)

1. CLAIMS 1.- A method for determining the condition / classification of automotive and / or other batteries, according to one or more parameters detected thereof, which comprises the steps of: a.) Applying a transient electrical test parameter of the terminals of a battery to be tested; b) use a characteristic of the battery reaction to the test parameter as a basis for the determination; c) the method further comprises the step of using the battery reaction with the transient test parameter to provide a measure of the condition of the battery by analyzing the shape or waveform profile of an electrical value representative of the reaction from the battery to the test parameter; characterized in that: d) the step of analyzing the waveform profile of the electrical value of the battery reaction with the test parameter is performed without using a neural network but by means of an algorithm adapted to identify a corresponding one of two or more 'classifications of battery condition, and the corresponding classification exhibits a correlation with voltage aspects defined with time of the waveform profile of the electrical value.
2. 2. - Method for determining the condition / classification of automotive batteries and others according to one or more parameters detected thereof, comprising the steps of analyzing the shape or waveform profile of an electrical value of the reaction of a battery to a test parameter, without employing a neural network and using an algorithm adapted to identify a corresponding or two or more battery condition ratings and the corresponding classification exhibits a correlation with voltage aspect profile waveform profile of the electrical value or a portion or dimension thereof.
3. 3. - The method according to claim 1 or claim 2, characterized in that the electrical value representative of the battery reaction to the transient test parameter comprises one or more of the following characteristics, ie impedance, voltage bounce, spikes of voltage, voltage channels, shape or voltage profile, load absorption and start and end voltages.
4. 4. - The method according to any of the preceding claims, characterized in that the analysis stage is adapted to relate the value of the electrical value to two or more ranges of the values that are representative of two or more corresponding condition classifications of the battery.
5. 5. - The method according to any of the preceding claims, characterized in that the transient test parameter is applied for one to one hundred milliseconds.
6. 6. - The method according to any of the preceding claims, characterized in that the transient test parameter is repeated at intervals of two to one hundred times the duration of the test parameter itself.
7. 7. Apparatus for determining the condition / classification of automotive batteries and others according to one or more parameters detected thereof, comprising: a) means for applying a transient electrical test parameter of the terminals of a battery to be tested; b) analysis means adapted to use a characteristic of the reaction of the battery to the test parameter as a basis for the determination; c) the analysis means are adapted to use the reaction of the battery to the transient test parameter, to provide a measure of the condition of the battery by analyzing the shape or waveform profile of an electrical value representative of the reaction from the battery to the test parameter; characterized by: d) the means of analysis do not comprise a neural network and are adapted to carry out the analysis by means of an algorithm adapted to identify a corresponding one of two or more battery condition classifications, of which the classification -receives a correlation with aspects of voltage defined in time of the waveform profile of the electrical value.
8. 8. - Apparatus for examining the condition / classification of a car battery or other characterized by means of analysis adapted to analyze the waveform profile of an electrical profile of a reaction -of a battery with a test parameter, the means of analysis do not comprise a neural network and employ an algorithm to identify a classification of battery conditions that exhibits a correlation with voltage aspects of a detected waveform profile.
9. 9. - The apparatus according to claim 7 or claim 8, characterized in that the analysis means are adapted to carry out the analysis by reference to one or more of the following characteristics of battery ie impedance, voltage bounce, peaks of voltage, voltage channels, shape or voltage profile, load reception and start and end voltages.
10. 10. The apparatus according to any of claims 7 to 9, characterized in that the means of analysis are adapted to relate the value of the electric value with two or more ranges of the values that are representative of a corresponding one of two or more classifications. of battery condition.
11. 11. The compliance device according to claims 7 to 10, characterized in that the apparatus is adapted to apply the transient test parameter from one to one hundred milliseconds.
12. 12. - The apparatus according to any of claims 7 to 11, characterized in that the apparatus is adapted to repeat the transient test parameter at intervals of two to one hundred times the duration of the test parameter.
13. 13. - Method for producing the battery test apparatus for classification of automotive batteries and others, according to one or more parameters detected thereof, the method comprising: a) providing data feed coupling means adapted for attach a battery with data classification means; b) providing data classification means for coupling by means of coupling with the battery, such that the battery provides a data source enabling classification of the battery state, - c) the classification means provided are comprised of: a neural network and the neural network is adapted to carry out a stage of classification of the compliance data; characterized by: d) providing the neural network in the form of a self-organized software network (software), such as a self-organized Kohonen network; e) the method further comprises causing the self-organized software network to carry out (or rely on the software (software) that has been carried out) a controlled degree of training based on a feed comprising a representative sample of battery data, the representative sample of battery data comprises a plurality of battery parameters and selected from the group including voltage, current, internal resistance, capacitance / surface charge and thermal parameters; and f) the method further comprises providing the battery test apparatus with means of generating test data adapted to subject a given battery to test, to test routines to generate test data for the means of classification through the means of coupling, these test data are related to the representative sample of battery data where the software network was trained, whereby the test data can be easily classified by the data classification means.
14. 14. - Method for producing a battery test apparatus for classification of automotive batteries and others, and other electrical systems according to one or more parameters detected thereof, the method comprises classification means comprising a neural network self-organizing and provoking that the same is submitted (or is based on software (software) that has submitted) to a controlled degree of its training in representative data, the representative data comprise a plurality of battery parameters, the method further comprises providing means of generating adapted test data to generate related battery data from a battery under test.
15. 15. - The method according to claim 13 to claim 14, characterized by employing as the neural network a network of self-organizing software (software) Kohonen.
16. 16. The method according to any of claims 13 to 15, characterized in that the representative sample of battery data comprises data that is obtained from tests where a load to be consumed (load) transient or a transient charge (charge) is applied to a battery.
17. 17. - The method according to claim 16 characterized in that one or more loads to be consumed (loads) or transient charges (charges) are applied to the battery in sequence, the sequence comprises one or more load cycles to be consumed ( loads) or charges stored (charges) at transient intervals, the cycles of loads to be consumed (loads) or charges stored (charges) and the intervals are up to 100 milliseconds in duration.
18. 18. The method according to any of the preceding claims 13 to 17, characterized by using the self-organizing network to identify data groupings within the data of the battery.
19. 19. - The method according to any of the preceding claims 13 to 18, characterized by causing the neural network to perform classification of the battery as good, good discharged, or bad unusable.
20. 20. The method according to any of the preceding claims 13 to 19, characterized by the additional step of using the classification generated as a basis to control a charging system for the battery.
21. 21. The method according to claim 20 characterized by the step of controlling the battery charging system by means comprising a neural network.
22. 22. Apparatus for classification of automotive batteries and others, according to one or more electrical parameters detected thereof, the apparatus comprises: a) coupling means for data feed adapted to attach a battery to data classification means; b) data classification means adapted to be coupled by the coupling means to a battery, such that the battery provides a data source enabling classification of the state of the battery; c) the classification means comprise a neural network adapted to carry out a classification step in the compliance battery data; characterized by: d) the neural network comprises a self-organizing software network, such as a self-organizing Kohonen network; e) the self-organizing network has been submitted (or is based on software (software)) that has been subjected to a controlled degree of training based on the feeding of data comprising the representative sample of battery data, the sample representative of battery data comprises a plurality of battery parameters selected from the group including a voltage, current, internal resistance, load / surface capacitance and thermal parameters, the data provide a basis for the classification step required for the self-organizing network; and f) the apparatus further comprises means for generating test data adapted to subject a given battery to test, to test routines for generating test data for the classification means through the coupling means, this test data is related to the representative sample of the battery data, where the support network is trained, so that the test data can be easily classified by the data classification means.
23. 23. - Apparatus for classification of automotive batteries and other electrical systems according to one or more parameters detected thereof, the apparatus comprises classification means comprising a self-organizing neural network adapted to effect a controlled degree of training based on data of sample battery, comprising a plurality of battery parameters or other data, and the apparatus further comprises means for generating test data to subject a test battery to routines that generate data related to training data for the network of self organization
24. 24. The apparatus according to claim 22 or claim 23, characterized in that the neural network comprises a self-organizing Kohonen network.
25. 25. - The apparatus according to any of claims 22 to 24, characterized in that the data sample representative of battery data comprises data obtained from tests in which a load to be consumed (load) transient or a stored load (charge) transient applies to a battery.
26. 26.- The apparatus according to claim 25, characterized by one or more transient loads (loads) or transient loads that have been applied in sequence, comprising one or more load cycles to be consumed ( loads) or charges to be stored (charges) transient at transient intervals, the cycles of loads to be consumed (loads) or loads to store transients and intervals are up to 100 milliseconds in duration.
27. 27. The apparatus according to any of claims 22 to 26, characterized in that the self-organizing network is adapted to identify data groupings within the data of the battery.
28. 28. The apparatus according to any of claims 22 to 27, characterized in that the classification means are adapted to classify the battery as good, good but discharged or bad and unusable.
29. 29. - The apparatus according to any of claims 22 to 28, characterized by a battery charging system and control means for the battery charging system adapted to effect a degree of control in accordance with the generated battery classification.
30. 30. The apparatus according to claim 29, characterized in that the control means of the battery charging system comprises a neural network.
31. 31.- Method for classification of automotive batteries and others in accordance with one or more parameters detected, the method comprises providing an apparatus according to claim 10 or 11, characterized by the step of causing the means of generating test data submit the battery to test routines generating test data for the sorting media through the coupling means, the test data is related to the representative sample of battery data where the software network was trained , and the method comprises causing the classification means to classify the test data.
32. 32.- Method for determining the condition / classification of automotive batteries and others according to one or more parameters detected thereof, which comprises the steps of: a) applying a transient electrical test parameter to the terminals of a battery to be tested; b) use a characteristic of the battery reaction to the test parameter as a basis for the determination; c) the method further comprises the step of using the battery reaction to the transient test parameter, to provide a measure of the battery condition by analyzing the shape or waveform profile of an electrical value representative of the reaction of the battery to the test parameter; characterized by: d) applying at least one additional transient electrical test parameter to the battery to be tested; e) one or more transient electrical test parameters form a micro cycle of two or more loads to be stored (charges) transient and / or loads to be consumed (loads) transients applied to the battery in sequence; f) analyze data referring to the reaction of the battery to the micro cycle by the neural network and / or an algorithm to allow classification of the battery.
33. 33.- Method according to claim 32, characterized in that the analysis step to perform in relation to one or more of the following characteristics of the battery ie impedance, voltage bounce, voltage peaks, voltage channels, shape or voltage profile, load reception, start and end voltages.
34. 34.- Apparatus for determining the condition / classification of automotive batteries and others, according to one or more parameters detected thereof, which comprises-: a) means to apply a transient electrical test parameter to the terminals of a battery , Approve; b) analysis means adapted to use a characteristic of the reaction of the battery to the test parameter as a basis for the determination; c) the analysis means are further adapted to utilize the reaction of the battery to the transient test parameter to provide a measure of the battery condition by analysis of the waveform shape or profile of an electrical value representative of the reaction of the battery to the test parameter; characterized by: d) means for applying at least one additional transient electrical test parameter to the battery to be tested; e) the one and additional additional transient electrical test parameters that form a micro cycle of two or more loads to be stored (charges) and / or loads to be consumed (loads) transients applied to the battery in sequence; and f) the means of analysis are also adapted to analyze data relating to the reaction of the battery to the micro cycle by a neural network and / or an algorithm to allow classification of the battery.
35. 35. Apparatus according to claim 34, characterized in that the analysis means are adapted to carry out the analysis stages in relation to one or more of the following characteristics of the battery, ie impedance, voltage bounce, voltage peaks. , voltage channels, shape or voltage profile, load absorption and start and end voltages.