Estimating accuracy of a remaining useful life prediction model for a consumable using statistics based segmentation technique

• US 9,046,854 B2
• Filed: 06/27/2013
• Issued: 06/02/2015
• Est. Priority Date: 06/27/2013
• Status: Expired due to Fees

• Current Assignee: Xerox Corporation (Xerox Holdings Corp.)
• Original Assignee: Xerox Corporation (Xerox Holdings Corp.)
• Inventors: Yang, Ming, Stumbo, William K., Foley, Diane M., Li, Juan
• Primary Examiner(s): Lindsay, Jr., Walter L
• Assistant Examiner(s): Wenderoth, Frederick
• Application Number:
  US13/929,760
• Publication Number:
  US 20150003847A1
• Time in Patent Office: 705 Days
• Field of Search: 399/27
• US Class Current: 1/1
• CPC Class Codes: G03G 15/0856 Detection or control means for the developer level
• CPC Class Codes: G03G 15/556 for toner consumption, e.g. pixel counting, toner coverage detection or toner density measurement

Abstract

An apparatus and method of predicting the end of life of a consumable. A basic weighted least squares algorithm has been extended and augmented to compensate for observed common consumable/printer behavior. The system uses consumable usage data (such as toner level) acquired from the device to predict the current and future consumable level and to predict the remaining life. The apparatus and method monitors the consumable'"'"'s usage and updates the prediction so that when the predicted remaining life matches a preset threshold, it automatically triggers an order placement event to ship product to customer.

Claims

1. A method to rapidly detect anomalies in measurement and/or usage which would prevent accurate estimates of supply level and of remaining useful life of a consumable in an image reproduction device, the method comprising:

selectively segmenting consumables into groups which show statistically different levels of prediction accuracy by prediction models when a prediction was given by a prediction model applied to a historic consumable usage dataset;

wherein segmenting the consumables into groups comprises determining a mean and standard deviation or variance (Vur) of a usage rate of the consumable;

wherein the prediction accuracy is a prediction error based on a difference between a predicted target day (PTD) as predicted by the prediction models and an actual target day(ATD);

applying statistical metrics to the groups which show statistically different levels of prediction accuracy for a given time window;

wherein the statistical metrics is a percentage of consumables within a predetermined range from the actual target day (ATD);

determining, from the statistical metrics of the prediction accuracy, if an employed prediction model is likely to provide an inaccurate estimate of the remaining useful life of the consumable;

wherein if it is determined that the employed prediction model is likely to provide an inaccurate estimate, then sending a message suggesting changing the employed prediction model.

2. The method in accordance to claim 1, wherein segmenting the consumables into groups comprises determining a correlation coefficients (K) between a usage of the consumable and the output of the image reproduction device.

3. The method in accordance to claim 1, wherein segmenting the consumables into groups comprises determining a first moment and a standard deviation of a predicted time when the consumable reaches end of life.

4. The method in accordance to claim 1, wherein segmenting the consumables into groups comprises determining at least one of a first moment and a standard deviation of a predicted time when the consumable reaches end of life, a correlation coefficients (K) between a usage of the consumable and the output of the image reproduction device, or a mean and standard deviation or variance (Vur) of a usage rate of the consumable.

6. The method in accordance to claim 1, wherein segmenting the consumables into groups comprises determining a correlation coefficients (K) between a usage of the consumable and the output of an image reproduction device.

5. A non-transitory computer readable medium encoded with computer executable instructions, which when accessed, causes a machine to perform operations comprising:

selectively segmenting consumables into groups which show statistically different levels of prediction accuracy by prediction models when a prediction was given by a prediction model applied to a historic consumable usage dataset;

wherein segmenting the consumables into groups comprises determining a mean and standard deviation or variance (Vur) of a usage rate of the consumable;

wherein the prediction accuracy is a prediction error based on a difference between a predicted target day (PTD) as predicted by the prediction models and an actual target day(ATD);

applying statistical metrics to the groups which show statistically different levels of prediction accuracy a given time window;

wherein the statistical metrics is a percentage of consumables within a predetermined range from the actual target day (ATD);

determining, from the statistical metrics of the prediction accuracy, if an employed prediction model is likely to provide an inaccurate estimate of the remaining useful life of the consumable;

wherein if it is determined that the employed prediction model is likely to provide an inaccurate estimate, then sending a message suggesting changing the employed prediction model.

7. The non-transitory computer readable medium encoded with computer executable instructions in accordance to claim 5, wherein segmenting the consumables into groups comprises determining a first moment and a standard deviation of a predicted time when the consumable reaches end of life.

8. The non-transitory computer readable medium encoded with computer executable instructions in accordance to claim 5, wherein segmenting the consumables into groups comprises determining at least one of a first moment and a standard deviation of a predicted time when the consumable reaches end of life, a correlation coefficients (K) between a usage of the consumable and the output of an image reproduction device, or a mean and standard deviation or variance (Vur) of a usage rate of the consumable.

9. based consumable management platform, comprising:

a database operable to store information associated with at least one replaceable toner cartridge, wherein the stored information includes daily toner cartridge level data from replaceable cartridges, wherein the database is further operable to rapidly detect anomalies in measurement and/or usage which would prevent accurate estimates of a remaining life of a replaceable cartridge and to alert a user or a service/maintenance provider that an anomaly has been detected by;

selectively segmenting replaceable toner cartridges into groups which show statistically different levels of prediction accuracy when a prediction was given by a prediction model applied to a historic replaceable toner cartridge usage dataset;

wherein segmenting the consumables into groups comprises determining a mean and standard deviation or variance (Vur) of a usage rate of the consumable;

wherein segmenting the replaceable toner cartridges into groups comprise determining a correlation coefficients (K) between a usage of the replaceable toner cartridge and the output of an image reproduction device;

applying statistical metrics to the groups which show statistically different levels of prediction accuracy to identify when it is probable that a remaining life prediction models will not yield accurate results for a given time window, so that a different prediction model or an alternative shipment triggering algorithm can be employed;

wherein the statistical metrics is a percentage of consumables within a predetermined range from an actual target day (ATD);

determining, from the statistical metrics of the prediction accuracy, if an employed prediction model is likely to provide an inaccurate estimate of the remaining useful life of the consumable;

wherein if it is determined that the employed prediction model is likely to provide an inaccurate estimate, then sending a message suggesting changing the employed prediction model.

10. The dynamic cloud based consumable management platform in accordance to claim 9, wherein segmenting the replaceable toner cartridges into groups comprises determining at least one of a first moment and a standard deviation of a predicted time when the replaceable toner cartridge reaches end of life, a correlation coefficients (K) between a usage of the replaceable toner cartridge and the output of an image reproduction device, or a mean and standard deviation or variance (Vur) of a usage rate of the replaceable toner cartridge.

11. The dynamic cloud based consumable management platform in accordance to claim 9, wherein the prediction accuracy is a prediction error based on a difference between a predicted target day (PTD) as predicted by the prediction model and the actual target day (ATD).

12. The dynamic cloud based consumable management platform in accordance to claim 11, wherein the statistical metrics is a percentage of replaceable toner cartridges within a predetermined range from the predicted target day.